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

inline func are OK for extended precision

This commit is contained in:
Mamy André-Ratsimbazafy 2020-02-29 17:25:40 +01:00
parent 0fab0c8a42
commit 309a15ce8b
No known key found for this signature in database
GPG Key ID: 7B88AD1FE79492E1
1 changed files with 14 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
constantine/primitives/extended_precision.nim
View File

@ -37,7 +37,7 @@ func unsafeDiv2n1n*(q, r: var Ct[uint32], n_hi, n_lo, d: Ct[uint32]) {.inline.}=
q = (Ct[uint32])(dividend div divisor) q = (Ct[uint32])(dividend div divisor)
r = (Ct[uint32])(dividend mod divisor) r = (Ct[uint32])(dividend mod divisor)
template unsafeFMA*(hi, lo: var Ct[uint32], a, b, c: Ct[uint32]) = func unsafeFMA*(hi, lo: var Ct[uint32], a, b, c: Ct[uint32]) {.inline.} =
## Extended precision multiplication + addition ## Extended precision multiplication + addition
## This is constant-time on most hardware except some specific one like Cortex M0 ## This is constant-time on most hardware except some specific one like Cortex M0
## (hi, lo) <- a*b + c ## (hi, lo) <- a*b + c
@ -48,7 +48,7 @@ template unsafeFMA*(hi, lo: var Ct[uint32], a, b, c: Ct[uint32]) =
hi = Ct[uint32](dblPrec shr 31) hi = Ct[uint32](dblPrec shr 31)
lo = Ct[uint32](dblPrec) and Ct[uint32](1 shl 31 - 1) lo = Ct[uint32](dblPrec) and Ct[uint32](1 shl 31 - 1)
template unsafeFMA2*(hi, lo: var Ct[uint32], a1, b1, a2, b2, c1, c2: Ct[uint32]) = func unsafeFMA2*(hi, lo: var Ct[uint32], a1, b1, a2, b2, c1, c2: Ct[uint32]) {.inline.}=
## (hi, lo) <- a1 * b1 + a2 * b2 + c1 + c2 ## (hi, lo) <- a1 * b1 + a2 * b2 + c1 + c2
block: block:
# TODO: Can this overflow? # TODO: Can this overflow?
@ -59,7 +59,7 @@ template unsafeFMA2*(hi, lo: var Ct[uint32], a1, b1, a2, b2, c1, c2: Ct[uint32])
hi = Ct[uint32](dblPrec shr 31) hi = Ct[uint32](dblPrec shr 31)
lo = Ct[uint32](dblPrec) and Ct[uint32](1 shl 31 - 1) lo = Ct[uint32](dblPrec) and Ct[uint32](1 shl 31 - 1)
template unsafeFMA2_hi*(hi: var Ct[uint32], a1, b1, a2, b2, c1: Ct[uint32]) = func unsafeFMA2_hi*(hi: var Ct[uint32], a1, b1, a2, b2, c1: Ct[uint32]) {.inline.}=
## Returns the high word of the sum of extended precision multiply-adds ## Returns the high word of the sum of extended precision multiply-adds
## (hi, _) <- a1 * b1 + a2 * b2 + c ## (hi, _) <- a1 * b1 + a2 * b2 + c
block: block:
@ -111,7 +111,7 @@ when defined(gcc) or defined(clang) or defined(llvm_gcc):
: :
""" """
template unsafeFMA*(hi, lo: var Ct[uint64], a, b, c: Ct[uint64]) = func unsafeFMA*(hi, lo: var Ct[uint64], a, b, c: Ct[uint64]) {.inline.}=
## Extended precision multiplication + addition ## Extended precision multiplication + addition
## This is constant-time on most hardware except some specific one like Cortex M0 ## This is constant-time on most hardware except some specific one like Cortex M0
## (hi, lo) <- a*b + c ## (hi, lo) <- a*b + c
@ -121,10 +121,11 @@ when defined(gcc) or defined(clang) or defined(llvm_gcc):
var dblPrec {.noInit.}: uint128 var dblPrec {.noInit.}: uint128
{.emit:[dblPrec, " = (unsigned __int128)", a," * (unsigned __int128)", b, " + (unsigned __int128)",c,";"].} {.emit:[dblPrec, " = (unsigned __int128)", a," * (unsigned __int128)", b, " + (unsigned __int128)",c,";"].}
{.emit:[hi, " = (NU64)(", dblPrec," >> ", 63'u64, ");"].} # Don't forget to dereference the var param
{.emit:[lo, " = (NU64)", dblPrec," & ", 1'u64 shl 63 - 1, ";"].} {.emit:["*",hi, " = (NU64)(", dblPrec," >> ", 63'u64, ");"].}
{.emit:["*",lo, " = (NU64)", dblPrec," & ", 1'u64 shl 63 - 1, ";"].}
template unsafeFMA2*(hi, lo: var Ct[uint64], a1, b1, a2, b2, c1, c2: Ct[uint64]) = func unsafeFMA2*(hi, lo: var Ct[uint64], a1, b1, a2, b2, c1, c2: Ct[uint64]) {.inline.}=
## (hi, lo) <- a1 * b1 + a2 * b2 + c1 + c2 ## (hi, lo) <- a1 * b1 + a2 * b2 + c1 + c2
block: block:
# TODO: Can this overflow? # TODO: Can this overflow?
@ -133,10 +134,11 @@ when defined(gcc) or defined(clang) or defined(llvm_gcc):
" + (unsigned __int128)", a2," * (unsigned __int128)", b2, " + (unsigned __int128)", a2," * (unsigned __int128)", b2,
" + (unsigned __int128)", c1, " + (unsigned __int128)", c1,
" + (unsigned __int128)", c2, ";"].} " + (unsigned __int128)", c2, ";"].}
{.emit:[hi, " = (NU64)(", dblPrec," >> ", 63'u64, ");"].} # Don't forget to dereference the var param
{.emit:[lo, " = (NU64)", dblPrec," & ", (1'u64 shl 63 - 1), ";"].} {.emit:["*",hi, " = (NU64)(", dblPrec," >> ", 63'u64, ");"].}
{.emit:["*",lo, " = (NU64)", dblPrec," & ", (1'u64 shl 63 - 1), ";"].}
template unsafeFMA2_hi*(hi: var Ct[uint64], a1, b1, a2, b2, c: Ct[uint64]) = func unsafeFMA2_hi*(hi: var Ct[uint64], a1, b1, a2, b2, c: Ct[uint64]) {.inline.}=
## Returns the high word of the sum of extended precision multiply-adds ## Returns the high word of the sum of extended precision multiply-adds
## (hi, _) <- a1 * b1 + a2 * b2 + c ## (hi, _) <- a1 * b1 + a2 * b2 + c
block: block:
@ -144,7 +146,8 @@ when defined(gcc) or defined(clang) or defined(llvm_gcc):
{.emit:[dblPrec, " = (unsigned __int128)", a1," * (unsigned __int128)", b1, {.emit:[dblPrec, " = (unsigned __int128)", a1," * (unsigned __int128)", b1,
" + (unsigned __int128)", a2," * (unsigned __int128)", b2, " + (unsigned __int128)", a2," * (unsigned __int128)", b2,
" + (unsigned __int128)", c, ";"].} " + (unsigned __int128)", c, ";"].}
{.emit:[hi, " = (NU64)(", dblPrec," >> ", 63'u64, ");"].} # Don't forget to dereference the var param
{.emit:["*",hi, " = (NU64)(", dblPrec," >> ", 63'u64, ");"].}
else: else:
{.error: "Compiler not implemented".} {.error: "Compiler not implemented".}