codex-storage
/
constantine
mirror of https://github.com/codex-storage/constantine.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Use Assembly cmov on x86

This commit is contained in:
Mamy André-Ratsimbazafy 2020-02-29 18:27:20 +01:00
parent 309a15ce8b
commit ca817fcb69
No known key found for this signature in database
GPG Key ID: 7B88AD1FE79492E1
2 changed files with 62 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
benchmarks/bls12_381_fp.nim
View File

@ -43,8 +43,8 @@ proc addBench() =
x += y x += y
let stop = getMonotime() let stop = getMonotime()
echo &"Time for {Iters} conditional additions (constant-time 381-bit): {inMilliseconds(stop-start)} ms" echo &"Time for {Iters} additions in 𝔽p (constant-time 381-bit): {inMilliseconds(stop-start)} ms"
echo &"Time for 1 conditional addition ==> {inNanoseconds((stop-start) div Iters)} ns" echo &"Time for 1 addition in 𝔽p ==> {inNanoseconds((stop-start) div Iters)} ns"
addBench() addBench()
@ -60,7 +60,7 @@ proc mulBench() =
r.prod(x, y) r.prod(x, y)
let stop = getMonotime() let stop = getMonotime()
echo &"Time for {Iters} multiplications (constant-time 381-bit): {inMilliseconds(stop-start)} ms" echo &"Time for {Iters} multiplications 𝔽p (constant-time 381-bit): {inMilliseconds(stop-start)} ms"
echo &"Time for 1 multiplication ==> {inNanoseconds((stop-start) div Iters)} ns" echo &"Time for 1 multiplication 𝔽p ==> {inNanoseconds((stop-start) div Iters)} ns"
mulBench() mulBench()

70
constantine/primitives/constant_time.nim
View File

@ -227,7 +227,7 @@ template `<=`*[T: Ct](x, y: T): CTBool[T] =
template `xor`*[T: Ct](x, y: CTBool[T]): CTBool[T] = template `xor`*[T: Ct](x, y: CTBool[T]): CTBool[T] =
CTBool[T](noteq(T(x), T(y))) CTBool[T](noteq(T(x), T(y)))
template mux*[T: Ct](ctl: CTBool[T], x, y: T): T = func mux*[T](ctl: CTBool[T], x, y: T): T {.inline.}=
## Multiplexer / selector ## Multiplexer / selector
## Returns x if ctl is true ## Returns x if ctl is true
## else returns y ## else returns y
@ -237,22 +237,68 @@ template mux*[T: Ct](ctl: CTBool[T], x, y: T): T =
# Alternatives: # Alternatives:
# - https://cryptocoding.net/index.php/Coding_rules # - https://cryptocoding.net/index.php/Coding_rules
# - https://www.cl.cam.ac.uk/~rja14/Papers/whatyouc.pdf # - https://www.cl.cam.ac.uk/~rja14/Papers/whatyouc.pdf
# when defined(amd64) or defined(i386):
# TODO: assembly fastpath for conditional mov when sizeof(T) == 8:
let # Templates duplicate input params code var muxed = x
x_Mux = x asm """
y_Mux = y testq %[ctl], %[ctl]
y_Mux xor (-T(ctl) and (x_Mux xor y_Mux)) cmovzq %[y], %[muxed]
: [muxed] "+r" (`muxed`)
: [ctl] "r" (`ctl`), [y] "r" (`y`)
: "cc"
"""
muxed
elif sizeof(T) == 4:
var muxed = x
asm """
testl %[ctl], %[ctl]
cmovzl %[y], %[muxed]
: [muxed] "+r" (`muxed`)
: [ctl] "r" (`ctl`), [y] "r" (`y`)
: "cc"
"""
muxed
else:
{.error: "Unsupported word size".}
else:
let # Templates duplicate input params code
x_Mux = x
y_Mux = y
y_Mux xor (-T(ctl) and (x_Mux xor y_Mux))
template mux*[T: CTBool](ctl: CTBool, x, y: T): T = func mux*[T: CTBool](ctl: CTBool, x, y: T): T {.inline.}=
## Multiplexer / selector ## Multiplexer / selector
## Returns x if ctl is true ## Returns x if ctl is true
## else returns y ## else returns y
## So equivalent to ctl? x: y ## So equivalent to ctl? x: y
let # Templates duplicate input params code when defined(amd64) or defined(i386):
x_Mux = x when sizeof(T) == 8:
y_Mux = y var muxed = x
T(T.T(y_Mux) xor (-T.T(ctl) and T.T(x_Mux xor y_Mux))) asm """
testq %[ctl], %[ctl]
cmovzq %[y], %[muxed]
: [muxed] "+r" (`muxed`)
: [ctl] "r" (`ctl`), [y] "r" (`y`)
: "cc"
"""
muxed
elif sizeof(T) == 4:
var muxed = x
asm """
testl %[ctl], %[ctl]
cmovzl %[y], %[muxed]
: [muxed] "+r" (`muxed`)
: [ctl] "r" (`ctl`), [y] "r" (`y`)
: "cc"
"""
muxed
else:
{.error: "Unsupported word size".}
else:
let # Templates duplicate input params code
x_Mux = x
y_Mux = y
T(T.T(y_Mux) xor (-T.T(ctl) and T.T(x_Mux xor y_Mux)))
# ############################################################ # ############################################################
# #