2
0
mirror of synced 2025-02-23 14:58:12 +00:00
mobile/bind/genjava.go

1125 lines
31 KiB
Go
Raw Normal View History

// Copyright 2014 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package bind
import (
"fmt"
"go/constant"
"go/types"
"math"
"strings"
)
// TODO(crawshaw): disallow basic android java type names in exported symbols.
// TODO(crawshaw): consider introducing Java functions for casting to and from interfaces at runtime.
type javaGen struct {
// javaPkg is the custom name of the Java pkg that contains the generated classes. If empty,
// use a package name generated from the Go package name.
javaPkg string
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
*generator
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *javaGen) genStruct(obj *types.TypeName, T *types.Struct) {
fields := exportedFields(T)
methods := exportedMethodSet(types.NewPointer(obj.Type()))
var impls []string
pT := types.NewPointer(obj.Type())
for _, iface := range g.allIntf {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
if types.AssignableTo(pT, iface.obj.Type()) {
n := iface.obj.Name()
if p := iface.obj.Pkg(); p != g.pkg {
n = fmt.Sprintf("%s.%s.%s", g.javaPkgName(p), className(p), n)
}
impls = append(impls, n)
}
}
g.Printf("public static final class %s extends Seq.Proxy", obj.Name())
if len(impls) > 0 {
g.Printf(" implements %s", strings.Join(impls, ", "))
}
g.Printf(" {\n")
g.Indent()
n := obj.Name()
g.Printf("private %s(go.Seq.Ref ref) { super(ref); }\n\n", n)
for _, f := range fields {
if t := f.Type(); !g.isSupported(t) {
g.Printf("// skipped field %s.%s with unsupported type: %T\n\n", n, f.Name(), t)
continue
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("public final native %s get%s();\n", g.javaType(f.Type()), f.Name())
g.Printf("public final native void set%s(%s v);\n\n", f.Name(), g.javaType(f.Type()))
}
var isStringer bool
for _, m := range methods {
if !g.isSigSupported(m.Type()) {
g.Printf("// skipped method %s.%s with unsupported parameter or return types\n\n", obj.Name(), m.Name())
continue
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genFuncSignature(m, false, false)
t := m.Type().(*types.Signature)
isStringer = isStringer || (m.Name() == "String" && t.Params().Len() == 0 && t.Results().Len() == 1 &&
types.Identical(t.Results().At(0).Type(), types.Typ[types.String]))
}
g.Printf("@Override public boolean equals(Object o) {\n")
g.Indent()
g.Printf("if (o == null || !(o instanceof %s)) {\n return false;\n}\n", n)
g.Printf("%s that = (%s)o;\n", n, n)
for _, f := range fields {
if t := f.Type(); !g.isSupported(t) {
g.Printf("// skipped field %s.%s with unsupported type: %T\n\n", n, f.Name(), t)
continue
}
nf := f.Name()
g.Printf("%s this%s = get%s();\n", g.javaType(f.Type()), nf, nf)
g.Printf("%s that%s = that.get%s();\n", g.javaType(f.Type()), nf, nf)
if isJavaPrimitive(f.Type()) {
g.Printf("if (this%s != that%s) {\n return false;\n}\n", nf, nf)
} else {
g.Printf("if (this%s == null) {\n", nf)
g.Indent()
g.Printf("if (that%s != null) {\n return false;\n}\n", nf)
g.Outdent()
g.Printf("} else if (!this%s.equals(that%s)) {\n return false;\n}\n", nf, nf)
}
}
g.Printf("return true;\n")
g.Outdent()
g.Printf("}\n\n")
g.Printf("@Override public int hashCode() {\n")
g.Printf(" return java.util.Arrays.hashCode(new Object[] {")
idx := 0
for _, f := range fields {
if t := f.Type(); !g.isSupported(t) {
continue
}
if idx > 0 {
g.Printf(", ")
}
idx++
g.Printf("get%s()", f.Name())
}
g.Printf("});\n")
g.Printf("}\n\n")
g.Printf("@Override public String toString() {\n")
g.Indent()
if isStringer {
g.Printf("return String();\n")
} else {
g.Printf("StringBuilder b = new StringBuilder();\n")
g.Printf(`b.append("%s").append("{");`, obj.Name())
g.Printf("\n")
for _, f := range fields {
if t := f.Type(); !g.isSupported(t) {
continue
}
n := f.Name()
g.Printf(`b.append("%s:").append(get%s()).append(",");`, n, n)
g.Printf("\n")
}
g.Printf(`return b.append("}").toString();`)
g.Printf("\n")
}
g.Outdent()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("}\n")
g.Outdent()
g.Printf("}\n\n")
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *javaGen) genInterface(iface interfaceInfo) {
var exts []string
numM := iface.t.NumMethods()
for _, other := range g.allIntf {
// Only extend interfaces with fewer methods to avoid circular references
if other.t.NumMethods() < numM && types.AssignableTo(iface.t, other.t) {
n := other.obj.Name()
if p := other.obj.Pkg(); p != g.pkg {
n = fmt.Sprintf("%s.%s.%s", g.javaPkgName(p), className(p), n)
}
exts = append(exts, n)
}
}
g.Printf("public interface %s", iface.obj.Name())
if len(exts) > 0 {
g.Printf(" extends %s", strings.Join(exts, ", "))
}
g.Printf(" {\n")
g.Indent()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
for _, m := range iface.summary.callable {
if !g.isSigSupported(m.Type()) {
g.Printf("// skipped method %s.%s with unsupported parameter or return types\n\n", iface.obj.Name(), m.Name())
continue
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genFuncSignature(m, false, true)
}
g.Outdent()
g.Printf("}\n")
g.Printf("\n")
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf(javaProxyPreamble, iface.obj.Name())
g.Indent()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
for _, m := range iface.summary.callable {
if !g.isSigSupported(m.Type()) {
g.Printf("// skipped method %s.%s with unsupported parameter or return types\n\n", iface.obj.Name(), m.Name())
continue
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genFuncSignature(m, false, false)
}
g.Outdent()
g.Printf("}\n\n")
}
func isJavaPrimitive(T types.Type) bool {
b, ok := T.(*types.Basic)
if !ok {
return false
}
switch b.Kind() {
case types.Bool, types.Uint8, types.Float32, types.Float64,
types.Int, types.Int8, types.Int16, types.Int32, types.Int64:
return true
}
return false
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
// jniType returns a string that can be used as a JNI type.
func (g *javaGen) jniType(T types.Type) string {
if isErrorType(T) {
// The error type is usually translated into an exception in
// Java, however the type can be exposed in other ways, such
// as an exported field.
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
return g.jniType(types.Typ[types.String])
}
switch T := T.(type) {
case *types.Basic:
switch T.Kind() {
case types.Bool, types.UntypedBool:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
return "jboolean"
case types.Int:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
return "jlong"
case types.Int8:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
return "jbyte"
case types.Int16:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
return "jshort"
case types.Int32, types.UntypedRune: // types.Rune
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
return "jint"
case types.Int64, types.UntypedInt:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
return "jlong"
case types.Uint8: // types.Byte
// TODO(crawshaw): Java bytes are signed, so this is
// questionable, but vital.
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
return "jbyte"
// TODO(crawshaw): case types.Uint, types.Uint16, types.Uint32, types.Uint64:
case types.Float32:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
return "jfloat"
case types.Float64, types.UntypedFloat:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
return "jdouble"
case types.String, types.UntypedString:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
return "jstring"
default:
g.errorf("unsupported basic type: %s", T)
return "TODO"
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
case *types.Slice:
return "jbyteArray"
case *types.Pointer:
if _, ok := T.Elem().(*types.Named); ok {
return g.jniType(T.Elem())
}
g.errorf("unsupported pointer to type: %s", T)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
case *types.Named:
return "jobject"
default:
g.errorf("unsupported jniType: %#+v, %s\n", T, T)
}
return "TODO"
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
func (g *javaGen) javaBasicType(T *types.Basic) string {
switch T.Kind() {
case types.Bool, types.UntypedBool:
return "boolean"
case types.Int:
return "long"
case types.Int8:
return "byte"
case types.Int16:
return "short"
case types.Int32, types.UntypedRune: // types.Rune
return "int"
case types.Int64, types.UntypedInt:
return "long"
case types.Uint8: // types.Byte
// TODO(crawshaw): Java bytes are signed, so this is
// questionable, but vital.
return "byte"
// TODO(crawshaw): case types.Uint, types.Uint16, types.Uint32, types.Uint64:
case types.Float32:
return "float"
case types.Float64, types.UntypedFloat:
return "double"
case types.String, types.UntypedString:
return "String"
default:
g.errorf("unsupported basic type: %s", T)
return "TODO"
}
}
// javaType returns a string that can be used as a Java type.
func (g *javaGen) javaType(T types.Type) string {
if isErrorType(T) {
// The error type is usually translated into an exception in
// Java, however the type can be exposed in other ways, such
// as an exported field.
return "String"
}
switch T := T.(type) {
case *types.Basic:
return g.javaBasicType(T)
case *types.Slice:
elem := g.javaType(T.Elem())
return elem + "[]"
case *types.Pointer:
if _, ok := T.Elem().(*types.Named); ok {
return g.javaType(T.Elem())
}
g.errorf("unsupported pointer to type: %s", T)
case *types.Named:
n := T.Obj()
nPkg := n.Pkg()
if !g.validPkg(nPkg) {
g.errorf("type %s is in %s, which is not bound", n.Name(), nPkg)
break
}
// TODO(crawshaw): more checking here
if nPkg != g.pkg {
return fmt.Sprintf("%s.%s.%s", g.javaPkgName(nPkg), className(nPkg), n.Name())
} else {
return n.Name()
}
default:
g.errorf("unsupported javaType: %#+v, %s\n", T, T)
}
return "TODO"
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *javaGen) genJNIFuncSignature(o *types.Func, sName string, proxy bool) {
sig := o.Type().(*types.Signature)
res := sig.Results()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
var ret string
switch res.Len() {
case 2:
ret = g.jniType(res.At(0).Type())
case 1:
if isErrorType(res.At(0).Type()) {
ret = "void"
} else {
ret = g.jniType(res.At(0).Type())
}
case 0:
ret = "void"
default:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.errorf("too many result values: %s", o)
return
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("JNIEXPORT %s JNICALL\n", ret)
g.Printf("Java_%s_%s", g.jniPkgName(), g.className())
if sName != "" {
// 0024 is the mangled form of $, for naming inner classes.
g.Printf("_00024")
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
if proxy {
g.Printf("proxy")
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
g.Printf("%s", sName)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
g.Printf("_%s(JNIEnv* env, ", o.Name())
if sName != "" {
g.Printf("jobject this")
} else {
g.Printf("jclass clazz")
}
params := sig.Params()
for i := 0; i < params.Len(); i++ {
g.Printf(", ")
v := sig.Params().At(i)
name := paramName(params, i)
jt := g.jniType(v.Type())
g.Printf("%s %s", jt, name)
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf(")")
}
func (g *javaGen) jniPkgName() string {
return strings.Replace(g.javaPkgName(g.pkg), ".", "_", -1)
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *javaGen) genFuncSignature(o *types.Func, static, header bool) {
sig := o.Type().(*types.Signature)
res := sig.Results()
var returnsError bool
var ret string
switch res.Len() {
case 2:
if !isErrorType(res.At(1).Type()) {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.errorf("second result value must be of type error: %s", o)
return
}
returnsError = true
ret = g.javaType(res.At(0).Type())
case 1:
if isErrorType(res.At(0).Type()) {
returnsError = true
ret = "void"
} else {
ret = g.javaType(res.At(0).Type())
}
case 0:
ret = "void"
default:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.errorf("too many result values: %s", o)
return
}
g.Printf("public ")
if static {
g.Printf("static ")
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
if !header {
g.Printf("native ")
}
oName := o.Name()
g.Printf("%s %s(", ret, oName)
params := sig.Params()
for i := 0; i < params.Len(); i++ {
if i > 0 {
g.Printf(", ")
}
v := sig.Params().At(i)
name := paramName(params, i)
jt := g.javaType(v.Type())
g.Printf("%s %s", jt, name)
}
g.Printf(")")
if returnsError {
g.Printf(" throws Exception")
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf(";\n")
}
func (g *javaGen) genVar(o *types.Var) {
if t := o.Type(); !g.isSupported(t) {
g.Printf("// skipped variable %s with unsupported type: %T\n\n", o.Name(), t)
return
}
jType := g.javaType(o.Type())
// setter
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("public static native void set%s(%s v);\n", o.Name(), jType)
// getter
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("public static native %s get%s();\n\n", jType, o.Name())
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *javaGen) genJavaToC(varName string, t types.Type, mode varMode) {
if isErrorType(t) {
g.genJavaToC(varName, types.Typ[types.String], mode)
return
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
switch t := t.(type) {
case *types.Basic:
switch t.Kind() {
case types.String:
mobile/bind: avoid intermediate []rune copy converting Java string to Go Converting a Go string to a string suitable use a specialized function, UTF16Encode, that can encode the string directly to a malloc'ed buffer. That way, only two copies are made when strings are passed from Go to Java; once for UTF-8 to UTF-16 encoding and once for the creation of the Java String. This CL implements the same optimization in the other direction, with a UTF-16 to UTF-8 decoder implemented in C. Unfortunately, while calling into a Go decoder also saves the extra copy, the Cgo overhead makes the calls much slower for short strings. To alleviate the risk of introducing decoding bugs, I've added the tests from the encoding/utf16 package to SeqTest. As a sideeffect, both Java and ObjC now always copy strings, regardless of the argument mode. The cpy argument can therefore be removed from the string conversion functions. Furthermore, the modeRetained and modeReturned modes can be collapsed into just one. While we're here, delete a leftover function from seq/strings.go that wasn't removed when the old seq buffers went away. Benchmarks, as compared with benchstat over 5 runs: name old time/op new time/op delta JavaStringShort 11.4µs ±13% 11.6µs ± 4% ~ (p=0.859 n=10+5) JavaStringShortDirect 19.5µs ± 9% 20.3µs ± 2% +3.68% (p=0.019 n=9+5) JavaStringLong 103µs ± 8% 24µs ± 4% -77.13% (p=0.001 n=9+5) JavaStringLongDirect 113µs ± 9% 32µs ± 7% -71.63% (p=0.001 n=9+5) JavaStringShortUnicode 11.1µs ±16% 10.7µs ± 5% ~ (p=0.190 n=9+5) JavaStringShortUnicodeDirect 19.6µs ± 7% 20.2µs ± 1% +2.78% (p=0.029 n=9+5) JavaStringLongUnicode 97.1µs ± 9% 28.0µs ± 5% -71.17% (p=0.001 n=9+5) JavaStringLongUnicodeDirect 105µs ±10% 34µs ± 5% -67.23% (p=0.002 n=8+5) JavaStringRetShort 14.2µs ± 2% 13.9µs ± 1% -2.15% (p=0.006 n=8+5) JavaStringRetShortDirect 20.8µs ± 2% 20.4µs ± 2% ~ (p=0.065 n=8+5) JavaStringRetLong 42.2µs ± 9% 42.4µs ± 3% ~ (p=0.190 n=9+5) JavaStringRetLongDirect 51.2µs ±21% 50.8µs ± 8% ~ (p=0.518 n=9+5) GoStringShort 23.4µs ± 7% 22.5µs ± 3% -3.55% (p=0.019 n=9+5) GoStringLong 51.9µs ± 9% 53.1µs ± 3% ~ (p=0.240 n=9+5) GoStringShortUnicode 24.2µs ± 6% 22.8µs ± 1% -5.54% (p=0.002 n=9+5) GoStringLongUnicode 58.6µs ± 8% 57.6µs ± 3% ~ (p=0.518 n=9+5) GoStringRetShort 27.6µs ± 1% 23.2µs ± 2% -15.87% (p=0.003 n=7+5) GoStringRetLong 129µs ±12% 33µs ± 2% -74.03% (p=0.001 n=10+5) Change-Id: Icb9481981493ffca8defed9fb80a9433d6048937 Reviewed-on: https://go-review.googlesource.com/20250 Reviewed-by: David Crawshaw <crawshaw@golang.org>
2016-03-04 17:59:04 +01:00
g.Printf("nstring _%s = go_seq_from_java_string(env, %s);\n", varName, varName)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
default:
g.Printf("%s _%s = (%s)%s;\n", g.cgoType(t), varName, g.cgoType(t), varName)
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
case *types.Slice:
switch e := t.Elem().(type) {
case *types.Basic:
switch e.Kind() {
case types.Uint8: // Byte.
mobile/bind: avoid intermediate []rune copy converting Java string to Go Converting a Go string to a string suitable use a specialized function, UTF16Encode, that can encode the string directly to a malloc'ed buffer. That way, only two copies are made when strings are passed from Go to Java; once for UTF-8 to UTF-16 encoding and once for the creation of the Java String. This CL implements the same optimization in the other direction, with a UTF-16 to UTF-8 decoder implemented in C. Unfortunately, while calling into a Go decoder also saves the extra copy, the Cgo overhead makes the calls much slower for short strings. To alleviate the risk of introducing decoding bugs, I've added the tests from the encoding/utf16 package to SeqTest. As a sideeffect, both Java and ObjC now always copy strings, regardless of the argument mode. The cpy argument can therefore be removed from the string conversion functions. Furthermore, the modeRetained and modeReturned modes can be collapsed into just one. While we're here, delete a leftover function from seq/strings.go that wasn't removed when the old seq buffers went away. Benchmarks, as compared with benchstat over 5 runs: name old time/op new time/op delta JavaStringShort 11.4µs ±13% 11.6µs ± 4% ~ (p=0.859 n=10+5) JavaStringShortDirect 19.5µs ± 9% 20.3µs ± 2% +3.68% (p=0.019 n=9+5) JavaStringLong 103µs ± 8% 24µs ± 4% -77.13% (p=0.001 n=9+5) JavaStringLongDirect 113µs ± 9% 32µs ± 7% -71.63% (p=0.001 n=9+5) JavaStringShortUnicode 11.1µs ±16% 10.7µs ± 5% ~ (p=0.190 n=9+5) JavaStringShortUnicodeDirect 19.6µs ± 7% 20.2µs ± 1% +2.78% (p=0.029 n=9+5) JavaStringLongUnicode 97.1µs ± 9% 28.0µs ± 5% -71.17% (p=0.001 n=9+5) JavaStringLongUnicodeDirect 105µs ±10% 34µs ± 5% -67.23% (p=0.002 n=8+5) JavaStringRetShort 14.2µs ± 2% 13.9µs ± 1% -2.15% (p=0.006 n=8+5) JavaStringRetShortDirect 20.8µs ± 2% 20.4µs ± 2% ~ (p=0.065 n=8+5) JavaStringRetLong 42.2µs ± 9% 42.4µs ± 3% ~ (p=0.190 n=9+5) JavaStringRetLongDirect 51.2µs ±21% 50.8µs ± 8% ~ (p=0.518 n=9+5) GoStringShort 23.4µs ± 7% 22.5µs ± 3% -3.55% (p=0.019 n=9+5) GoStringLong 51.9µs ± 9% 53.1µs ± 3% ~ (p=0.240 n=9+5) GoStringShortUnicode 24.2µs ± 6% 22.8µs ± 1% -5.54% (p=0.002 n=9+5) GoStringLongUnicode 58.6µs ± 8% 57.6µs ± 3% ~ (p=0.518 n=9+5) GoStringRetShort 27.6µs ± 1% 23.2µs ± 2% -15.87% (p=0.003 n=7+5) GoStringRetLong 129µs ±12% 33µs ± 2% -74.03% (p=0.001 n=10+5) Change-Id: Icb9481981493ffca8defed9fb80a9433d6048937 Reviewed-on: https://go-review.googlesource.com/20250 Reviewed-by: David Crawshaw <crawshaw@golang.org>
2016-03-04 17:59:04 +01:00
g.Printf("nbyteslice _%s = go_seq_from_java_bytearray(env, %s, %d);\n", varName, varName, g.toCFlag(mode == modeRetained))
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
default:
g.errorf("unsupported type: %s", t)
}
default:
g.errorf("unsupported type: %s", t)
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
case *types.Named:
switch u := t.Underlying().(type) {
case *types.Interface:
g.Printf("int32_t _%s = go_seq_to_refnum(env, %s);\n", varName, varName)
default:
g.errorf("unsupported named type: %s / %T", u, u)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
case *types.Pointer:
g.Printf("int32_t _%s = go_seq_to_refnum(env, %s);\n", varName, varName)
default:
g.Printf("%s _%s = (%s)%s;\n", g.cgoType(t), varName, g.cgoType(t), varName)
}
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *javaGen) genCToJava(toName, fromName string, t types.Type, mode varMode) {
if isErrorType(t) {
g.genCToJava(toName, fromName, types.Typ[types.String], mode)
return
}
switch t := t.(type) {
case *types.Basic:
switch t.Kind() {
case types.String:
g.Printf("jstring %s = go_seq_to_java_string(env, %s);\n", toName, fromName)
case types.Bool:
g.Printf("jboolean %s = %s ? JNI_TRUE : JNI_FALSE;\n", toName, fromName)
default:
g.Printf("%s %s = (%s)%s;\n", g.jniType(t), toName, g.jniType(t), fromName)
}
case *types.Slice:
switch e := t.Elem().(type) {
case *types.Basic:
switch e.Kind() {
case types.Uint8: // Byte.
mobile/bind: avoid intermediate []rune copy converting Java string to Go Converting a Go string to a string suitable use a specialized function, UTF16Encode, that can encode the string directly to a malloc'ed buffer. That way, only two copies are made when strings are passed from Go to Java; once for UTF-8 to UTF-16 encoding and once for the creation of the Java String. This CL implements the same optimization in the other direction, with a UTF-16 to UTF-8 decoder implemented in C. Unfortunately, while calling into a Go decoder also saves the extra copy, the Cgo overhead makes the calls much slower for short strings. To alleviate the risk of introducing decoding bugs, I've added the tests from the encoding/utf16 package to SeqTest. As a sideeffect, both Java and ObjC now always copy strings, regardless of the argument mode. The cpy argument can therefore be removed from the string conversion functions. Furthermore, the modeRetained and modeReturned modes can be collapsed into just one. While we're here, delete a leftover function from seq/strings.go that wasn't removed when the old seq buffers went away. Benchmarks, as compared with benchstat over 5 runs: name old time/op new time/op delta JavaStringShort 11.4µs ±13% 11.6µs ± 4% ~ (p=0.859 n=10+5) JavaStringShortDirect 19.5µs ± 9% 20.3µs ± 2% +3.68% (p=0.019 n=9+5) JavaStringLong 103µs ± 8% 24µs ± 4% -77.13% (p=0.001 n=9+5) JavaStringLongDirect 113µs ± 9% 32µs ± 7% -71.63% (p=0.001 n=9+5) JavaStringShortUnicode 11.1µs ±16% 10.7µs ± 5% ~ (p=0.190 n=9+5) JavaStringShortUnicodeDirect 19.6µs ± 7% 20.2µs ± 1% +2.78% (p=0.029 n=9+5) JavaStringLongUnicode 97.1µs ± 9% 28.0µs ± 5% -71.17% (p=0.001 n=9+5) JavaStringLongUnicodeDirect 105µs ±10% 34µs ± 5% -67.23% (p=0.002 n=8+5) JavaStringRetShort 14.2µs ± 2% 13.9µs ± 1% -2.15% (p=0.006 n=8+5) JavaStringRetShortDirect 20.8µs ± 2% 20.4µs ± 2% ~ (p=0.065 n=8+5) JavaStringRetLong 42.2µs ± 9% 42.4µs ± 3% ~ (p=0.190 n=9+5) JavaStringRetLongDirect 51.2µs ±21% 50.8µs ± 8% ~ (p=0.518 n=9+5) GoStringShort 23.4µs ± 7% 22.5µs ± 3% -3.55% (p=0.019 n=9+5) GoStringLong 51.9µs ± 9% 53.1µs ± 3% ~ (p=0.240 n=9+5) GoStringShortUnicode 24.2µs ± 6% 22.8µs ± 1% -5.54% (p=0.002 n=9+5) GoStringLongUnicode 58.6µs ± 8% 57.6µs ± 3% ~ (p=0.518 n=9+5) GoStringRetShort 27.6µs ± 1% 23.2µs ± 2% -15.87% (p=0.003 n=7+5) GoStringRetLong 129µs ±12% 33µs ± 2% -74.03% (p=0.001 n=10+5) Change-Id: Icb9481981493ffca8defed9fb80a9433d6048937 Reviewed-on: https://go-review.googlesource.com/20250 Reviewed-by: David Crawshaw <crawshaw@golang.org>
2016-03-04 17:59:04 +01:00
g.Printf("jbyteArray %s = go_seq_to_java_bytearray(env, %s, %d);\n", toName, fromName, g.toCFlag(mode == modeRetained))
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
default:
g.errorf("unsupported type: %s", t)
}
default:
g.errorf("unsupported type: %s", t)
}
case *types.Pointer:
// TODO(crawshaw): test *int
// TODO(crawshaw): test **Generator
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
switch t := t.Elem().(type) {
case *types.Named:
g.genFromRefnum(toName, fromName, t, t.Obj())
default:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.errorf("unsupported type %s", t)
}
case *types.Named:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
switch t.Underlying().(type) {
case *types.Interface, *types.Pointer:
g.genFromRefnum(toName, fromName, t, t.Obj())
default:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.errorf("unsupported, direct named type %s", t)
}
default:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("%s %s = (%s)%s;\n", g.jniType(t), toName, g.jniType(t), fromName)
}
}
func (g *javaGen) genFromRefnum(toName, fromName string, t types.Type, o *types.TypeName) {
oPkg := o.Pkg()
if !g.validPkg(oPkg) {
g.errorf("type %s is defined in package %s, which is not bound", t, oPkg)
return
}
p := pkgPrefix(oPkg)
g.Printf("jobject %s = go_seq_from_refnum(env, %s, proxy_class_%s_%s, proxy_class_%s_%s_cons);\n", toName, fromName, p, o.Name(), p, o.Name())
}
func (g *javaGen) gobindOpts() string {
opts := []string{"-lang=java"}
if g.javaPkg != "" {
opts = append(opts, "-javapkg="+g.javaPkg)
}
return strings.Join(opts, " ")
}
var javaNameReplacer = strings.NewReplacer(
"-", "_",
".", "_",
)
func (g *javaGen) javaPkgName(pkg *types.Package) string {
if g.javaPkg != "" {
return g.javaPkg
}
s := javaNameReplacer.Replace(pkg.Name())
// Look for Java keywords that are not Go keywords, and avoid using
// them as a package name.
//
// This is not a problem for normal Go identifiers as we only expose
// exported symbols. The upper case first letter saves everything
// from accidentally matching except for the package name.
//
// Note that basic type names (like int) are not keywords in Go.
switch s {
case "abstract", "assert", "boolean", "byte", "catch", "char", "class",
"do", "double", "enum", "extends", "final", "finally", "float",
"implements", "instanceof", "int", "long", "native", "private",
"protected", "public", "short", "static", "strictfp", "super",
"synchronized", "this", "throw", "throws", "transient", "try",
"void", "volatile", "while":
s += "_"
}
return "go." + s
}
func (g *javaGen) className() string {
return className(g.pkg)
}
func className(pkg *types.Package) string {
return strings.Title(javaNameReplacer.Replace(pkg.Name()))
}
func (g *javaGen) genConst(o *types.Const) {
if _, ok := o.Type().(*types.Basic); !ok {
g.Printf("// skipped const %s with unsupported type: %T\n\n", o.Name(), o)
return
}
// TODO(hyangah): should const names use upper cases + "_"?
// TODO(hyangah): check invalid names.
jType := g.javaType(o.Type())
val := constExactString(o)
switch b := o.Type().(*types.Basic); b.Kind() {
case types.Int64, types.UntypedInt:
i, exact := constant.Int64Val(o.Val())
if !exact {
g.errorf("const value %s for %s cannot be represented as %s", val, o.Name(), jType)
return
}
val = fmt.Sprintf("%dL", i)
case types.Float32:
f, _ := constant.Float32Val(o.Val())
val = fmt.Sprintf("%gf", f)
case types.Float64, types.UntypedFloat:
f, _ := constant.Float64Val(o.Val())
if math.IsInf(f, 0) || math.Abs(f) > math.MaxFloat64 {
g.errorf("const value %s for %s cannot be represented as %s", val, o.Name(), jType)
return
}
val = fmt.Sprintf("%g", f)
}
g.Printf("public static final %s %s = %s;\n", g.javaType(o.Type()), o.Name(), val)
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *javaGen) genJNIField(o *types.TypeName, f *types.Var) {
if t := f.Type(); !g.isSupported(t) {
g.Printf("// skipped field %s with unsupported type: %T\n\n", o.Name(), t)
return
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
// setter
g.Printf("JNIEXPORT void JNICALL\n")
g.Printf("Java_%s_%s_00024%s_set%s(JNIEnv *env, jobject this, %s v) {\n", g.jniPkgName(), g.className(), o.Name(), f.Name(), g.jniType(f.Type()))
g.Indent()
g.Printf("int32_t o = go_seq_to_refnum(env, this);\n")
g.genJavaToC("v", f.Type(), modeRetained)
g.Printf("proxy%s_%s_%s_Set(o, _v);\n", g.pkgPrefix, o.Name(), f.Name())
g.genRelease("v", f.Type(), modeRetained)
g.Outdent()
g.Printf("}\n\n")
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
// getter
g.Printf("JNIEXPORT %s JNICALL\n", g.jniType(f.Type()))
g.Printf("Java_%s_%s_00024%s_get%s(JNIEnv *env, jobject this) {\n", g.jniPkgName(), g.className(), o.Name(), f.Name())
g.Indent()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("int32_t o = go_seq_to_refnum(env, this);\n")
g.Printf("%s r0 = ", g.cgoType(f.Type()))
g.Printf("proxy%s_%s_%s_Get(o);\n", g.pkgPrefix, o.Name(), f.Name())
mobile/bind: avoid intermediate []rune copy converting Java string to Go Converting a Go string to a string suitable use a specialized function, UTF16Encode, that can encode the string directly to a malloc'ed buffer. That way, only two copies are made when strings are passed from Go to Java; once for UTF-8 to UTF-16 encoding and once for the creation of the Java String. This CL implements the same optimization in the other direction, with a UTF-16 to UTF-8 decoder implemented in C. Unfortunately, while calling into a Go decoder also saves the extra copy, the Cgo overhead makes the calls much slower for short strings. To alleviate the risk of introducing decoding bugs, I've added the tests from the encoding/utf16 package to SeqTest. As a sideeffect, both Java and ObjC now always copy strings, regardless of the argument mode. The cpy argument can therefore be removed from the string conversion functions. Furthermore, the modeRetained and modeReturned modes can be collapsed into just one. While we're here, delete a leftover function from seq/strings.go that wasn't removed when the old seq buffers went away. Benchmarks, as compared with benchstat over 5 runs: name old time/op new time/op delta JavaStringShort 11.4µs ±13% 11.6µs ± 4% ~ (p=0.859 n=10+5) JavaStringShortDirect 19.5µs ± 9% 20.3µs ± 2% +3.68% (p=0.019 n=9+5) JavaStringLong 103µs ± 8% 24µs ± 4% -77.13% (p=0.001 n=9+5) JavaStringLongDirect 113µs ± 9% 32µs ± 7% -71.63% (p=0.001 n=9+5) JavaStringShortUnicode 11.1µs ±16% 10.7µs ± 5% ~ (p=0.190 n=9+5) JavaStringShortUnicodeDirect 19.6µs ± 7% 20.2µs ± 1% +2.78% (p=0.029 n=9+5) JavaStringLongUnicode 97.1µs ± 9% 28.0µs ± 5% -71.17% (p=0.001 n=9+5) JavaStringLongUnicodeDirect 105µs ±10% 34µs ± 5% -67.23% (p=0.002 n=8+5) JavaStringRetShort 14.2µs ± 2% 13.9µs ± 1% -2.15% (p=0.006 n=8+5) JavaStringRetShortDirect 20.8µs ± 2% 20.4µs ± 2% ~ (p=0.065 n=8+5) JavaStringRetLong 42.2µs ± 9% 42.4µs ± 3% ~ (p=0.190 n=9+5) JavaStringRetLongDirect 51.2µs ±21% 50.8µs ± 8% ~ (p=0.518 n=9+5) GoStringShort 23.4µs ± 7% 22.5µs ± 3% -3.55% (p=0.019 n=9+5) GoStringLong 51.9µs ± 9% 53.1µs ± 3% ~ (p=0.240 n=9+5) GoStringShortUnicode 24.2µs ± 6% 22.8µs ± 1% -5.54% (p=0.002 n=9+5) GoStringLongUnicode 58.6µs ± 8% 57.6µs ± 3% ~ (p=0.518 n=9+5) GoStringRetShort 27.6µs ± 1% 23.2µs ± 2% -15.87% (p=0.003 n=7+5) GoStringRetLong 129µs ±12% 33µs ± 2% -74.03% (p=0.001 n=10+5) Change-Id: Icb9481981493ffca8defed9fb80a9433d6048937 Reviewed-on: https://go-review.googlesource.com/20250 Reviewed-by: David Crawshaw <crawshaw@golang.org>
2016-03-04 17:59:04 +01:00
g.genCToJava("_r0", "r0", f.Type(), modeRetained)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("return _r0;\n")
g.Outdent()
g.Printf("}\n\n")
}
func (g *javaGen) genJNIVar(o *types.Var) {
if t := o.Type(); !g.isSupported(t) {
g.Printf("// skipped variable %s with unsupported type: %T\n\n", o.Name(), t)
return
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
// setter
g.Printf("JNIEXPORT void JNICALL\n")
g.Printf("Java_%s_%s_set%s(JNIEnv *env, jclass clazz, %s v) {\n", g.jniPkgName(), g.className(), o.Name(), g.jniType(o.Type()))
g.Indent()
g.genJavaToC("v", o.Type(), modeRetained)
g.Printf("var_set%s_%s(_v);\n", g.pkgPrefix, o.Name())
g.genRelease("v", o.Type(), modeRetained)
g.Outdent()
g.Printf("}\n\n")
// getter
g.Printf("JNIEXPORT %s JNICALL\n", g.jniType(o.Type()))
g.Printf("Java_%s_%s_get%s(JNIEnv *env, jclass clazz) {\n", g.jniPkgName(), g.className(), o.Name())
g.Indent()
g.Printf("%s r0 = ", g.cgoType(o.Type()))
g.Printf("var_get%s_%s();\n", g.pkgPrefix, o.Name())
mobile/bind: avoid intermediate []rune copy converting Java string to Go Converting a Go string to a string suitable use a specialized function, UTF16Encode, that can encode the string directly to a malloc'ed buffer. That way, only two copies are made when strings are passed from Go to Java; once for UTF-8 to UTF-16 encoding and once for the creation of the Java String. This CL implements the same optimization in the other direction, with a UTF-16 to UTF-8 decoder implemented in C. Unfortunately, while calling into a Go decoder also saves the extra copy, the Cgo overhead makes the calls much slower for short strings. To alleviate the risk of introducing decoding bugs, I've added the tests from the encoding/utf16 package to SeqTest. As a sideeffect, both Java and ObjC now always copy strings, regardless of the argument mode. The cpy argument can therefore be removed from the string conversion functions. Furthermore, the modeRetained and modeReturned modes can be collapsed into just one. While we're here, delete a leftover function from seq/strings.go that wasn't removed when the old seq buffers went away. Benchmarks, as compared with benchstat over 5 runs: name old time/op new time/op delta JavaStringShort 11.4µs ±13% 11.6µs ± 4% ~ (p=0.859 n=10+5) JavaStringShortDirect 19.5µs ± 9% 20.3µs ± 2% +3.68% (p=0.019 n=9+5) JavaStringLong 103µs ± 8% 24µs ± 4% -77.13% (p=0.001 n=9+5) JavaStringLongDirect 113µs ± 9% 32µs ± 7% -71.63% (p=0.001 n=9+5) JavaStringShortUnicode 11.1µs ±16% 10.7µs ± 5% ~ (p=0.190 n=9+5) JavaStringShortUnicodeDirect 19.6µs ± 7% 20.2µs ± 1% +2.78% (p=0.029 n=9+5) JavaStringLongUnicode 97.1µs ± 9% 28.0µs ± 5% -71.17% (p=0.001 n=9+5) JavaStringLongUnicodeDirect 105µs ±10% 34µs ± 5% -67.23% (p=0.002 n=8+5) JavaStringRetShort 14.2µs ± 2% 13.9µs ± 1% -2.15% (p=0.006 n=8+5) JavaStringRetShortDirect 20.8µs ± 2% 20.4µs ± 2% ~ (p=0.065 n=8+5) JavaStringRetLong 42.2µs ± 9% 42.4µs ± 3% ~ (p=0.190 n=9+5) JavaStringRetLongDirect 51.2µs ±21% 50.8µs ± 8% ~ (p=0.518 n=9+5) GoStringShort 23.4µs ± 7% 22.5µs ± 3% -3.55% (p=0.019 n=9+5) GoStringLong 51.9µs ± 9% 53.1µs ± 3% ~ (p=0.240 n=9+5) GoStringShortUnicode 24.2µs ± 6% 22.8µs ± 1% -5.54% (p=0.002 n=9+5) GoStringLongUnicode 58.6µs ± 8% 57.6µs ± 3% ~ (p=0.518 n=9+5) GoStringRetShort 27.6µs ± 1% 23.2µs ± 2% -15.87% (p=0.003 n=7+5) GoStringRetLong 129µs ±12% 33µs ± 2% -74.03% (p=0.001 n=10+5) Change-Id: Icb9481981493ffca8defed9fb80a9433d6048937 Reviewed-on: https://go-review.googlesource.com/20250 Reviewed-by: David Crawshaw <crawshaw@golang.org>
2016-03-04 17:59:04 +01:00
g.genCToJava("_r0", "r0", o.Type(), modeRetained)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("return _r0;\n")
g.Outdent()
g.Printf("}\n\n")
}
func (g *javaGen) genJNIFunc(o *types.Func, sName string, proxy bool) {
if !g.isSigSupported(o.Type()) {
n := o.Name()
if sName != "" {
n = sName + "." + n
}
g.Printf("// skipped function %s with unsupported parameter or return types\n\n", o.Name())
return
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genJNIFuncSignature(o, sName, proxy)
sig := o.Type().(*types.Signature)
res := sig.Results()
g.Printf(" {\n")
g.Indent()
if sName != "" {
g.Printf("int32_t o = go_seq_to_refnum(env, this);\n")
}
params := sig.Params()
for i := 0; i < params.Len(); i++ {
name := paramName(params, i)
g.genJavaToC(name, params.At(i).Type(), modeTransient)
}
resPrefix := ""
if res.Len() > 0 {
if res.Len() == 1 {
g.Printf("%s r0 = ", g.cgoType(res.At(0).Type()))
} else {
resPrefix = "res."
g.Printf("struct proxy%s_%s_%s_return res = ", g.pkgPrefix, sName, o.Name())
}
}
g.Printf("proxy%s_%s_%s(", g.pkgPrefix, sName, o.Name())
if sName != "" {
g.Printf("o")
}
for i := 0; i < params.Len(); i++ {
if i > 0 || sName != "" {
g.Printf(", ")
}
g.Printf("_%s", paramName(params, i))
}
g.Printf(");\n")
for i := 0; i < params.Len(); i++ {
g.genRelease(paramName(params, i), params.At(i).Type(), modeTransient)
}
for i := 0; i < res.Len(); i++ {
tn := fmt.Sprintf("_r%d", i)
t := res.At(i).Type()
mobile/bind: avoid intermediate []rune copy converting Java string to Go Converting a Go string to a string suitable use a specialized function, UTF16Encode, that can encode the string directly to a malloc'ed buffer. That way, only two copies are made when strings are passed from Go to Java; once for UTF-8 to UTF-16 encoding and once for the creation of the Java String. This CL implements the same optimization in the other direction, with a UTF-16 to UTF-8 decoder implemented in C. Unfortunately, while calling into a Go decoder also saves the extra copy, the Cgo overhead makes the calls much slower for short strings. To alleviate the risk of introducing decoding bugs, I've added the tests from the encoding/utf16 package to SeqTest. As a sideeffect, both Java and ObjC now always copy strings, regardless of the argument mode. The cpy argument can therefore be removed from the string conversion functions. Furthermore, the modeRetained and modeReturned modes can be collapsed into just one. While we're here, delete a leftover function from seq/strings.go that wasn't removed when the old seq buffers went away. Benchmarks, as compared with benchstat over 5 runs: name old time/op new time/op delta JavaStringShort 11.4µs ±13% 11.6µs ± 4% ~ (p=0.859 n=10+5) JavaStringShortDirect 19.5µs ± 9% 20.3µs ± 2% +3.68% (p=0.019 n=9+5) JavaStringLong 103µs ± 8% 24µs ± 4% -77.13% (p=0.001 n=9+5) JavaStringLongDirect 113µs ± 9% 32µs ± 7% -71.63% (p=0.001 n=9+5) JavaStringShortUnicode 11.1µs ±16% 10.7µs ± 5% ~ (p=0.190 n=9+5) JavaStringShortUnicodeDirect 19.6µs ± 7% 20.2µs ± 1% +2.78% (p=0.029 n=9+5) JavaStringLongUnicode 97.1µs ± 9% 28.0µs ± 5% -71.17% (p=0.001 n=9+5) JavaStringLongUnicodeDirect 105µs ±10% 34µs ± 5% -67.23% (p=0.002 n=8+5) JavaStringRetShort 14.2µs ± 2% 13.9µs ± 1% -2.15% (p=0.006 n=8+5) JavaStringRetShortDirect 20.8µs ± 2% 20.4µs ± 2% ~ (p=0.065 n=8+5) JavaStringRetLong 42.2µs ± 9% 42.4µs ± 3% ~ (p=0.190 n=9+5) JavaStringRetLongDirect 51.2µs ±21% 50.8µs ± 8% ~ (p=0.518 n=9+5) GoStringShort 23.4µs ± 7% 22.5µs ± 3% -3.55% (p=0.019 n=9+5) GoStringLong 51.9µs ± 9% 53.1µs ± 3% ~ (p=0.240 n=9+5) GoStringShortUnicode 24.2µs ± 6% 22.8µs ± 1% -5.54% (p=0.002 n=9+5) GoStringLongUnicode 58.6µs ± 8% 57.6µs ± 3% ~ (p=0.518 n=9+5) GoStringRetShort 27.6µs ± 1% 23.2µs ± 2% -15.87% (p=0.003 n=7+5) GoStringRetLong 129µs ±12% 33µs ± 2% -74.03% (p=0.001 n=10+5) Change-Id: Icb9481981493ffca8defed9fb80a9433d6048937 Reviewed-on: https://go-review.googlesource.com/20250 Reviewed-by: David Crawshaw <crawshaw@golang.org>
2016-03-04 17:59:04 +01:00
g.genCToJava(tn, fmt.Sprintf("%sr%d", resPrefix, i), t, modeRetained)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
// Go backwards so that any exception is thrown before
// the return.
for i := res.Len() - 1; i >= 0; i-- {
t := res.At(i).Type()
if !isErrorType(t) {
g.Printf("return _r%d;\n", i)
} else {
g.Printf("go_seq_maybe_throw_exception(env, _r%d);\n", i)
}
}
g.Outdent()
g.Printf("}\n\n")
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
// genRelease cleans up arguments that weren't copied in genJavaToC.
func (g *javaGen) genRelease(varName string, t types.Type, mode varMode) {
if isErrorType(t) {
g.genRelease(varName, types.Typ[types.String], mode)
return
}
switch t := t.(type) {
case *types.Basic:
case *types.Slice:
switch e := t.Elem().(type) {
case *types.Basic:
switch e.Kind() {
case types.Uint8: // Byte.
mobile/bind: avoid intermediate []rune copy converting Java string to Go Converting a Go string to a string suitable use a specialized function, UTF16Encode, that can encode the string directly to a malloc'ed buffer. That way, only two copies are made when strings are passed from Go to Java; once for UTF-8 to UTF-16 encoding and once for the creation of the Java String. This CL implements the same optimization in the other direction, with a UTF-16 to UTF-8 decoder implemented in C. Unfortunately, while calling into a Go decoder also saves the extra copy, the Cgo overhead makes the calls much slower for short strings. To alleviate the risk of introducing decoding bugs, I've added the tests from the encoding/utf16 package to SeqTest. As a sideeffect, both Java and ObjC now always copy strings, regardless of the argument mode. The cpy argument can therefore be removed from the string conversion functions. Furthermore, the modeRetained and modeReturned modes can be collapsed into just one. While we're here, delete a leftover function from seq/strings.go that wasn't removed when the old seq buffers went away. Benchmarks, as compared with benchstat over 5 runs: name old time/op new time/op delta JavaStringShort 11.4µs ±13% 11.6µs ± 4% ~ (p=0.859 n=10+5) JavaStringShortDirect 19.5µs ± 9% 20.3µs ± 2% +3.68% (p=0.019 n=9+5) JavaStringLong 103µs ± 8% 24µs ± 4% -77.13% (p=0.001 n=9+5) JavaStringLongDirect 113µs ± 9% 32µs ± 7% -71.63% (p=0.001 n=9+5) JavaStringShortUnicode 11.1µs ±16% 10.7µs ± 5% ~ (p=0.190 n=9+5) JavaStringShortUnicodeDirect 19.6µs ± 7% 20.2µs ± 1% +2.78% (p=0.029 n=9+5) JavaStringLongUnicode 97.1µs ± 9% 28.0µs ± 5% -71.17% (p=0.001 n=9+5) JavaStringLongUnicodeDirect 105µs ±10% 34µs ± 5% -67.23% (p=0.002 n=8+5) JavaStringRetShort 14.2µs ± 2% 13.9µs ± 1% -2.15% (p=0.006 n=8+5) JavaStringRetShortDirect 20.8µs ± 2% 20.4µs ± 2% ~ (p=0.065 n=8+5) JavaStringRetLong 42.2µs ± 9% 42.4µs ± 3% ~ (p=0.190 n=9+5) JavaStringRetLongDirect 51.2µs ±21% 50.8µs ± 8% ~ (p=0.518 n=9+5) GoStringShort 23.4µs ± 7% 22.5µs ± 3% -3.55% (p=0.019 n=9+5) GoStringLong 51.9µs ± 9% 53.1µs ± 3% ~ (p=0.240 n=9+5) GoStringShortUnicode 24.2µs ± 6% 22.8µs ± 1% -5.54% (p=0.002 n=9+5) GoStringLongUnicode 58.6µs ± 8% 57.6µs ± 3% ~ (p=0.518 n=9+5) GoStringRetShort 27.6µs ± 1% 23.2µs ± 2% -15.87% (p=0.003 n=7+5) GoStringRetLong 129µs ±12% 33µs ± 2% -74.03% (p=0.001 n=10+5) Change-Id: Icb9481981493ffca8defed9fb80a9433d6048937 Reviewed-on: https://go-review.googlesource.com/20250 Reviewed-by: David Crawshaw <crawshaw@golang.org>
2016-03-04 17:59:04 +01:00
if mode == modeTransient {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("if (_%s.ptr != NULL) {\n", varName)
g.Printf(" (*env)->ReleaseByteArrayElements(env, %s, _%s.ptr, 0);\n", varName, varName)
g.Printf("}\n")
}
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
}
}
func (g *javaGen) genMethodInterfaceProxy(oName string, m *types.Func) {
if !g.isSigSupported(m.Type()) {
g.Printf("// skipped method %s with unsupported parameter or return types\n\n", oName)
return
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
sig := m.Type().(*types.Signature)
params := sig.Params()
res := sig.Results()
g.genInterfaceMethodSignature(m, oName, false)
g.Indent()
// Push a JNI reference frame with a conservative capacity of two for each per parameter (Seq.Ref and Seq.Object),
// plus extra space for the receiver, the return value, and exception (if any).
g.Printf("JNIEnv *env = go_seq_push_local_frame(%d);\n", 2*params.Len()+10)
g.Printf("jobject o = go_seq_from_refnum(env, refnum, proxy_class_%s_%s, proxy_class_%s_%s_cons);\n", g.pkgPrefix, oName, g.pkgPrefix, oName)
for i := 0; i < params.Len(); i++ {
pn := paramName(params, i)
g.genCToJava("_"+pn, pn, params.At(i).Type(), modeTransient)
}
if res.Len() > 0 && !isErrorType(res.At(0).Type()) {
t := res.At(0).Type()
g.Printf("%s res = (*env)->Call%sMethod(env, o, ", g.jniType(t), g.jniCallType(t))
} else {
g.Printf("(*env)->CallVoidMethod(env, o, ")
}
g.Printf("mid_%s_%s", oName, m.Name())
for i := 0; i < params.Len(); i++ {
g.Printf(", _%s", paramName(params, i))
}
g.Printf(");\n")
var retName string
if res.Len() > 0 {
var rets []string
t := res.At(0).Type()
if !isErrorType(t) {
mobile/bind: avoid intermediate []rune copy converting Java string to Go Converting a Go string to a string suitable use a specialized function, UTF16Encode, that can encode the string directly to a malloc'ed buffer. That way, only two copies are made when strings are passed from Go to Java; once for UTF-8 to UTF-16 encoding and once for the creation of the Java String. This CL implements the same optimization in the other direction, with a UTF-16 to UTF-8 decoder implemented in C. Unfortunately, while calling into a Go decoder also saves the extra copy, the Cgo overhead makes the calls much slower for short strings. To alleviate the risk of introducing decoding bugs, I've added the tests from the encoding/utf16 package to SeqTest. As a sideeffect, both Java and ObjC now always copy strings, regardless of the argument mode. The cpy argument can therefore be removed from the string conversion functions. Furthermore, the modeRetained and modeReturned modes can be collapsed into just one. While we're here, delete a leftover function from seq/strings.go that wasn't removed when the old seq buffers went away. Benchmarks, as compared with benchstat over 5 runs: name old time/op new time/op delta JavaStringShort 11.4µs ±13% 11.6µs ± 4% ~ (p=0.859 n=10+5) JavaStringShortDirect 19.5µs ± 9% 20.3µs ± 2% +3.68% (p=0.019 n=9+5) JavaStringLong 103µs ± 8% 24µs ± 4% -77.13% (p=0.001 n=9+5) JavaStringLongDirect 113µs ± 9% 32µs ± 7% -71.63% (p=0.001 n=9+5) JavaStringShortUnicode 11.1µs ±16% 10.7µs ± 5% ~ (p=0.190 n=9+5) JavaStringShortUnicodeDirect 19.6µs ± 7% 20.2µs ± 1% +2.78% (p=0.029 n=9+5) JavaStringLongUnicode 97.1µs ± 9% 28.0µs ± 5% -71.17% (p=0.001 n=9+5) JavaStringLongUnicodeDirect 105µs ±10% 34µs ± 5% -67.23% (p=0.002 n=8+5) JavaStringRetShort 14.2µs ± 2% 13.9µs ± 1% -2.15% (p=0.006 n=8+5) JavaStringRetShortDirect 20.8µs ± 2% 20.4µs ± 2% ~ (p=0.065 n=8+5) JavaStringRetLong 42.2µs ± 9% 42.4µs ± 3% ~ (p=0.190 n=9+5) JavaStringRetLongDirect 51.2µs ±21% 50.8µs ± 8% ~ (p=0.518 n=9+5) GoStringShort 23.4µs ± 7% 22.5µs ± 3% -3.55% (p=0.019 n=9+5) GoStringLong 51.9µs ± 9% 53.1µs ± 3% ~ (p=0.240 n=9+5) GoStringShortUnicode 24.2µs ± 6% 22.8µs ± 1% -5.54% (p=0.002 n=9+5) GoStringLongUnicode 58.6µs ± 8% 57.6µs ± 3% ~ (p=0.518 n=9+5) GoStringRetShort 27.6µs ± 1% 23.2µs ± 2% -15.87% (p=0.003 n=7+5) GoStringRetLong 129µs ±12% 33µs ± 2% -74.03% (p=0.001 n=10+5) Change-Id: Icb9481981493ffca8defed9fb80a9433d6048937 Reviewed-on: https://go-review.googlesource.com/20250 Reviewed-by: David Crawshaw <crawshaw@golang.org>
2016-03-04 17:59:04 +01:00
g.genJavaToC("res", t, modeRetained)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
retName = "_res"
rets = append(rets, retName)
}
if res.Len() == 2 || isErrorType(t) {
g.Printf("jstring exc = go_seq_get_exception_message(env);\n")
st := types.Typ[types.String]
mobile/bind: avoid intermediate []rune copy converting Java string to Go Converting a Go string to a string suitable use a specialized function, UTF16Encode, that can encode the string directly to a malloc'ed buffer. That way, only two copies are made when strings are passed from Go to Java; once for UTF-8 to UTF-16 encoding and once for the creation of the Java String. This CL implements the same optimization in the other direction, with a UTF-16 to UTF-8 decoder implemented in C. Unfortunately, while calling into a Go decoder also saves the extra copy, the Cgo overhead makes the calls much slower for short strings. To alleviate the risk of introducing decoding bugs, I've added the tests from the encoding/utf16 package to SeqTest. As a sideeffect, both Java and ObjC now always copy strings, regardless of the argument mode. The cpy argument can therefore be removed from the string conversion functions. Furthermore, the modeRetained and modeReturned modes can be collapsed into just one. While we're here, delete a leftover function from seq/strings.go that wasn't removed when the old seq buffers went away. Benchmarks, as compared with benchstat over 5 runs: name old time/op new time/op delta JavaStringShort 11.4µs ±13% 11.6µs ± 4% ~ (p=0.859 n=10+5) JavaStringShortDirect 19.5µs ± 9% 20.3µs ± 2% +3.68% (p=0.019 n=9+5) JavaStringLong 103µs ± 8% 24µs ± 4% -77.13% (p=0.001 n=9+5) JavaStringLongDirect 113µs ± 9% 32µs ± 7% -71.63% (p=0.001 n=9+5) JavaStringShortUnicode 11.1µs ±16% 10.7µs ± 5% ~ (p=0.190 n=9+5) JavaStringShortUnicodeDirect 19.6µs ± 7% 20.2µs ± 1% +2.78% (p=0.029 n=9+5) JavaStringLongUnicode 97.1µs ± 9% 28.0µs ± 5% -71.17% (p=0.001 n=9+5) JavaStringLongUnicodeDirect 105µs ±10% 34µs ± 5% -67.23% (p=0.002 n=8+5) JavaStringRetShort 14.2µs ± 2% 13.9µs ± 1% -2.15% (p=0.006 n=8+5) JavaStringRetShortDirect 20.8µs ± 2% 20.4µs ± 2% ~ (p=0.065 n=8+5) JavaStringRetLong 42.2µs ± 9% 42.4µs ± 3% ~ (p=0.190 n=9+5) JavaStringRetLongDirect 51.2µs ±21% 50.8µs ± 8% ~ (p=0.518 n=9+5) GoStringShort 23.4µs ± 7% 22.5µs ± 3% -3.55% (p=0.019 n=9+5) GoStringLong 51.9µs ± 9% 53.1µs ± 3% ~ (p=0.240 n=9+5) GoStringShortUnicode 24.2µs ± 6% 22.8µs ± 1% -5.54% (p=0.002 n=9+5) GoStringLongUnicode 58.6µs ± 8% 57.6µs ± 3% ~ (p=0.518 n=9+5) GoStringRetShort 27.6µs ± 1% 23.2µs ± 2% -15.87% (p=0.003 n=7+5) GoStringRetLong 129µs ±12% 33µs ± 2% -74.03% (p=0.001 n=10+5) Change-Id: Icb9481981493ffca8defed9fb80a9433d6048937 Reviewed-on: https://go-review.googlesource.com/20250 Reviewed-by: David Crawshaw <crawshaw@golang.org>
2016-03-04 17:59:04 +01:00
g.genJavaToC("exc", st, modeRetained)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
retName = "_exc"
rets = append(rets, "_exc")
}
if res.Len() > 1 {
g.Printf("cproxy%s_%s_%s_return sres = {\n", g.pkgPrefix, oName, m.Name())
g.Printf(" %s\n", strings.Join(rets, ", "))
g.Printf("};\n")
retName = "sres"
}
}
g.Printf("go_seq_pop_local_frame(env);\n")
if retName != "" {
g.Printf("return %s;\n", retName)
}
g.Outdent()
g.Printf("}\n\n")
}
func (g *javaGen) genH() error {
g.Printf(hPreamble, g.gobindOpts(), g.pkg.Path(), g.className())
for _, iface := range g.interfaces {
g.Printf("extern jclass proxy_class_%s_%s;\n", g.pkgPrefix, iface.obj.Name())
g.Printf("extern jmethodID proxy_class_%s_%s_cons;\n", g.pkgPrefix, iface.obj.Name())
g.Printf("\n")
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
for _, m := range iface.summary.callable {
if !g.isSigSupported(m.Type()) {
g.Printf("// skipped method %s.%s with unsupported parameter or return types\n\n", iface.obj.Name(), m.Name())
continue
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genInterfaceMethodSignature(m, iface.obj.Name(), true)
g.Printf("\n")
}
}
for _, s := range g.structs {
g.Printf("extern jclass proxy_class_%s_%s;\n", g.pkgPrefix, s.obj.Name())
g.Printf("extern jmethodID proxy_class_%s_%s_cons;\n", g.pkgPrefix, s.obj.Name())
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("#endif\n")
if len(g.err) > 0 {
return g.err
}
return nil
}
func (g *javaGen) jniCallType(t types.Type) string {
if isErrorType(t) {
return g.jniCallType(types.Typ[types.String])
}
switch t := t.(type) {
case *types.Basic:
switch t.Kind() {
case types.Bool, types.UntypedBool:
return "Boolean"
case types.Int:
return "Long"
case types.Int8, types.Uint8: // types.Byte
return "Byte"
case types.Int16:
return "Short"
case types.Int32, types.UntypedRune: // types.Rune
return "Int"
case types.Int64, types.UntypedInt:
return "Long"
case types.Float32:
return "Float"
case types.Float64, types.UntypedFloat:
return "Double"
case types.String, types.UntypedString:
return "Object"
default:
g.errorf("unsupported basic type: %s", t)
}
case *types.Slice:
return "Object"
case *types.Pointer:
if _, ok := t.Elem().(*types.Named); ok {
return g.jniCallType(t.Elem())
}
g.errorf("unsupported pointer to type: %s", t)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
case *types.Named:
return "Object"
default:
return "Object"
}
return "TODO"
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
func (g *javaGen) jniClassSigPrefix(pkg *types.Package) string {
return strings.Replace(g.javaPkgName(pkg), ".", "/", -1) + "/" + className(pkg) + "$"
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
func (g *javaGen) jniSigType(T types.Type) string {
if isErrorType(T) {
return g.jniSigType(types.Typ[types.String])
}
switch T := T.(type) {
case *types.Basic:
switch T.Kind() {
case types.Bool, types.UntypedBool:
return "Z"
case types.Int:
return "J"
case types.Int8:
return "B"
case types.Int16:
return "S"
case types.Int32, types.UntypedRune: // types.Rune
return "I"
case types.Int64, types.UntypedInt:
return "J"
case types.Uint8: // types.Byte
return "B"
case types.Float32:
return "F"
case types.Float64, types.UntypedFloat:
return "D"
case types.String, types.UntypedString:
return "Ljava/lang/String;"
default:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.errorf("unsupported basic type: %s", T)
return "TODO"
}
case *types.Slice:
return "[" + g.jniSigType(T.Elem())
case *types.Pointer:
if _, ok := T.Elem().(*types.Named); ok {
return g.jniSigType(T.Elem())
}
g.errorf("unsupported pointer to type: %s", T)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
case *types.Named:
return "L" + g.jniClassSigPrefix(T.Obj().Pkg()) + T.Obj().Name() + ";"
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
default:
g.errorf("unsupported jniType: %#+v, %s\n", T, T)
}
return "TODO"
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
func (g *javaGen) genC() error {
g.Printf(cPreamble, g.gobindOpts(), g.pkg.Path(), g.pkg.Name())
g.Printf("#include %q\n", g.pkg.Name()+".h")
for _, pkg := range g.pkg.Imports() {
if g.validPkg(pkg) {
g.Printf("#include \"%s.h\"\n", pkg.Name())
}
}
g.Printf("\n")
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
for _, iface := range g.interfaces {
g.Printf("jclass proxy_class_%s_%s;\n", g.pkgPrefix, iface.obj.Name())
g.Printf("jmethodID proxy_class_%s_%s_cons;\n", g.pkgPrefix, iface.obj.Name())
for _, m := range iface.summary.callable {
if !g.isSigSupported(m.Type()) {
g.Printf("// skipped method %s.%s with unsupported parameter or return types\n\n", iface.obj.Name(), m.Name())
continue
}
g.Printf("static jmethodID mid_%s_%s;\n", iface.obj.Name(), m.Name())
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
}
for _, s := range g.structs {
g.Printf("jclass proxy_class_%s_%s;\n", g.pkgPrefix, s.obj.Name())
g.Printf("jmethodID proxy_class_%s_%s_cons;\n", g.pkgPrefix, s.obj.Name())
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
g.Printf("\n")
g.Printf("JNIEXPORT void JNICALL\n")
g.Printf("Java_%s_%s_init(JNIEnv *env, jclass _unused) {\n", g.jniPkgName(), g.className())
g.Indent()
g.Printf("jclass clazz;\n")
for _, s := range g.structs {
g.Printf("clazz = (*env)->FindClass(env, %q);\n", g.jniClassSigPrefix(s.obj.Pkg())+s.obj.Name())
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("proxy_class_%s_%s = (*env)->NewGlobalRef(env, clazz);\n", g.pkgPrefix, s.obj.Name())
g.Printf("proxy_class_%s_%s_cons = (*env)->GetMethodID(env, clazz, \"<init>\", \"(Lgo/Seq$Ref;)V\");\n", g.pkgPrefix, s.obj.Name())
}
for _, iface := range g.interfaces {
g.Printf("clazz = (*env)->FindClass(env, %q);\n", g.jniClassSigPrefix(iface.obj.Pkg())+"proxy"+iface.obj.Name())
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("proxy_class_%s_%s = (*env)->NewGlobalRef(env, clazz);\n", g.pkgPrefix, iface.obj.Name())
g.Printf("proxy_class_%s_%s_cons = (*env)->GetMethodID(env, clazz, \"<init>\", \"(Lgo/Seq$Ref;)V\");\n", g.pkgPrefix, iface.obj.Name())
g.Printf("clazz = (*env)->FindClass(env, %q);\n", g.jniClassSigPrefix(iface.obj.Pkg())+iface.obj.Name())
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
for _, m := range iface.summary.callable {
if !g.isSigSupported(m.Type()) {
g.Printf("// skipped method %s.%s with unsupported parameter or return types\n\n", iface.obj.Name(), m.Name())
continue
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
sig := m.Type().(*types.Signature)
res := sig.Results()
retSig := "V"
if res.Len() > 0 {
if t := res.At(0).Type(); !isErrorType(t) {
retSig = g.jniSigType(t)
}
}
var jniParams string
params := sig.Params()
for i := 0; i < params.Len(); i++ {
jniParams += g.jniSigType(params.At(i).Type())
}
g.Printf("mid_%s_%s = (*env)->GetMethodID(env, clazz, %q, \"(%s)%s\");\n",
iface.obj.Name(), m.Name(), m.Name(), jniParams, retSig)
}
g.Printf("\n")
}
g.Outdent()
g.Printf("}\n\n")
for _, f := range g.funcs {
g.genJNIFunc(f, "", false)
}
for _, s := range g.structs {
sName := s.obj.Name()
for _, m := range exportedMethodSet(types.NewPointer(s.obj.Type())) {
g.genJNIFunc(m, sName, false)
}
for _, f := range exportedFields(s.t) {
g.genJNIField(s.obj, f)
}
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
for _, iface := range g.interfaces {
for _, m := range iface.summary.callable {
g.genJNIFunc(m, iface.obj.Name(), true)
g.genMethodInterfaceProxy(iface.obj.Name(), m)
}
}
for _, v := range g.vars {
g.genJNIVar(v)
}
if len(g.err) > 0 {
return g.err
}
return nil
}
func (g *javaGen) genJava() error {
g.Printf(javaPreamble, g.javaPkgName(g.pkg), g.className(), g.gobindOpts(), g.pkg.Path())
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("public abstract class %s {\n", g.className())
g.Indent()
g.Printf("static {\n")
g.Indent()
g.Printf("Seq.touch(); // for loading the native library\n")
for _, p := range g.pkg.Imports() {
if g.validPkg(p) {
g.Printf("%s.%s.touch();\n", g.javaPkgName(p), className(p))
}
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("init();\n")
g.Outdent()
g.Printf("}\n\n")
g.Printf("private %s() {} // uninstantiable\n\n", g.className())
g.Printf("// touch is called from other bound packages to initialize this package\n")
g.Printf("public static void touch() {}\n\n")
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("private static native void init();\n\n")
for _, s := range g.structs {
g.genStruct(s.obj, s.t)
}
for _, iface := range g.interfaces {
g.genInterface(iface)
}
for _, c := range g.constants {
g.genConst(c)
}
g.Printf("\n")
for _, v := range g.vars {
g.genVar(v)
}
for _, f := range g.funcs {
if !g.isSigSupported(f.Type()) {
g.Printf("// skipped function %s with unsupported parameter or return types\n\n", f.Name())
continue
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genFuncSignature(f, true, false)
}
g.Outdent()
g.Printf("}\n")
if len(g.err) > 0 {
return g.err
}
return nil
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
const (
javaProxyPreamble = `private static final class proxy%[1]s extends Seq.Proxy implements %[1]s {
proxy%[1]s(Seq.Ref ref) { super(ref); }
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
`
javaPreamble = `// Java class %[1]s.%[2]s is a proxy for talking to a Go program.
// gobind %[3]s %[4]s
//
// File is generated by gobind. Do not edit.
package %[1]s;
import go.Seq;
`
cPreamble = `// JNI functions for the Go <=> Java bridge.
// gobind %[1]s %[2]s
//
// File is generated by gobind. Do not edit.
#include <android/log.h>
#include <stdint.h>
#include "seq.h"
#include "_cgo_export.h"
`
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
hPreamble = `// JNI function headers for the Go <=> Java bridge.
// gobind %[1]s %[2]s
//
// File is generated by gobind. Do not edit.
#ifndef __%[3]s_H__
#define __%[3]s_H__
#include <jni.h>
`
)