mirror of
https://github.com/emmansun/gmsm.git
synced 2025-04-28 05:06:18 +08:00
store SM2 precomputed basepoint table in source
This commit is contained in:
parent
8a2098aa26
commit
e1fa144e15
96
sm2/gen_p256_table.go
Normal file
96
sm2/gen_p256_table.go
Normal file
@ -0,0 +1,96 @@
|
|||||||
|
// Copyright 2021 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.
|
||||||
|
|
||||||
|
//go:build amd64
|
||||||
|
// +build amd64
|
||||||
|
|
||||||
|
package sm2
|
||||||
|
|
||||||
|
import (
|
||||||
|
"bytes"
|
||||||
|
"encoding/binary"
|
||||||
|
"fmt"
|
||||||
|
"go/format"
|
||||||
|
)
|
||||||
|
|
||||||
|
func GenTables() {
|
||||||
|
buf := new(bytes.Buffer)
|
||||||
|
fmt.Fprint(buf, `
|
||||||
|
// Generated by gen_p256_table.go. DO NOT EDIT.
|
||||||
|
//go:build amd64
|
||||||
|
// +build amd64
|
||||||
|
package sm2
|
||||||
|
`[1:])
|
||||||
|
|
||||||
|
// Generate precomputed p256 tables.
|
||||||
|
var pre [43][32 * 8]uint64
|
||||||
|
basePoint := []uint64{
|
||||||
|
0x61328990f418029e, 0x3e7981eddca6c050, 0xd6a1ed99ac24c3c3, 0x91167a5ee1c13b05,
|
||||||
|
0xc1354e593c2d0ddd, 0xc1f5e5788d3295fa, 0x8d4cfb066e2a48f8, 0x63cd65d481d735bd,
|
||||||
|
0x0000000000000001, 0x00000000ffffffff, 0x0000000000000000, 0x0000000100000000,
|
||||||
|
}
|
||||||
|
t1 := make([]uint64, 12)
|
||||||
|
t2 := make([]uint64, 12)
|
||||||
|
copy(t2, basePoint)
|
||||||
|
zInv := make([]uint64, 4)
|
||||||
|
zInvSq := make([]uint64, 4)
|
||||||
|
for j := 0; j < 32; j++ {
|
||||||
|
copy(t1, t2)
|
||||||
|
for i := 0; i < 43; i++ {
|
||||||
|
// The window size is 6 so we need to double 6 times.
|
||||||
|
if i != 0 {
|
||||||
|
for k := 0; k < 6; k++ {
|
||||||
|
p256PointDoubleAsm(t1, t1)
|
||||||
|
}
|
||||||
|
}
|
||||||
|
// Convert the point to affine form. (Its values are
|
||||||
|
// still in Montgomery form however.)
|
||||||
|
p256Inverse(zInv, t1[8:12])
|
||||||
|
p256Sqr(zInvSq, zInv, 1)
|
||||||
|
p256Mul(zInv, zInv, zInvSq)
|
||||||
|
p256Mul(t1[:4], t1[:4], zInvSq)
|
||||||
|
p256Mul(t1[4:8], t1[4:8], zInv)
|
||||||
|
copy(t1[8:12], basePoint[8:12])
|
||||||
|
// Update the table entry
|
||||||
|
copy(pre[i][j*8:], t1[:8])
|
||||||
|
}
|
||||||
|
if j == 0 {
|
||||||
|
p256PointDoubleAsm(t2, basePoint)
|
||||||
|
} else {
|
||||||
|
p256PointAddAsm(t2, t2, basePoint)
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
fmt.Fprint(buf, "const p256Precomputed = \"\" +\n\n")
|
||||||
|
|
||||||
|
// Dump the precomputed tables, flattened, little-endian.
|
||||||
|
// These tables are used directly by assembly on little-endian platforms.
|
||||||
|
// Putting the data in a const string lets it be stored readonly.
|
||||||
|
for i := range &pre {
|
||||||
|
for j, v := range &pre[i] {
|
||||||
|
fmt.Fprintf(buf, "\"")
|
||||||
|
var u8 [8]byte
|
||||||
|
binary.LittleEndian.PutUint64(u8[:], v)
|
||||||
|
for _, b := range &u8 {
|
||||||
|
fmt.Fprintf(buf, "\\x%02x", b)
|
||||||
|
}
|
||||||
|
fmt.Fprintf(buf, "\"")
|
||||||
|
if i < len(pre)-1 || j < len(pre[i])-1 {
|
||||||
|
fmt.Fprint(buf, "+")
|
||||||
|
}
|
||||||
|
if j%8 == 7 {
|
||||||
|
fmt.Fprint(buf, "\n")
|
||||||
|
}
|
||||||
|
}
|
||||||
|
fmt.Fprint(buf, "\n")
|
||||||
|
}
|
||||||
|
|
||||||
|
src := buf.Bytes()
|
||||||
|
fmtsrc, fmterr := format.Source(src)
|
||||||
|
// If formatting failed, keep the original source for debugging.
|
||||||
|
if fmterr == nil {
|
||||||
|
src = fmtsrc
|
||||||
|
}
|
||||||
|
fmt.Println(string(src))
|
||||||
|
}
|
@ -15,7 +15,6 @@ package sm2
|
|||||||
import (
|
import (
|
||||||
"crypto/elliptic"
|
"crypto/elliptic"
|
||||||
"math/big"
|
"math/big"
|
||||||
"sync"
|
|
||||||
)
|
)
|
||||||
|
|
||||||
type (
|
type (
|
||||||
@ -30,8 +29,6 @@ type (
|
|||||||
|
|
||||||
var (
|
var (
|
||||||
p256 p256Curve
|
p256 p256Curve
|
||||||
p256Precomputed *[43][32 * 8]uint64
|
|
||||||
precomputeOnce sync.Once
|
|
||||||
)
|
)
|
||||||
|
|
||||||
func initP256() {
|
func initP256() {
|
||||||
@ -82,7 +79,7 @@ func p256LittleToBig(res []byte, in []uint64)
|
|||||||
func p256Select(point, table []uint64, idx int)
|
func p256Select(point, table []uint64, idx int)
|
||||||
|
|
||||||
//go:noescape
|
//go:noescape
|
||||||
func p256SelectBase(point, table []uint64, idx int)
|
func p256SelectBase(point *[12]uint64, table string, idx int)
|
||||||
|
|
||||||
// Montgomery multiplication modulo Ord(G)
|
// Montgomery multiplication modulo Ord(G)
|
||||||
//go:noescape
|
//go:noescape
|
||||||
@ -446,52 +443,10 @@ func boothW6(in uint) (int, int) {
|
|||||||
return int(d), int(s & 1)
|
return int(d), int(s & 1)
|
||||||
}
|
}
|
||||||
|
|
||||||
// table[i][j] = (2^(6*i))*(j+1)*G mod P
|
|
||||||
func initTable() {
|
|
||||||
p256Precomputed = new([43][32 * 8]uint64)
|
|
||||||
|
|
||||||
t1 := make([]uint64, 12)
|
|
||||||
t2 := make([]uint64, 12)
|
|
||||||
copy(t2, basePoint)
|
|
||||||
|
|
||||||
zInv := make([]uint64, 4)
|
|
||||||
zInvSq := make([]uint64, 4)
|
|
||||||
for j := 0; j < 32; j++ {
|
|
||||||
copy(t1, t2)
|
|
||||||
for i := 0; i < 43; i++ {
|
|
||||||
// The window size is 6 so we need to double 6 times.
|
|
||||||
if i != 0 {
|
|
||||||
for k := 0; k < 6; k++ {
|
|
||||||
p256PointDoubleAsm(t1, t1)
|
|
||||||
}
|
|
||||||
}
|
|
||||||
// Convert the point to affine form. (Its values are
|
|
||||||
// still in Montgomery form however.)
|
|
||||||
p256Inverse(zInv, t1[8:12])
|
|
||||||
p256Sqr(zInvSq, zInv, 1)
|
|
||||||
p256Mul(zInv, zInv, zInvSq)
|
|
||||||
|
|
||||||
p256Mul(t1[:4], t1[:4], zInvSq)
|
|
||||||
p256Mul(t1[4:8], t1[4:8], zInv)
|
|
||||||
|
|
||||||
copy(t1[8:12], basePoint[8:12])
|
|
||||||
// Update the table entry
|
|
||||||
copy(p256Precomputed[i][j*8:], t1[:8])
|
|
||||||
}
|
|
||||||
if j == 0 {
|
|
||||||
p256PointDoubleAsm(t2, basePoint)
|
|
||||||
} else {
|
|
||||||
p256PointAddAsm(t2, t2, basePoint)
|
|
||||||
}
|
|
||||||
}
|
|
||||||
}
|
|
||||||
|
|
||||||
func (p *p256Point) p256BaseMult(scalar []uint64) {
|
func (p *p256Point) p256BaseMult(scalar []uint64) {
|
||||||
precomputeOnce.Do(initTable)
|
|
||||||
|
|
||||||
wvalue := (scalar[0] << 1) & 0x7f
|
wvalue := (scalar[0] << 1) & 0x7f
|
||||||
sel, sign := boothW6(uint(wvalue))
|
sel, sign := boothW6(uint(wvalue))
|
||||||
p256SelectBase(p.xyz[0:8], p256Precomputed[0][0:], sel)
|
p256SelectBase(&p.xyz, p256Precomputed, sel)
|
||||||
p256NegCond(p.xyz[4:8], sign)
|
p256NegCond(p.xyz[4:8], sign)
|
||||||
|
|
||||||
// (This is one, in the Montgomery domain.)
|
// (This is one, in the Montgomery domain.)
|
||||||
@ -518,7 +473,7 @@ func (p *p256Point) p256BaseMult(scalar []uint64) {
|
|||||||
}
|
}
|
||||||
index += 6
|
index += 6
|
||||||
sel, sign = boothW6(uint(wvalue))
|
sel, sign = boothW6(uint(wvalue))
|
||||||
p256SelectBase(t0.xyz[0:8], p256Precomputed[i][0:], sel)
|
p256SelectBase(&t0.xyz, p256Precomputed[i*32*8*8:], sel)
|
||||||
p256PointAddAffineAsm(p.xyz[0:12], p.xyz[0:12], t0.xyz[0:8], sign, sel, zero)
|
p256PointAddAffineAsm(p.xyz[0:12], p.xyz[0:12], t0.xyz[0:8], sign, sel, zero)
|
||||||
zero |= sel
|
zero |= sel
|
||||||
}
|
}
|
||||||
|
@ -773,10 +773,10 @@ loop_select:
|
|||||||
RET
|
RET
|
||||||
/* ---------------------------------------*/
|
/* ---------------------------------------*/
|
||||||
// Constant time point access to base point table.
|
// Constant time point access to base point table.
|
||||||
// func p256SelectBase(point, table []uint64, idx int)
|
// func p256SelectBase(point *[12]uint64, table string, idx int)
|
||||||
TEXT ·p256SelectBase(SB),NOSPLIT,$0
|
TEXT ·p256SelectBase(SB),NOSPLIT,$0
|
||||||
MOVQ idx+48(FP),AX
|
MOVQ idx+24(FP),AX
|
||||||
MOVQ table+24(FP),DI
|
MOVQ table+8(FP),DI
|
||||||
MOVQ point+0(FP),DX
|
MOVQ point+0(FP),DX
|
||||||
|
|
||||||
PXOR X15, X15 // X15 = 0
|
PXOR X15, X15 // X15 = 0
|
||||||
|
1425
sm2/p256_asm_table.go
Normal file
1425
sm2/p256_asm_table.go
Normal file
File diff suppressed because it is too large
Load Diff
65
sm2/p256_asm_table_test.go
Normal file
65
sm2/p256_asm_table_test.go
Normal file
@ -0,0 +1,65 @@
|
|||||||
|
// Copyright 2021 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.
|
||||||
|
|
||||||
|
//go:build amd64
|
||||||
|
// +build amd64
|
||||||
|
|
||||||
|
package sm2
|
||||||
|
|
||||||
|
import (
|
||||||
|
"encoding/binary"
|
||||||
|
"reflect"
|
||||||
|
"testing"
|
||||||
|
)
|
||||||
|
|
||||||
|
func TestP256PrecomputedTable(t *testing.T) {
|
||||||
|
|
||||||
|
basePoint := []uint64{
|
||||||
|
0x61328990f418029e, 0x3e7981eddca6c050, 0xd6a1ed99ac24c3c3, 0x91167a5ee1c13b05,
|
||||||
|
0xc1354e593c2d0ddd, 0xc1f5e5788d3295fa, 0x8d4cfb066e2a48f8, 0x63cd65d481d735bd,
|
||||||
|
0x0000000000000001, 0x00000000ffffffff, 0x0000000000000000, 0x0000000100000000,
|
||||||
|
}
|
||||||
|
t1 := make([]uint64, 12)
|
||||||
|
t2 := make([]uint64, 12)
|
||||||
|
copy(t2, basePoint)
|
||||||
|
|
||||||
|
zInv := make([]uint64, 4)
|
||||||
|
zInvSq := make([]uint64, 4)
|
||||||
|
for j := 0; j < 32; j++ {
|
||||||
|
copy(t1, t2)
|
||||||
|
for i := 0; i < 43; i++ {
|
||||||
|
// The window size is 6 so we need to double 6 times.
|
||||||
|
if i != 0 {
|
||||||
|
for k := 0; k < 6; k++ {
|
||||||
|
p256PointDoubleAsm(t1, t1)
|
||||||
|
}
|
||||||
|
}
|
||||||
|
// Convert the point to affine form. (Its values are
|
||||||
|
// still in Montgomery form however.)
|
||||||
|
p256Inverse(zInv, t1[8:12])
|
||||||
|
p256Sqr(zInvSq, zInv, 1)
|
||||||
|
p256Mul(zInv, zInv, zInvSq)
|
||||||
|
|
||||||
|
p256Mul(t1[:4], t1[:4], zInvSq)
|
||||||
|
p256Mul(t1[4:8], t1[4:8], zInv)
|
||||||
|
|
||||||
|
copy(t1[8:12], basePoint[8:12])
|
||||||
|
|
||||||
|
buf := make([]byte, 8*8)
|
||||||
|
for i, u := range t1[:8] {
|
||||||
|
binary.LittleEndian.PutUint64(buf[i*8:i*8+8], u)
|
||||||
|
}
|
||||||
|
start := i*32*8*8 + j*8*8
|
||||||
|
if got, want := p256Precomputed[start:start+64], string(buf); !reflect.DeepEqual(got, want) {
|
||||||
|
t.Fatalf("Unexpected table entry at [%d][%d:%d]: got %v, want %v", i, j*8, (j*8)+8, got, want)
|
||||||
|
}
|
||||||
|
}
|
||||||
|
if j == 0 {
|
||||||
|
p256PointDoubleAsm(t2, basePoint)
|
||||||
|
} else {
|
||||||
|
p256PointAddAsm(t2, t2, basePoint)
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
}
|
Loading…
x
Reference in New Issue
Block a user