go.crypto/curve25519: add package.
This consists of ~2000 lines of amd64 assembly and a, much slower, generic Go version in curve25519.go. The assembly has been ported from djb's public domain sources and the only semantic alterations are to deal with Go's split stacks. R=rsc CC=golang-dev https://golang.org/cl/5786045
This commit is contained in:
Adam Langley
@@ -0,0 +1,18 @@
|
||||
// Copyright 2012 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.
|
||||
|
||||
// This code was translated into a form compatible with 6a from the public
|
||||
// domain sources in SUPERCOP: http://bench.cr.yp.to/supercop.html
|
||||
|
||||
DATA ·REDMASK51(SB)/8, $0x0007FFFFFFFFFFFF
|
||||
GLOBL ·REDMASK51(SB), $8
|
||||
|
||||
DATA ·_121666_213(SB)/8, $996687872
|
||||
GLOBL ·_121666_213(SB), $8
|
||||
|
||||
DATA ·_2P0(SB)/8, $0xFFFFFFFFFFFDA
|
||||
GLOBL ·_2P0(SB), $8
|
||||
|
||||
DATA ·_2P1234(SB)/8, $0xFFFFFFFFFFFFE
|
||||
GLOBL ·_2P1234(SB), $8
|
||||
@@ -0,0 +1,86 @@
|
||||
// Copyright 2012 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.
|
||||
|
||||
// This code was translated into a form compatible with 6a from the public
|
||||
// domain sources in SUPERCOP: http://bench.cr.yp.to/supercop.html
|
||||
|
||||
// func cswap(inout *[5]uint64, v uint64)
|
||||
TEXT ·cswap(SB),7,$0
|
||||
MOVQ inout+0(FP),DI
|
||||
MOVQ v+8(FP),SI
|
||||
|
||||
CMPQ SI,$1
|
||||
MOVQ 0(DI),SI
|
||||
MOVQ 80(DI),DX
|
||||
MOVQ 8(DI),CX
|
||||
MOVQ 88(DI),R8
|
||||
MOVQ SI,R9
|
||||
CMOVQEQ DX,SI
|
||||
CMOVQEQ R9,DX
|
||||
MOVQ CX,R9
|
||||
CMOVQEQ R8,CX
|
||||
CMOVQEQ R9,R8
|
||||
MOVQ SI,0(DI)
|
||||
MOVQ DX,80(DI)
|
||||
MOVQ CX,8(DI)
|
||||
MOVQ R8,88(DI)
|
||||
MOVQ 16(DI),SI
|
||||
MOVQ 96(DI),DX
|
||||
MOVQ 24(DI),CX
|
||||
MOVQ 104(DI),R8
|
||||
MOVQ SI,R9
|
||||
CMOVQEQ DX,SI
|
||||
CMOVQEQ R9,DX
|
||||
MOVQ CX,R9
|
||||
CMOVQEQ R8,CX
|
||||
CMOVQEQ R9,R8
|
||||
MOVQ SI,16(DI)
|
||||
MOVQ DX,96(DI)
|
||||
MOVQ CX,24(DI)
|
||||
MOVQ R8,104(DI)
|
||||
MOVQ 32(DI),SI
|
||||
MOVQ 112(DI),DX
|
||||
MOVQ 40(DI),CX
|
||||
MOVQ 120(DI),R8
|
||||
MOVQ SI,R9
|
||||
CMOVQEQ DX,SI
|
||||
CMOVQEQ R9,DX
|
||||
MOVQ CX,R9
|
||||
CMOVQEQ R8,CX
|
||||
CMOVQEQ R9,R8
|
||||
MOVQ SI,32(DI)
|
||||
MOVQ DX,112(DI)
|
||||
MOVQ CX,40(DI)
|
||||
MOVQ R8,120(DI)
|
||||
MOVQ 48(DI),SI
|
||||
MOVQ 128(DI),DX
|
||||
MOVQ 56(DI),CX
|
||||
MOVQ 136(DI),R8
|
||||
MOVQ SI,R9
|
||||
CMOVQEQ DX,SI
|
||||
CMOVQEQ R9,DX
|
||||
MOVQ CX,R9
|
||||
CMOVQEQ R8,CX
|
||||
CMOVQEQ R9,R8
|
||||
MOVQ SI,48(DI)
|
||||
MOVQ DX,128(DI)
|
||||
MOVQ CX,56(DI)
|
||||
MOVQ R8,136(DI)
|
||||
MOVQ 64(DI),SI
|
||||
MOVQ 144(DI),DX
|
||||
MOVQ 72(DI),CX
|
||||
MOVQ 152(DI),R8
|
||||
MOVQ SI,R9
|
||||
CMOVQEQ DX,SI
|
||||
CMOVQEQ R9,DX
|
||||
MOVQ CX,R9
|
||||
CMOVQEQ R8,CX
|
||||
CMOVQEQ R9,R8
|
||||
MOVQ SI,64(DI)
|
||||
MOVQ DX,144(DI)
|
||||
MOVQ CX,72(DI)
|
||||
MOVQ R8,152(DI)
|
||||
MOVQ DI,AX
|
||||
MOVQ SI,DX
|
||||
RET
|
||||
@@ -0,0 +1,136 @@
|
||||
// Copyright 2012 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 curve25519
|
||||
|
||||
// We have a implementation in amd64 assembly so this code is only run on
|
||||
// non-amd64 platforms.
|
||||
// +build !amd64
|
||||
|
||||
import (
|
||||
"math/big"
|
||||
)
|
||||
|
||||
// p is the prime order of the underlying field: 2^255-19
|
||||
var p *big.Int
|
||||
|
||||
// pMinus2 is p-2
|
||||
var pMinus2 *big.Int
|
||||
|
||||
// a is a parameter of the elliptic curve: 486662
|
||||
var a *big.Int
|
||||
|
||||
func init() {
|
||||
p, _ = new(big.Int).SetString("7fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffed", 16)
|
||||
pMinus2, _ = new(big.Int).SetString("7fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffeb", 16)
|
||||
a = new(big.Int).SetInt64(486662)
|
||||
}
|
||||
|
||||
// context contains state shared throughout the computation, including scratch
|
||||
// variables to save on allocation.
|
||||
type context struct {
|
||||
tmp1, tmp2, tmp3, tmp4 *big.Int
|
||||
x1 *big.Int
|
||||
}
|
||||
|
||||
// add sets (outx, outz) to the sum of two points in the elliptic curve group.
|
||||
// See http://hyperelliptic.org/EFD/g1p/auto-montgom-xz.html#diffadd-dadd-1987-m
|
||||
// outx and outz should not alias any of the other inputs.
|
||||
func (c *context) add(outx, outz, xn, zn, xm, zm *big.Int) {
|
||||
// x₃ = 4(x·x′ - z·z′)² · z1
|
||||
// (z1 == 1 here)
|
||||
c.tmp1.Mul(xn, xm)
|
||||
c.tmp2.Mul(zn, zm)
|
||||
c.tmp3.Sub(c.tmp1, c.tmp2)
|
||||
outx.Mul(c.tmp3, c.tmp3)
|
||||
outx.Lsh(outx, 2)
|
||||
outx.Mod(outx, p)
|
||||
|
||||
// z₃ = 4(x·z′ - z·x′)² · x1
|
||||
// (x1 == 1 here)
|
||||
c.tmp1.Mul(xm, zn)
|
||||
c.tmp2.Mul(zm, xn)
|
||||
c.tmp3.Sub(c.tmp1, c.tmp2)
|
||||
outz.Mul(c.tmp3, c.tmp3)
|
||||
outz.Mul(outz, c.x1)
|
||||
outz.Lsh(outz, 2)
|
||||
outz.Mod(outz, p)
|
||||
|
||||
return
|
||||
}
|
||||
|
||||
// double sets (outx, outz) to 2*(x,z) in the elliptic curve group. See
|
||||
// http://hyperelliptic.org/EFD/g1p/auto-montgom-xz.html#doubling-dbl-1987-m
|
||||
// outx and outz should not alias any of the other inputs.
|
||||
func (c *context) double(outx, outz, x, z *big.Int) {
|
||||
// x₂ = (x² - z²)²
|
||||
c.tmp1.Mul(x, x)
|
||||
c.tmp2.Mul(z, z)
|
||||
c.tmp3.Sub(c.tmp1, c.tmp2)
|
||||
outx.Mul(c.tmp3, c.tmp3)
|
||||
outx.Mod(outx, p)
|
||||
|
||||
// z₂ = 4xz·(x² + Axz + z²)
|
||||
c.tmp3.Add(c.tmp1, c.tmp2)
|
||||
c.tmp1.Mul(x, z)
|
||||
c.tmp2.Mul(c.tmp1, a)
|
||||
outz.Add(c.tmp3, c.tmp2)
|
||||
c.tmp2.Lsh(c.tmp1, 2)
|
||||
outz.Mul(outz, c.tmp2)
|
||||
outz.Mod(outz, p)
|
||||
|
||||
return
|
||||
}
|
||||
|
||||
func scalarMult(out, in, base *[32]byte) {
|
||||
var baseReversed, inCopy [32]byte
|
||||
for i := 0; i < 32; i++ {
|
||||
baseReversed[31-i] = base[i]
|
||||
inCopy[i] = in[i]
|
||||
}
|
||||
|
||||
inCopy[31] &= 127
|
||||
inCopy[31] |= 64
|
||||
inCopy[0] &= 248
|
||||
|
||||
c := &context{new(big.Int), new(big.Int), new(big.Int), new(big.Int), nil}
|
||||
c.x1 = new(big.Int).SetBytes(baseReversed[:])
|
||||
|
||||
x1 := new(big.Int).SetInt64(1)
|
||||
z1 := new(big.Int)
|
||||
x2 := new(big.Int).Set(c.x1)
|
||||
z2 := new(big.Int).SetInt64(1)
|
||||
outx := new(big.Int)
|
||||
outz := new(big.Int)
|
||||
|
||||
for i := 0; i < 32; i++ {
|
||||
b := inCopy[31-i]
|
||||
for j := 0; j < 8; j++ {
|
||||
if b&0x80 != 0 {
|
||||
c.add(outx, outz, x1, z1, x2, z2)
|
||||
x1, z1, outx, outz = outx, outz, x1, z1
|
||||
c.double(outx, outz, x2, z2)
|
||||
x2, z2, outx, outz = outx, outz, x2, z2
|
||||
} else {
|
||||
c.add(outx, outz, x1, z1, x2, z2)
|
||||
x2, z2, outx, outz = outx, outz, x2, z2
|
||||
c.double(outx, outz, x1, z1)
|
||||
x1, z1, outx, outz = outx, outz, x1, z1
|
||||
}
|
||||
b <<= 1
|
||||
}
|
||||
}
|
||||
|
||||
c.tmp1.Exp(z1, pMinus2, p)
|
||||
c.tmp2.Mul(x1, c.tmp1)
|
||||
c.tmp3.Mod(c.tmp2, p)
|
||||
|
||||
outReversed := c.tmp3.Bytes()
|
||||
for i := 0; i < len(outReversed); i++ {
|
||||
out[i] = outReversed[len(outReversed)-(1+i)]
|
||||
}
|
||||
for i := len(outReversed); i < 32; i++ {
|
||||
out[i] = 0
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,29 @@
|
||||
// Copyright 2012 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 curve25519
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"testing"
|
||||
)
|
||||
|
||||
const expectedHex = "89161fde887b2b53de549af483940106ecc114d6982daa98256de23bdf77661a"
|
||||
|
||||
func TestBaseScalarMult(t *testing.T) {
|
||||
var a, b [32]byte
|
||||
in := &a
|
||||
out := &b
|
||||
a[0] = 1
|
||||
|
||||
for i := 0; i < 200; i++ {
|
||||
ScalarBaseMult(out, in)
|
||||
in, out = out, in
|
||||
}
|
||||
|
||||
result := fmt.Sprintf("%x", in[:])
|
||||
if result != expectedHex {
|
||||
t.Errorf("incorrect result: got %s, want %s", result, expectedHex)
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,23 @@
|
||||
// Copyright 2012 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 curve25519 provides an implementation of scalar multiplication on
|
||||
// the elliptic curve known as curve25519. See http://cr.yp.to/ecdh.html
|
||||
package curve25519
|
||||
|
||||
// basePoint is the x coordinate of the generator of the curve.
|
||||
var basePoint = [32]byte{9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
|
||||
|
||||
// ScalarMult sets dst to the product in*base where dst and base are the x
|
||||
// coordinates of group points and all values are in little-endian form.
|
||||
func ScalarMult(dst, in, base *[32]byte) {
|
||||
scalarMult(dst, in, base)
|
||||
}
|
||||
|
||||
// ScalarBaseMult sets dst to the product in*base where dst and base are the x
|
||||
// coordinates of group points, base is the standard generator and all values
|
||||
// are in little-endian form.
|
||||
func ScalarBaseMult(dst, in *[32]byte) {
|
||||
ScalarMult(dst, in, &basePoint)
|
||||
}
|
||||
@@ -0,0 +1,92 @@
|
||||
// Copyright 2012 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.
|
||||
|
||||
// This code was translated into a form compatible with 6a from the public
|
||||
// domain sources in SUPERCOP: http://bench.cr.yp.to/supercop.html
|
||||
|
||||
// func freeze(inout *[5]uint64)
|
||||
TEXT ·freeze(SB),7,$96-8
|
||||
MOVQ inout+0(FP), DI
|
||||
|
||||
MOVQ SP,R11
|
||||
MOVQ $31,CX
|
||||
NOTQ CX
|
||||
ANDQ CX,SP
|
||||
ADDQ $32,SP
|
||||
|
||||
MOVQ R11,0(SP)
|
||||
MOVQ R12,8(SP)
|
||||
MOVQ R13,16(SP)
|
||||
MOVQ R14,24(SP)
|
||||
MOVQ R15,32(SP)
|
||||
MOVQ BX,40(SP)
|
||||
MOVQ BP,48(SP)
|
||||
MOVQ 0(DI),SI
|
||||
MOVQ 8(DI),DX
|
||||
MOVQ 16(DI),CX
|
||||
MOVQ 24(DI),R8
|
||||
MOVQ 32(DI),R9
|
||||
MOVQ ·REDMASK51(SB),AX
|
||||
MOVQ AX,R10
|
||||
SUBQ $18,R10
|
||||
MOVQ $3,R11
|
||||
REDUCELOOP:
|
||||
MOVQ SI,R12
|
||||
SHRQ $51,R12
|
||||
ANDQ AX,SI
|
||||
ADDQ R12,DX
|
||||
MOVQ DX,R12
|
||||
SHRQ $51,R12
|
||||
ANDQ AX,DX
|
||||
ADDQ R12,CX
|
||||
MOVQ CX,R12
|
||||
SHRQ $51,R12
|
||||
ANDQ AX,CX
|
||||
ADDQ R12,R8
|
||||
MOVQ R8,R12
|
||||
SHRQ $51,R12
|
||||
ANDQ AX,R8
|
||||
ADDQ R12,R9
|
||||
MOVQ R9,R12
|
||||
SHRQ $51,R12
|
||||
ANDQ AX,R9
|
||||
IMUL3Q $19,R12,R12
|
||||
ADDQ R12,SI
|
||||
SUBQ $1,R11
|
||||
JA REDUCELOOP
|
||||
MOVQ $1,R12
|
||||
CMPQ R10,SI
|
||||
CMOVQLT R11,R12
|
||||
CMPQ AX,DX
|
||||
CMOVQNE R11,R12
|
||||
CMPQ AX,CX
|
||||
CMOVQNE R11,R12
|
||||
CMPQ AX,R8
|
||||
CMOVQNE R11,R12
|
||||
CMPQ AX,R9
|
||||
CMOVQNE R11,R12
|
||||
NEGQ R12
|
||||
ANDQ R12,AX
|
||||
ANDQ R12,R10
|
||||
SUBQ R10,SI
|
||||
SUBQ AX,DX
|
||||
SUBQ AX,CX
|
||||
SUBQ AX,R8
|
||||
SUBQ AX,R9
|
||||
MOVQ SI,0(DI)
|
||||
MOVQ DX,8(DI)
|
||||
MOVQ CX,16(DI)
|
||||
MOVQ R8,24(DI)
|
||||
MOVQ R9,32(DI)
|
||||
MOVQ 0(SP),R11
|
||||
MOVQ 8(SP),R12
|
||||
MOVQ 16(SP),R13
|
||||
MOVQ 24(SP),R14
|
||||
MOVQ 32(SP),R15
|
||||
MOVQ 40(SP),BX
|
||||
MOVQ 48(SP),BP
|
||||
MOVQ R11,SP
|
||||
MOVQ DI,AX
|
||||
MOVQ SI,DX
|
||||
RET
|
||||
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,223 @@
|
||||
// Copyright 2012 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 curve25519
|
||||
|
||||
// These functions are implemented in the .s files. The names of the functions
|
||||
// in the rest of the file are also taken from the SUPERCOP sources to help
|
||||
// people following along.
|
||||
func cswap(*[5]uint64, uint64)
|
||||
func ladderstep(*[5][5]uint64)
|
||||
func freeze(inout *[5]uint64)
|
||||
func mul(dest, a, b *[5]uint64)
|
||||
func square(out, in *[5]uint64)
|
||||
|
||||
// mladder uses a Montgomery ladder to calculate (xr/zr) *= s.
|
||||
func mladder(xr, zr *[5]uint64, s *[32]byte) {
|
||||
var work [5][5]uint64
|
||||
|
||||
work[0] = *xr
|
||||
setint(&work[1], 1)
|
||||
setint(&work[2], 0)
|
||||
work[3] = *xr
|
||||
setint(&work[4], 1)
|
||||
|
||||
j := uint(6)
|
||||
var prevbit byte
|
||||
|
||||
for i := 31; i >= 0; i-- {
|
||||
for j < 8 {
|
||||
bit := ((*s)[i] >> j) & 1
|
||||
swap := bit ^ prevbit
|
||||
prevbit = bit
|
||||
cswap(&work[1], uint64(swap))
|
||||
ladderstep(&work)
|
||||
j--
|
||||
}
|
||||
j = 7
|
||||
}
|
||||
|
||||
*xr = work[1]
|
||||
*zr = work[2]
|
||||
}
|
||||
|
||||
func scalarMult(out, in, base *[32]byte) {
|
||||
var e [32]byte
|
||||
copy(e[:], (*in)[:])
|
||||
e[0] &= 248
|
||||
e[31] &= 127
|
||||
e[31] |= 64
|
||||
|
||||
var t, z [5]uint64
|
||||
unpack(&t, base)
|
||||
mladder(&t, &z, &e)
|
||||
invert(&z, &z)
|
||||
mul(&t, &t, &z)
|
||||
pack(out, &t)
|
||||
}
|
||||
|
||||
func setint(r *[5]uint64, v uint64) {
|
||||
r[0] = v
|
||||
r[1] = 0
|
||||
r[2] = 0
|
||||
r[3] = 0
|
||||
r[4] = 0
|
||||
}
|
||||
|
||||
// unpack sets r = x where r consists of 5, 51-bit limbs in little-endian
|
||||
// order.
|
||||
func unpack(r *[5]uint64, x *[32]byte) {
|
||||
r[0] = uint64(x[0]) |
|
||||
uint64(x[1])<<8 |
|
||||
uint64(x[2])<<16 |
|
||||
uint64(x[3])<<24 |
|
||||
uint64(x[4])<<32 |
|
||||
uint64(x[5])<<40 |
|
||||
uint64(x[6]&7)<<48
|
||||
|
||||
r[1] = uint64(x[6])>>3 |
|
||||
uint64(x[7])<<5 |
|
||||
uint64(x[8])<<13 |
|
||||
uint64(x[9])<<21 |
|
||||
uint64(x[10])<<29 |
|
||||
uint64(x[11])<<37 |
|
||||
uint64(x[12]&63)<<45
|
||||
|
||||
r[2] = uint64(x[12])>>6 |
|
||||
uint64(x[13])<<2 |
|
||||
uint64(x[14])<<10 |
|
||||
uint64(x[15])<<18 |
|
||||
uint64(x[16])<<26 |
|
||||
uint64(x[17])<<34 |
|
||||
uint64(x[18])<<42 |
|
||||
uint64(x[19]&1)<<50
|
||||
|
||||
r[3] = uint64(x[19])>>1 |
|
||||
uint64(x[20])<<7 |
|
||||
uint64(x[21])<<15 |
|
||||
uint64(x[22])<<23 |
|
||||
uint64(x[23])<<31 |
|
||||
uint64(x[24])<<39 |
|
||||
uint64(x[25]&15)<<47
|
||||
|
||||
r[4] = uint64(x[25])>>4 |
|
||||
uint64(x[26])<<4 |
|
||||
uint64(x[27])<<12 |
|
||||
uint64(x[28])<<20 |
|
||||
uint64(x[29])<<28 |
|
||||
uint64(x[30])<<36 |
|
||||
uint64(x[31]&127)<<44
|
||||
}
|
||||
|
||||
// pack sets out = x where out is the usual, little-endian form of the 5,
|
||||
// 51-bit limbs in x.
|
||||
func pack(out *[32]byte, x *[5]uint64) {
|
||||
t := *x
|
||||
freeze(&t)
|
||||
|
||||
out[0] = byte(t[0])
|
||||
out[1] = byte(t[0] >> 8)
|
||||
out[2] = byte(t[0] >> 16)
|
||||
out[3] = byte(t[0] >> 24)
|
||||
out[4] = byte(t[0] >> 32)
|
||||
out[5] = byte(t[0] >> 40)
|
||||
out[6] = byte(t[0] >> 48)
|
||||
|
||||
out[6] ^= byte(t[1]<<3) & 0xf8
|
||||
out[7] = byte(t[1] >> 5)
|
||||
out[8] = byte(t[1] >> 13)
|
||||
out[9] = byte(t[1] >> 21)
|
||||
out[10] = byte(t[1] >> 29)
|
||||
out[11] = byte(t[1] >> 37)
|
||||
out[12] = byte(t[1] >> 45)
|
||||
|
||||
out[12] ^= byte(t[2]<<6) & 0xc0
|
||||
out[13] = byte(t[2] >> 2)
|
||||
out[14] = byte(t[2] >> 10)
|
||||
out[15] = byte(t[2] >> 18)
|
||||
out[16] = byte(t[2] >> 26)
|
||||
out[17] = byte(t[2] >> 34)
|
||||
out[18] = byte(t[2] >> 42)
|
||||
out[19] = byte(t[2] >> 50)
|
||||
|
||||
out[19] ^= byte(t[3]<<1) & 0xfe
|
||||
out[20] = byte(t[3] >> 7)
|
||||
out[21] = byte(t[3] >> 15)
|
||||
out[22] = byte(t[3] >> 23)
|
||||
out[23] = byte(t[3] >> 31)
|
||||
out[24] = byte(t[3] >> 39)
|
||||
out[25] = byte(t[3] >> 47)
|
||||
|
||||
out[25] ^= byte(t[4]<<4) & 0xf0
|
||||
out[26] = byte(t[4] >> 4)
|
||||
out[27] = byte(t[4] >> 12)
|
||||
out[28] = byte(t[4] >> 20)
|
||||
out[29] = byte(t[4] >> 28)
|
||||
out[30] = byte(t[4] >> 36)
|
||||
out[31] = byte(t[4] >> 44)
|
||||
}
|
||||
|
||||
// invert calculates r = x^-1 mod p using Fermat's little theorem.
|
||||
func invert(r *[5]uint64, x *[5]uint64) {
|
||||
var z2, z9, z11, z2_5_0, z2_10_0, z2_20_0, z2_50_0, z2_100_0, t [5]uint64
|
||||
|
||||
square(&z2, x) /* 2 */
|
||||
square(&t, &z2) /* 4 */
|
||||
square(&t, &t) /* 8 */
|
||||
mul(&z9, &t, x) /* 9 */
|
||||
mul(&z11, &z9, &z2) /* 11 */
|
||||
square(&t, &z11) /* 22 */
|
||||
mul(&z2_5_0, &t, &z9) /* 2^5 - 2^0 = 31 */
|
||||
|
||||
square(&t, &z2_5_0) /* 2^6 - 2^1 */
|
||||
for i := 1; i < 5; i++ { /* 2^20 - 2^10 */
|
||||
square(&t, &t)
|
||||
}
|
||||
mul(&z2_10_0, &t, &z2_5_0) /* 2^10 - 2^0 */
|
||||
|
||||
square(&t, &z2_10_0) /* 2^11 - 2^1 */
|
||||
for i := 1; i < 10; i++ { /* 2^20 - 2^10 */
|
||||
square(&t, &t)
|
||||
}
|
||||
mul(&z2_20_0, &t, &z2_10_0) /* 2^20 - 2^0 */
|
||||
|
||||
square(&t, &z2_20_0) /* 2^21 - 2^1 */
|
||||
for i := 1; i < 20; i++ { /* 2^40 - 2^20 */
|
||||
square(&t, &t)
|
||||
}
|
||||
mul(&t, &t, &z2_20_0) /* 2^40 - 2^0 */
|
||||
|
||||
square(&t, &t) /* 2^41 - 2^1 */
|
||||
for i := 1; i < 10; i++ { /* 2^50 - 2^10 */
|
||||
square(&t, &t)
|
||||
}
|
||||
mul(&z2_50_0, &t, &z2_10_0) /* 2^50 - 2^0 */
|
||||
|
||||
square(&t, &z2_50_0) /* 2^51 - 2^1 */
|
||||
for i := 1; i < 50; i++ { /* 2^100 - 2^50 */
|
||||
square(&t, &t)
|
||||
}
|
||||
mul(&z2_100_0, &t, &z2_50_0) /* 2^100 - 2^0 */
|
||||
|
||||
square(&t, &z2_100_0) /* 2^101 - 2^1 */
|
||||
for i := 1; i < 100; i++ { /* 2^200 - 2^100 */
|
||||
square(&t, &t)
|
||||
}
|
||||
mul(&t, &t, &z2_100_0) /* 2^200 - 2^0 */
|
||||
|
||||
square(&t, &t) /* 2^201 - 2^1 */
|
||||
for i := 1; i < 50; i++ { /* 2^250 - 2^50 */
|
||||
square(&t, &t)
|
||||
}
|
||||
mul(&t, &t, &z2_50_0) /* 2^250 - 2^0 */
|
||||
|
||||
square(&t, &t) /* 2^251 - 2^1 */
|
||||
square(&t, &t) /* 2^252 - 2^2 */
|
||||
square(&t, &t) /* 2^253 - 2^3 */
|
||||
|
||||
square(&t, &t) /* 2^254 - 2^4 */
|
||||
|
||||
square(&t, &t) /* 2^255 - 2^5 */
|
||||
mul(r, &t, &z11) /* 2^255 - 21 */
|
||||
}
|
||||
@@ -0,0 +1,189 @@
|
||||
// Copyright 2012 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.
|
||||
|
||||
// This code was translated into a form compatible with 6a from the public
|
||||
// domain sources in SUPERCOP: http://bench.cr.yp.to/supercop.html
|
||||
|
||||
// func mul(dest, a, b *[5]uint64)
|
||||
TEXT ·mul(SB),0,$128-24
|
||||
MOVQ dest+0(FP), DI
|
||||
MOVQ a+8(FP), SI
|
||||
MOVQ b+16(FP), DX
|
||||
|
||||
MOVQ SP,R11
|
||||
MOVQ $31,CX
|
||||
NOTQ CX
|
||||
ANDQ CX,SP
|
||||
ADDQ $32,SP
|
||||
|
||||
MOVQ R11,0(SP)
|
||||
MOVQ R12,8(SP)
|
||||
MOVQ R13,16(SP)
|
||||
MOVQ R14,24(SP)
|
||||
MOVQ R15,32(SP)
|
||||
MOVQ BX,40(SP)
|
||||
MOVQ BP,48(SP)
|
||||
MOVQ DI,56(SP)
|
||||
MOVQ DX,CX
|
||||
MOVQ 24(SI),DX
|
||||
IMUL3Q $19,DX,AX
|
||||
MOVQ AX,64(SP)
|
||||
MULQ 16(CX)
|
||||
MOVQ AX,R8
|
||||
MOVQ DX,R9
|
||||
MOVQ 32(SI),DX
|
||||
IMUL3Q $19,DX,AX
|
||||
MOVQ AX,72(SP)
|
||||
MULQ 8(CX)
|
||||
ADDQ AX,R8
|
||||
ADCQ DX,R9
|
||||
MOVQ 0(SI),AX
|
||||
MULQ 0(CX)
|
||||
ADDQ AX,R8
|
||||
ADCQ DX,R9
|
||||
MOVQ 0(SI),AX
|
||||
MULQ 8(CX)
|
||||
MOVQ AX,R10
|
||||
MOVQ DX,R11
|
||||
MOVQ 0(SI),AX
|
||||
MULQ 16(CX)
|
||||
MOVQ AX,R12
|
||||
MOVQ DX,R13
|
||||
MOVQ 0(SI),AX
|
||||
MULQ 24(CX)
|
||||
MOVQ AX,R14
|
||||
MOVQ DX,R15
|
||||
MOVQ 0(SI),AX
|
||||
MULQ 32(CX)
|
||||
MOVQ AX,BX
|
||||
MOVQ DX,BP
|
||||
MOVQ 8(SI),AX
|
||||
MULQ 0(CX)
|
||||
ADDQ AX,R10
|
||||
ADCQ DX,R11
|
||||
MOVQ 8(SI),AX
|
||||
MULQ 8(CX)
|
||||
ADDQ AX,R12
|
||||
ADCQ DX,R13
|
||||
MOVQ 8(SI),AX
|
||||
MULQ 16(CX)
|
||||
ADDQ AX,R14
|
||||
ADCQ DX,R15
|
||||
MOVQ 8(SI),AX
|
||||
MULQ 24(CX)
|
||||
ADDQ AX,BX
|
||||
ADCQ DX,BP
|
||||
MOVQ 8(SI),DX
|
||||
IMUL3Q $19,DX,AX
|
||||
MULQ 32(CX)
|
||||
ADDQ AX,R8
|
||||
ADCQ DX,R9
|
||||
MOVQ 16(SI),AX
|
||||
MULQ 0(CX)
|
||||
ADDQ AX,R12
|
||||
ADCQ DX,R13
|
||||
MOVQ 16(SI),AX
|
||||
MULQ 8(CX)
|
||||
ADDQ AX,R14
|
||||
ADCQ DX,R15
|
||||
MOVQ 16(SI),AX
|
||||
MULQ 16(CX)
|
||||
ADDQ AX,BX
|
||||
ADCQ DX,BP
|
||||
MOVQ 16(SI),DX
|
||||
IMUL3Q $19,DX,AX
|
||||
MULQ 24(CX)
|
||||
ADDQ AX,R8
|
||||
ADCQ DX,R9
|
||||
MOVQ 16(SI),DX
|
||||
IMUL3Q $19,DX,AX
|
||||
MULQ 32(CX)
|
||||
ADDQ AX,R10
|
||||
ADCQ DX,R11
|
||||
MOVQ 24(SI),AX
|
||||
MULQ 0(CX)
|
||||
ADDQ AX,R14
|
||||
ADCQ DX,R15
|
||||
MOVQ 24(SI),AX
|
||||
MULQ 8(CX)
|
||||
ADDQ AX,BX
|
||||
ADCQ DX,BP
|
||||
MOVQ 64(SP),AX
|
||||
MULQ 24(CX)
|
||||
ADDQ AX,R10
|
||||
ADCQ DX,R11
|
||||
MOVQ 64(SP),AX
|
||||
MULQ 32(CX)
|
||||
ADDQ AX,R12
|
||||
ADCQ DX,R13
|
||||
MOVQ 32(SI),AX
|
||||
MULQ 0(CX)
|
||||
ADDQ AX,BX
|
||||
ADCQ DX,BP
|
||||
MOVQ 72(SP),AX
|
||||
MULQ 16(CX)
|
||||
ADDQ AX,R10
|
||||
ADCQ DX,R11
|
||||
MOVQ 72(SP),AX
|
||||
MULQ 24(CX)
|
||||
ADDQ AX,R12
|
||||
ADCQ DX,R13
|
||||
MOVQ 72(SP),AX
|
||||
MULQ 32(CX)
|
||||
ADDQ AX,R14
|
||||
ADCQ DX,R15
|
||||
MOVQ ·REDMASK51(SB),SI
|
||||
SHLQ $13,R9:R8
|
||||
ANDQ SI,R8
|
||||
SHLQ $13,R11:R10
|
||||
ANDQ SI,R10
|
||||
ADDQ R9,R10
|
||||
SHLQ $13,R13:R12
|
||||
ANDQ SI,R12
|
||||
ADDQ R11,R12
|
||||
SHLQ $13,R15:R14
|
||||
ANDQ SI,R14
|
||||
ADDQ R13,R14
|
||||
SHLQ $13,BP:BX
|
||||
ANDQ SI,BX
|
||||
ADDQ R15,BX
|
||||
IMUL3Q $19,BP,DX
|
||||
ADDQ DX,R8
|
||||
MOVQ R8,DX
|
||||
SHRQ $51,DX
|
||||
ADDQ R10,DX
|
||||
MOVQ DX,CX
|
||||
SHRQ $51,DX
|
||||
ANDQ SI,R8
|
||||
ADDQ R12,DX
|
||||
MOVQ DX,R9
|
||||
SHRQ $51,DX
|
||||
ANDQ SI,CX
|
||||
ADDQ R14,DX
|
||||
MOVQ DX,AX
|
||||
SHRQ $51,DX
|
||||
ANDQ SI,R9
|
||||
ADDQ BX,DX
|
||||
MOVQ DX,R10
|
||||
SHRQ $51,DX
|
||||
ANDQ SI,AX
|
||||
IMUL3Q $19,DX,DX
|
||||
ADDQ DX,R8
|
||||
ANDQ SI,R10
|
||||
MOVQ R8,0(DI)
|
||||
MOVQ CX,8(DI)
|
||||
MOVQ R9,16(DI)
|
||||
MOVQ AX,24(DI)
|
||||
MOVQ R10,32(DI)
|
||||
MOVQ 0(SP),R11
|
||||
MOVQ 8(SP),R12
|
||||
MOVQ 16(SP),R13
|
||||
MOVQ 24(SP),R14
|
||||
MOVQ 32(SP),R15
|
||||
MOVQ 40(SP),BX
|
||||
MOVQ 48(SP),BP
|
||||
MOVQ R11,SP
|
||||
MOVQ DI,AX
|
||||
MOVQ SI,DX
|
||||
RET
|
||||
@@ -0,0 +1,151 @@
|
||||
// Copyright 2012 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.
|
||||
|
||||
// This code was translated into a form compatible with 6a from the public
|
||||
// domain sources in SUPERCOP: http://bench.cr.yp.to/supercop.html
|
||||
|
||||
// func square(dest, src *[5]uint64)
|
||||
TEXT ·square(SB),7,$96-16
|
||||
MOVQ dest+0(FP), DI
|
||||
MOVQ src+8(FP), SI
|
||||
|
||||
MOVQ SP,R11
|
||||
MOVQ $31,CX
|
||||
NOTQ CX
|
||||
ANDQ CX,SP
|
||||
ADDQ $32, SP
|
||||
|
||||
MOVQ R11,0(SP)
|
||||
MOVQ R12,8(SP)
|
||||
MOVQ R13,16(SP)
|
||||
MOVQ R14,24(SP)
|
||||
MOVQ R15,32(SP)
|
||||
MOVQ BX,40(SP)
|
||||
MOVQ BP,48(SP)
|
||||
MOVQ 0(SI),AX
|
||||
MULQ 0(SI)
|
||||
MOVQ AX,CX
|
||||
MOVQ DX,R8
|
||||
MOVQ 0(SI),AX
|
||||
SHLQ $1,AX
|
||||
MULQ 8(SI)
|
||||
MOVQ AX,R9
|
||||
MOVQ DX,R10
|
||||
MOVQ 0(SI),AX
|
||||
SHLQ $1,AX
|
||||
MULQ 16(SI)
|
||||
MOVQ AX,R11
|
||||
MOVQ DX,R12
|
||||
MOVQ 0(SI),AX
|
||||
SHLQ $1,AX
|
||||
MULQ 24(SI)
|
||||
MOVQ AX,R13
|
||||
MOVQ DX,R14
|
||||
MOVQ 0(SI),AX
|
||||
SHLQ $1,AX
|
||||
MULQ 32(SI)
|
||||
MOVQ AX,R15
|
||||
MOVQ DX,BX
|
||||
MOVQ 8(SI),AX
|
||||
MULQ 8(SI)
|
||||
ADDQ AX,R11
|
||||
ADCQ DX,R12
|
||||
MOVQ 8(SI),AX
|
||||
SHLQ $1,AX
|
||||
MULQ 16(SI)
|
||||
ADDQ AX,R13
|
||||
ADCQ DX,R14
|
||||
MOVQ 8(SI),AX
|
||||
SHLQ $1,AX
|
||||
MULQ 24(SI)
|
||||
ADDQ AX,R15
|
||||
ADCQ DX,BX
|
||||
MOVQ 8(SI),DX
|
||||
IMUL3Q $38,DX,AX
|
||||
MULQ 32(SI)
|
||||
ADDQ AX,CX
|
||||
ADCQ DX,R8
|
||||
MOVQ 16(SI),AX
|
||||
MULQ 16(SI)
|
||||
ADDQ AX,R15
|
||||
ADCQ DX,BX
|
||||
MOVQ 16(SI),DX
|
||||
IMUL3Q $38,DX,AX
|
||||
MULQ 24(SI)
|
||||
ADDQ AX,CX
|
||||
ADCQ DX,R8
|
||||
MOVQ 16(SI),DX
|
||||
IMUL3Q $38,DX,AX
|
||||
MULQ 32(SI)
|
||||
ADDQ AX,R9
|
||||
ADCQ DX,R10
|
||||
MOVQ 24(SI),DX
|
||||
IMUL3Q $19,DX,AX
|
||||
MULQ 24(SI)
|
||||
ADDQ AX,R9
|
||||
ADCQ DX,R10
|
||||
MOVQ 24(SI),DX
|
||||
IMUL3Q $38,DX,AX
|
||||
MULQ 32(SI)
|
||||
ADDQ AX,R11
|
||||
ADCQ DX,R12
|
||||
MOVQ 32(SI),DX
|
||||
IMUL3Q $19,DX,AX
|
||||
MULQ 32(SI)
|
||||
ADDQ AX,R13
|
||||
ADCQ DX,R14
|
||||
MOVQ ·REDMASK51(SB),SI
|
||||
SHLQ $13,R8:CX
|
||||
ANDQ SI,CX
|
||||
SHLQ $13,R10:R9
|
||||
ANDQ SI,R9
|
||||
ADDQ R8,R9
|
||||
SHLQ $13,R12:R11
|
||||
ANDQ SI,R11
|
||||
ADDQ R10,R11
|
||||
SHLQ $13,R14:R13
|
||||
ANDQ SI,R13
|
||||
ADDQ R12,R13
|
||||
SHLQ $13,BX:R15
|
||||
ANDQ SI,R15
|
||||
ADDQ R14,R15
|
||||
IMUL3Q $19,BX,DX
|
||||
ADDQ DX,CX
|
||||
MOVQ CX,DX
|
||||
SHRQ $51,DX
|
||||
ADDQ R9,DX
|
||||
ANDQ SI,CX
|
||||
MOVQ DX,R8
|
||||
SHRQ $51,DX
|
||||
ADDQ R11,DX
|
||||
ANDQ SI,R8
|
||||
MOVQ DX,R9
|
||||
SHRQ $51,DX
|
||||
ADDQ R13,DX
|
||||
ANDQ SI,R9
|
||||
MOVQ DX,AX
|
||||
SHRQ $51,DX
|
||||
ADDQ R15,DX
|
||||
ANDQ SI,AX
|
||||
MOVQ DX,R10
|
||||
SHRQ $51,DX
|
||||
IMUL3Q $19,DX,DX
|
||||
ADDQ DX,CX
|
||||
ANDQ SI,R10
|
||||
MOVQ CX,0(DI)
|
||||
MOVQ R8,8(DI)
|
||||
MOVQ R9,16(DI)
|
||||
MOVQ AX,24(DI)
|
||||
MOVQ R10,32(DI)
|
||||
MOVQ 0(SP),R11
|
||||
MOVQ 8(SP),R12
|
||||
MOVQ 16(SP),R13
|
||||
MOVQ 24(SP),R14
|
||||
MOVQ 32(SP),R15
|
||||
MOVQ 40(SP),BX
|
||||
MOVQ 48(SP),BP
|
||||
MOVQ R11,SP
|
||||
MOVQ DI,AX
|
||||
MOVQ SI,DX
|
||||
RET
|
||||
Reference in New Issue
Block a user