1
0
mirror of https://github.com/henrydcase/nobs.git synced 2024-11-22 15:18:57 +00:00
nobs/dh/csidh/fp511.go

291 lines
8.1 KiB
Go
Raw Permalink Normal View History

// +build noasm amd64 arm64 ppc64le riscv64
cSIDH-511: (#26) Implementation of Commutative Supersingular Isogeny Diffie Hellman, based on "A faster way to CSIDH" paper (2018/782). * For fast isogeny calculation, implementation converts a curve from Montgomery to Edwards. All calculations are done on Edwards curve and then converted back to Montgomery. * As multiplication in a field Fp511 is most expensive operation the implementation contains multiple multiplications. It has most performant, assembly implementation which uses BMI2 and ADOX/ADCX instructions for modern CPUs. It also contains slower implementation which will run on older CPUs * Benchmarks (Intel SkyLake): BenchmarkGeneratePrivate 6459 172213 ns/op 0 B/op 0 allocs/op BenchmarkGenerateKeyPair 25 45800356 ns/op 0 B/op 0 allocs/op BenchmarkValidate 297 3915983 ns/op 0 B/op 0 allocs/op BenchmarkValidateRandom 184683 6231 ns/op 0 B/op 0 allocs/op BenchmarkValidateGenerated 25 48481306 ns/op 0 B/op 0 allocs/op BenchmarkDerive 19 60928763 ns/op 0 B/op 0 allocs/op BenchmarkDeriveGenerated 8 137342421 ns/op 0 B/op 0 allocs/op BenchmarkXMul 2311 494267 ns/op 1 B/op 0 allocs/op BenchmarkXAdd 2396754 501 ns/op 0 B/op 0 allocs/op BenchmarkXDbl 2072690 571 ns/op 0 B/op 0 allocs/op BenchmarkIsom 78004 15171 ns/op 0 B/op 0 allocs/op BenchmarkFp512Sub 224635152 5.33 ns/op 0 B/op 0 allocs/op BenchmarkFp512Mul 246633255 4.90 ns/op 0 B/op 0 allocs/op BenchmarkCSwap 233228547 5.10 ns/op 0 B/op 0 allocs/op BenchmarkAddRdc 87348240 12.6 ns/op 0 B/op 0 allocs/op BenchmarkSubRdc 95112787 11.7 ns/op 0 B/op 0 allocs/op BenchmarkModExpRdc 25436 46878 ns/op 0 B/op 0 allocs/op BenchmarkMulBmiAsm 19527573 60.1 ns/op 0 B/op 0 allocs/op BenchmarkMulGeneric 7117650 164 ns/op 0 B/op 0 allocs/op * Go code has very similar performance when compared to C implementation. Results from sidh_torturer (4e2996e12d68364761064341cbe1d1b47efafe23) github.com:henrydcase/sidh-torture/csidh | TestName |Go | C | |------------------|----------|----------| |TestSharedSecret | 57.95774 | 57.91092 | |TestKeyGeneration | 62.23614 | 58.12980 | |TestSharedSecret | 55.28988 | 57.23132 | |TestKeyGeneration | 61.68745 | 58.66396 | |TestSharedSecret | 63.19408 | 58.64774 | |TestKeyGeneration | 62.34022 | 61.62539 | |TestSharedSecret | 62.85453 | 68.74503 | |TestKeyGeneration | 52.58518 | 58.40115 | |TestSharedSecret | 50.77081 | 61.91699 | |TestKeyGeneration | 59.91843 | 61.09266 | |TestSharedSecret | 59.97962 | 62.98151 | |TestKeyGeneration | 64.57525 | 56.22863 | |TestSharedSecret | 56.40521 | 55.77447 | |TestKeyGeneration | 67.85850 | 58.52604 | |TestSharedSecret | 60.54290 | 65.14052 | |TestKeyGeneration | 65.45766 | 58.42823 | On average Go implementation is 2% faster.
2019-11-24 03:39:35 +00:00
package csidh
import (
"math/bits"
"golang.org/x/sys/cpu"
)
// CPU Capabilities. Those flags are referred by assembly code. According to
// https://github.com/golang/go/issues/28230, variables referred from the
// assembly must be in the same package.
// We declare variables not constants, in order to facilitate testing.
var (
// Signals support for BMI2 (MULX)
2020-05-14 00:48:43 +01:00
hasBMI2 = cpu.X86.HasBMI2 //nolint
cSIDH-511: (#26) Implementation of Commutative Supersingular Isogeny Diffie Hellman, based on "A faster way to CSIDH" paper (2018/782). * For fast isogeny calculation, implementation converts a curve from Montgomery to Edwards. All calculations are done on Edwards curve and then converted back to Montgomery. * As multiplication in a field Fp511 is most expensive operation the implementation contains multiple multiplications. It has most performant, assembly implementation which uses BMI2 and ADOX/ADCX instructions for modern CPUs. It also contains slower implementation which will run on older CPUs * Benchmarks (Intel SkyLake): BenchmarkGeneratePrivate 6459 172213 ns/op 0 B/op 0 allocs/op BenchmarkGenerateKeyPair 25 45800356 ns/op 0 B/op 0 allocs/op BenchmarkValidate 297 3915983 ns/op 0 B/op 0 allocs/op BenchmarkValidateRandom 184683 6231 ns/op 0 B/op 0 allocs/op BenchmarkValidateGenerated 25 48481306 ns/op 0 B/op 0 allocs/op BenchmarkDerive 19 60928763 ns/op 0 B/op 0 allocs/op BenchmarkDeriveGenerated 8 137342421 ns/op 0 B/op 0 allocs/op BenchmarkXMul 2311 494267 ns/op 1 B/op 0 allocs/op BenchmarkXAdd 2396754 501 ns/op 0 B/op 0 allocs/op BenchmarkXDbl 2072690 571 ns/op 0 B/op 0 allocs/op BenchmarkIsom 78004 15171 ns/op 0 B/op 0 allocs/op BenchmarkFp512Sub 224635152 5.33 ns/op 0 B/op 0 allocs/op BenchmarkFp512Mul 246633255 4.90 ns/op 0 B/op 0 allocs/op BenchmarkCSwap 233228547 5.10 ns/op 0 B/op 0 allocs/op BenchmarkAddRdc 87348240 12.6 ns/op 0 B/op 0 allocs/op BenchmarkSubRdc 95112787 11.7 ns/op 0 B/op 0 allocs/op BenchmarkModExpRdc 25436 46878 ns/op 0 B/op 0 allocs/op BenchmarkMulBmiAsm 19527573 60.1 ns/op 0 B/op 0 allocs/op BenchmarkMulGeneric 7117650 164 ns/op 0 B/op 0 allocs/op * Go code has very similar performance when compared to C implementation. Results from sidh_torturer (4e2996e12d68364761064341cbe1d1b47efafe23) github.com:henrydcase/sidh-torture/csidh | TestName |Go | C | |------------------|----------|----------| |TestSharedSecret | 57.95774 | 57.91092 | |TestKeyGeneration | 62.23614 | 58.12980 | |TestSharedSecret | 55.28988 | 57.23132 | |TestKeyGeneration | 61.68745 | 58.66396 | |TestSharedSecret | 63.19408 | 58.64774 | |TestKeyGeneration | 62.34022 | 61.62539 | |TestSharedSecret | 62.85453 | 68.74503 | |TestKeyGeneration | 52.58518 | 58.40115 | |TestSharedSecret | 50.77081 | 61.91699 | |TestKeyGeneration | 59.91843 | 61.09266 | |TestSharedSecret | 59.97962 | 62.98151 | |TestKeyGeneration | 64.57525 | 56.22863 | |TestSharedSecret | 56.40521 | 55.77447 | |TestKeyGeneration | 67.85850 | 58.52604 | |TestSharedSecret | 60.54290 | 65.14052 | |TestKeyGeneration | 65.45766 | 58.42823 | On average Go implementation is 2% faster.
2019-11-24 03:39:35 +00:00
// Signals support for ADX and BMI2
hasADXandBMI2 = cpu.X86.HasBMI2 && cpu.X86.HasADX
)
// Constant time select.
// if pick == 0xFF..FF (out = in1)
// if pick == 0 (out = in2)
2020-05-14 00:48:43 +01:00
// else out is undefined.
cSIDH-511: (#26) Implementation of Commutative Supersingular Isogeny Diffie Hellman, based on "A faster way to CSIDH" paper (2018/782). * For fast isogeny calculation, implementation converts a curve from Montgomery to Edwards. All calculations are done on Edwards curve and then converted back to Montgomery. * As multiplication in a field Fp511 is most expensive operation the implementation contains multiple multiplications. It has most performant, assembly implementation which uses BMI2 and ADOX/ADCX instructions for modern CPUs. It also contains slower implementation which will run on older CPUs * Benchmarks (Intel SkyLake): BenchmarkGeneratePrivate 6459 172213 ns/op 0 B/op 0 allocs/op BenchmarkGenerateKeyPair 25 45800356 ns/op 0 B/op 0 allocs/op BenchmarkValidate 297 3915983 ns/op 0 B/op 0 allocs/op BenchmarkValidateRandom 184683 6231 ns/op 0 B/op 0 allocs/op BenchmarkValidateGenerated 25 48481306 ns/op 0 B/op 0 allocs/op BenchmarkDerive 19 60928763 ns/op 0 B/op 0 allocs/op BenchmarkDeriveGenerated 8 137342421 ns/op 0 B/op 0 allocs/op BenchmarkXMul 2311 494267 ns/op 1 B/op 0 allocs/op BenchmarkXAdd 2396754 501 ns/op 0 B/op 0 allocs/op BenchmarkXDbl 2072690 571 ns/op 0 B/op 0 allocs/op BenchmarkIsom 78004 15171 ns/op 0 B/op 0 allocs/op BenchmarkFp512Sub 224635152 5.33 ns/op 0 B/op 0 allocs/op BenchmarkFp512Mul 246633255 4.90 ns/op 0 B/op 0 allocs/op BenchmarkCSwap 233228547 5.10 ns/op 0 B/op 0 allocs/op BenchmarkAddRdc 87348240 12.6 ns/op 0 B/op 0 allocs/op BenchmarkSubRdc 95112787 11.7 ns/op 0 B/op 0 allocs/op BenchmarkModExpRdc 25436 46878 ns/op 0 B/op 0 allocs/op BenchmarkMulBmiAsm 19527573 60.1 ns/op 0 B/op 0 allocs/op BenchmarkMulGeneric 7117650 164 ns/op 0 B/op 0 allocs/op * Go code has very similar performance when compared to C implementation. Results from sidh_torturer (4e2996e12d68364761064341cbe1d1b47efafe23) github.com:henrydcase/sidh-torture/csidh | TestName |Go | C | |------------------|----------|----------| |TestSharedSecret | 57.95774 | 57.91092 | |TestKeyGeneration | 62.23614 | 58.12980 | |TestSharedSecret | 55.28988 | 57.23132 | |TestKeyGeneration | 61.68745 | 58.66396 | |TestSharedSecret | 63.19408 | 58.64774 | |TestKeyGeneration | 62.34022 | 61.62539 | |TestSharedSecret | 62.85453 | 68.74503 | |TestKeyGeneration | 52.58518 | 58.40115 | |TestSharedSecret | 50.77081 | 61.91699 | |TestKeyGeneration | 59.91843 | 61.09266 | |TestSharedSecret | 59.97962 | 62.98151 | |TestKeyGeneration | 64.57525 | 56.22863 | |TestSharedSecret | 56.40521 | 55.77447 | |TestKeyGeneration | 67.85850 | 58.52604 | |TestSharedSecret | 60.54290 | 65.14052 | |TestKeyGeneration | 65.45766 | 58.42823 | On average Go implementation is 2% faster.
2019-11-24 03:39:35 +00:00
func ctPick64(which uint64, in1, in2 uint64) uint64 {
return (in1 & which) | (in2 & ^which)
}
// ctIsNonZero64 returns 0 in case i == 0, otherwise it returns 1.
// Constant-time.
func ctIsNonZero64(i uint64) int {
// In case i==0 then i-1 will set MSB. Only in such case (i OR ~(i-1))
// will result in MSB being not set (logical implication: (i-1)=>i is
// false iff (i-1)==0 and i==non-zero). In every other case MSB is
// set and hence function returns 1.
return int((i | (^(i - 1))) >> 63)
}
func mulGeneric(r, x, y *fp) {
var s fp // keeps intermediate results
var t1, t2 [9]uint64
var c, q uint64
for i := 0; i < numWords-1; i++ {
q = ((x[i] * y[0]) + s[0]) * pNegInv[0]
mul576(&t1, &p, q)
mul576(&t2, y, x[i])
// x[i]*y + q_i*p
t1[0], c = bits.Add64(t1[0], t2[0], 0)
t1[1], c = bits.Add64(t1[1], t2[1], c)
t1[2], c = bits.Add64(t1[2], t2[2], c)
t1[3], c = bits.Add64(t1[3], t2[3], c)
t1[4], c = bits.Add64(t1[4], t2[4], c)
t1[5], c = bits.Add64(t1[5], t2[5], c)
t1[6], c = bits.Add64(t1[6], t2[6], c)
t1[7], c = bits.Add64(t1[7], t2[7], c)
t1[8], _ = bits.Add64(t1[8], t2[8], c)
// s = (s + x[i]*y + q_i * p) / R
_, c = bits.Add64(t1[0], s[0], 0)
s[0], c = bits.Add64(t1[1], s[1], c)
s[1], c = bits.Add64(t1[2], s[2], c)
s[2], c = bits.Add64(t1[3], s[3], c)
s[3], c = bits.Add64(t1[4], s[4], c)
s[4], c = bits.Add64(t1[5], s[5], c)
s[5], c = bits.Add64(t1[6], s[6], c)
s[6], c = bits.Add64(t1[7], s[7], c)
s[7], _ = bits.Add64(t1[8], 0, c)
}
// last iteration stores result in r
q = ((x[numWords-1] * y[0]) + s[0]) * pNegInv[0]
mul576(&t1, &p, q)
mul576(&t2, y, x[numWords-1])
t1[0], c = bits.Add64(t1[0], t2[0], c)
t1[1], c = bits.Add64(t1[1], t2[1], c)
t1[2], c = bits.Add64(t1[2], t2[2], c)
t1[3], c = bits.Add64(t1[3], t2[3], c)
t1[4], c = bits.Add64(t1[4], t2[4], c)
t1[5], c = bits.Add64(t1[5], t2[5], c)
t1[6], c = bits.Add64(t1[6], t2[6], c)
t1[7], c = bits.Add64(t1[7], t2[7], c)
t1[8], _ = bits.Add64(t1[8], t2[8], c)
_, c = bits.Add64(t1[0], s[0], 0)
r[0], c = bits.Add64(t1[1], s[1], c)
r[1], c = bits.Add64(t1[2], s[2], c)
r[2], c = bits.Add64(t1[3], s[3], c)
r[3], c = bits.Add64(t1[4], s[4], c)
r[4], c = bits.Add64(t1[5], s[5], c)
r[5], c = bits.Add64(t1[6], s[6], c)
r[6], c = bits.Add64(t1[7], s[7], c)
r[7], _ = bits.Add64(t1[8], 0, c)
}
// Returns result of x<y operation.
func isLess(x, y *fp) bool {
for i := numWords - 1; i >= 0; i-- {
v, c := bits.Sub64(y[i], x[i], 0)
if c != 0 {
return false
}
if v != 0 {
return true
}
}
// x == y
return false
}
// r = x + y mod p.
func addRdc(r, x, y *fp) {
var c uint64
var t fp
r[0], c = bits.Add64(x[0], y[0], 0)
r[1], c = bits.Add64(x[1], y[1], c)
r[2], c = bits.Add64(x[2], y[2], c)
r[3], c = bits.Add64(x[3], y[3], c)
r[4], c = bits.Add64(x[4], y[4], c)
r[5], c = bits.Add64(x[5], y[5], c)
r[6], c = bits.Add64(x[6], y[6], c)
r[7], _ = bits.Add64(x[7], y[7], c)
t[0], c = bits.Sub64(r[0], p[0], 0)
t[1], c = bits.Sub64(r[1], p[1], c)
t[2], c = bits.Sub64(r[2], p[2], c)
t[3], c = bits.Sub64(r[3], p[3], c)
t[4], c = bits.Sub64(r[4], p[4], c)
t[5], c = bits.Sub64(r[5], p[5], c)
t[6], c = bits.Sub64(r[6], p[6], c)
t[7], c = bits.Sub64(r[7], p[7], c)
var w = 0 - c
r[0] = ctPick64(w, r[0], t[0])
r[1] = ctPick64(w, r[1], t[1])
r[2] = ctPick64(w, r[2], t[2])
r[3] = ctPick64(w, r[3], t[3])
r[4] = ctPick64(w, r[4], t[4])
r[5] = ctPick64(w, r[5], t[5])
r[6] = ctPick64(w, r[6], t[6])
r[7] = ctPick64(w, r[7], t[7])
}
2020-05-14 00:48:43 +01:00
// r = x - y.
cSIDH-511: (#26) Implementation of Commutative Supersingular Isogeny Diffie Hellman, based on "A faster way to CSIDH" paper (2018/782). * For fast isogeny calculation, implementation converts a curve from Montgomery to Edwards. All calculations are done on Edwards curve and then converted back to Montgomery. * As multiplication in a field Fp511 is most expensive operation the implementation contains multiple multiplications. It has most performant, assembly implementation which uses BMI2 and ADOX/ADCX instructions for modern CPUs. It also contains slower implementation which will run on older CPUs * Benchmarks (Intel SkyLake): BenchmarkGeneratePrivate 6459 172213 ns/op 0 B/op 0 allocs/op BenchmarkGenerateKeyPair 25 45800356 ns/op 0 B/op 0 allocs/op BenchmarkValidate 297 3915983 ns/op 0 B/op 0 allocs/op BenchmarkValidateRandom 184683 6231 ns/op 0 B/op 0 allocs/op BenchmarkValidateGenerated 25 48481306 ns/op 0 B/op 0 allocs/op BenchmarkDerive 19 60928763 ns/op 0 B/op 0 allocs/op BenchmarkDeriveGenerated 8 137342421 ns/op 0 B/op 0 allocs/op BenchmarkXMul 2311 494267 ns/op 1 B/op 0 allocs/op BenchmarkXAdd 2396754 501 ns/op 0 B/op 0 allocs/op BenchmarkXDbl 2072690 571 ns/op 0 B/op 0 allocs/op BenchmarkIsom 78004 15171 ns/op 0 B/op 0 allocs/op BenchmarkFp512Sub 224635152 5.33 ns/op 0 B/op 0 allocs/op BenchmarkFp512Mul 246633255 4.90 ns/op 0 B/op 0 allocs/op BenchmarkCSwap 233228547 5.10 ns/op 0 B/op 0 allocs/op BenchmarkAddRdc 87348240 12.6 ns/op 0 B/op 0 allocs/op BenchmarkSubRdc 95112787 11.7 ns/op 0 B/op 0 allocs/op BenchmarkModExpRdc 25436 46878 ns/op 0 B/op 0 allocs/op BenchmarkMulBmiAsm 19527573 60.1 ns/op 0 B/op 0 allocs/op BenchmarkMulGeneric 7117650 164 ns/op 0 B/op 0 allocs/op * Go code has very similar performance when compared to C implementation. Results from sidh_torturer (4e2996e12d68364761064341cbe1d1b47efafe23) github.com:henrydcase/sidh-torture/csidh | TestName |Go | C | |------------------|----------|----------| |TestSharedSecret | 57.95774 | 57.91092 | |TestKeyGeneration | 62.23614 | 58.12980 | |TestSharedSecret | 55.28988 | 57.23132 | |TestKeyGeneration | 61.68745 | 58.66396 | |TestSharedSecret | 63.19408 | 58.64774 | |TestKeyGeneration | 62.34022 | 61.62539 | |TestSharedSecret | 62.85453 | 68.74503 | |TestKeyGeneration | 52.58518 | 58.40115 | |TestSharedSecret | 50.77081 | 61.91699 | |TestKeyGeneration | 59.91843 | 61.09266 | |TestSharedSecret | 59.97962 | 62.98151 | |TestKeyGeneration | 64.57525 | 56.22863 | |TestSharedSecret | 56.40521 | 55.77447 | |TestKeyGeneration | 67.85850 | 58.52604 | |TestSharedSecret | 60.54290 | 65.14052 | |TestKeyGeneration | 65.45766 | 58.42823 | On average Go implementation is 2% faster.
2019-11-24 03:39:35 +00:00
func sub512(r, x, y *fp) uint64 {
var c uint64
r[0], c = bits.Sub64(x[0], y[0], 0)
r[1], c = bits.Sub64(x[1], y[1], c)
r[2], c = bits.Sub64(x[2], y[2], c)
r[3], c = bits.Sub64(x[3], y[3], c)
r[4], c = bits.Sub64(x[4], y[4], c)
r[5], c = bits.Sub64(x[5], y[5], c)
r[6], c = bits.Sub64(x[6], y[6], c)
r[7], c = bits.Sub64(x[7], y[7], c)
return c
}
// r = x - y mod p.
func subRdc(r, x, y *fp) {
var c uint64
// Same as sub512(r,x,y). Unfortunately
// compiler is not able to inline it.
r[0], c = bits.Sub64(x[0], y[0], 0)
r[1], c = bits.Sub64(x[1], y[1], c)
r[2], c = bits.Sub64(x[2], y[2], c)
r[3], c = bits.Sub64(x[3], y[3], c)
r[4], c = bits.Sub64(x[4], y[4], c)
r[5], c = bits.Sub64(x[5], y[5], c)
r[6], c = bits.Sub64(x[6], y[6], c)
r[7], c = bits.Sub64(x[7], y[7], c)
// if x<y => r=x-y+p
var w = 0 - c
r[0], c = bits.Add64(r[0], ctPick64(w, p[0], 0), 0)
r[1], c = bits.Add64(r[1], ctPick64(w, p[1], 0), c)
r[2], c = bits.Add64(r[2], ctPick64(w, p[2], 0), c)
r[3], c = bits.Add64(r[3], ctPick64(w, p[3], 0), c)
r[4], c = bits.Add64(r[4], ctPick64(w, p[4], 0), c)
r[5], c = bits.Add64(r[5], ctPick64(w, p[5], 0), c)
r[6], c = bits.Add64(r[6], ctPick64(w, p[6], 0), c)
r[7], _ = bits.Add64(r[7], ctPick64(w, p[7], 0), c)
}
// Fixed-window mod exp for fpBitLen bit value with 4 bit window. Returned
// result is a number in montgomery domain.
// r = b ^ e (mod p).
// Constant time.
func modExpRdcCommon(r, b, e *fp, fpBitLen int) {
var precomp [16]fp
var t fp
var c uint64
// Precompute step, computes an array of small powers of 'b'. As this
// algorithm implements 4-bit window, we need 2^4=16 of such values.
// b^0 = 1, which is equal to R from REDC.
precomp[0] = one // b ^ 0
precomp[1] = *b // b ^ 1
for i := 2; i < 16; i = i + 2 {
2020-05-14 00:48:43 +01:00
// OPTIMIZE: implement fast squering. Then interleaving fast squaring
cSIDH-511: (#26) Implementation of Commutative Supersingular Isogeny Diffie Hellman, based on "A faster way to CSIDH" paper (2018/782). * For fast isogeny calculation, implementation converts a curve from Montgomery to Edwards. All calculations are done on Edwards curve and then converted back to Montgomery. * As multiplication in a field Fp511 is most expensive operation the implementation contains multiple multiplications. It has most performant, assembly implementation which uses BMI2 and ADOX/ADCX instructions for modern CPUs. It also contains slower implementation which will run on older CPUs * Benchmarks (Intel SkyLake): BenchmarkGeneratePrivate 6459 172213 ns/op 0 B/op 0 allocs/op BenchmarkGenerateKeyPair 25 45800356 ns/op 0 B/op 0 allocs/op BenchmarkValidate 297 3915983 ns/op 0 B/op 0 allocs/op BenchmarkValidateRandom 184683 6231 ns/op 0 B/op 0 allocs/op BenchmarkValidateGenerated 25 48481306 ns/op 0 B/op 0 allocs/op BenchmarkDerive 19 60928763 ns/op 0 B/op 0 allocs/op BenchmarkDeriveGenerated 8 137342421 ns/op 0 B/op 0 allocs/op BenchmarkXMul 2311 494267 ns/op 1 B/op 0 allocs/op BenchmarkXAdd 2396754 501 ns/op 0 B/op 0 allocs/op BenchmarkXDbl 2072690 571 ns/op 0 B/op 0 allocs/op BenchmarkIsom 78004 15171 ns/op 0 B/op 0 allocs/op BenchmarkFp512Sub 224635152 5.33 ns/op 0 B/op 0 allocs/op BenchmarkFp512Mul 246633255 4.90 ns/op 0 B/op 0 allocs/op BenchmarkCSwap 233228547 5.10 ns/op 0 B/op 0 allocs/op BenchmarkAddRdc 87348240 12.6 ns/op 0 B/op 0 allocs/op BenchmarkSubRdc 95112787 11.7 ns/op 0 B/op 0 allocs/op BenchmarkModExpRdc 25436 46878 ns/op 0 B/op 0 allocs/op BenchmarkMulBmiAsm 19527573 60.1 ns/op 0 B/op 0 allocs/op BenchmarkMulGeneric 7117650 164 ns/op 0 B/op 0 allocs/op * Go code has very similar performance when compared to C implementation. Results from sidh_torturer (4e2996e12d68364761064341cbe1d1b47efafe23) github.com:henrydcase/sidh-torture/csidh | TestName |Go | C | |------------------|----------|----------| |TestSharedSecret | 57.95774 | 57.91092 | |TestKeyGeneration | 62.23614 | 58.12980 | |TestSharedSecret | 55.28988 | 57.23132 | |TestKeyGeneration | 61.68745 | 58.66396 | |TestSharedSecret | 63.19408 | 58.64774 | |TestKeyGeneration | 62.34022 | 61.62539 | |TestSharedSecret | 62.85453 | 68.74503 | |TestKeyGeneration | 52.58518 | 58.40115 | |TestSharedSecret | 50.77081 | 61.91699 | |TestKeyGeneration | 59.91843 | 61.09266 | |TestSharedSecret | 59.97962 | 62.98151 | |TestKeyGeneration | 64.57525 | 56.22863 | |TestSharedSecret | 56.40521 | 55.77447 | |TestKeyGeneration | 67.85850 | 58.52604 | |TestSharedSecret | 60.54290 | 65.14052 | |TestKeyGeneration | 65.45766 | 58.42823 | On average Go implementation is 2% faster.
2019-11-24 03:39:35 +00:00
// with multiplication should improve performance.
mulRdc(&precomp[i], &precomp[i/2], &precomp[i/2]) // sqr
mulRdc(&precomp[i+1], &precomp[i], b)
}
*r = one
for i := fpBitLen/4 - 1; i >= 0; i-- {
for j := 0; j < 4; j++ {
mulRdc(r, r, r)
}
// note: non resistant to cache SCA
idx := (e[i/16] >> uint((i%16)*4)) & 15
mulRdc(r, r, &precomp[idx])
}
// if p <= r < 2p then r = r-p
t[0], c = bits.Sub64(r[0], p[0], 0)
t[1], c = bits.Sub64(r[1], p[1], c)
t[2], c = bits.Sub64(r[2], p[2], c)
t[3], c = bits.Sub64(r[3], p[3], c)
t[4], c = bits.Sub64(r[4], p[4], c)
t[5], c = bits.Sub64(r[5], p[5], c)
t[6], c = bits.Sub64(r[6], p[6], c)
t[7], c = bits.Sub64(r[7], p[7], c)
var w = 0 - c
r[0] = ctPick64(w, r[0], t[0])
r[1] = ctPick64(w, r[1], t[1])
r[2] = ctPick64(w, r[2], t[2])
r[3] = ctPick64(w, r[3], t[3])
r[4] = ctPick64(w, r[4], t[4])
r[5] = ctPick64(w, r[5], t[5])
r[6] = ctPick64(w, r[6], t[6])
r[7] = ctPick64(w, r[7], t[7])
}
// modExpRdc does modular exponentation of 512-bit number.
// Constant-time.
func modExpRdc512(r, b, e *fp) {
modExpRdcCommon(r, b, e, 512)
}
// modExpRdc does modular exponentation of 64-bit number.
// Constant-time.
func modExpRdc64(r, b *fp, e uint64) {
modExpRdcCommon(r, b, &fp{e}, 64)
}
// isNonQuadRes checks whether value v is quadratic residue.
// Implementation uses Fermat's little theorem (or
// Euler's criterion)
// a^(p-1) == 1, hence
// (a^2) ((p-1)/2) == 1
// Which means v is a quadratic residue iff v^((p-1)/2) == 1.
// Caller provided v must be in montgomery domain.
// Returns 0 in case v is quadratic residue or 1 in case
// v is quadratic non-residue.
func (v *fp) isNonQuadRes() int {
var res fp
var b uint64
modExpRdc512(&res, v, &pMin1By2)
for i := range res {
b |= res[i] ^ one[i]
}
return ctIsNonZero64(b)
}
// isZero returns false in case v is equal to 0, otherwise
// true. Constant time.
func (v *fp) isZero() bool {
var r uint64
for i := 0; i < numWords; i++ {
r |= v[i]
}
return ctIsNonZero64(r) == 0
}
2020-05-14 00:48:43 +01:00
// equal checks if v is equal to in. Constant time.
cSIDH-511: (#26) Implementation of Commutative Supersingular Isogeny Diffie Hellman, based on "A faster way to CSIDH" paper (2018/782). * For fast isogeny calculation, implementation converts a curve from Montgomery to Edwards. All calculations are done on Edwards curve and then converted back to Montgomery. * As multiplication in a field Fp511 is most expensive operation the implementation contains multiple multiplications. It has most performant, assembly implementation which uses BMI2 and ADOX/ADCX instructions for modern CPUs. It also contains slower implementation which will run on older CPUs * Benchmarks (Intel SkyLake): BenchmarkGeneratePrivate 6459 172213 ns/op 0 B/op 0 allocs/op BenchmarkGenerateKeyPair 25 45800356 ns/op 0 B/op 0 allocs/op BenchmarkValidate 297 3915983 ns/op 0 B/op 0 allocs/op BenchmarkValidateRandom 184683 6231 ns/op 0 B/op 0 allocs/op BenchmarkValidateGenerated 25 48481306 ns/op 0 B/op 0 allocs/op BenchmarkDerive 19 60928763 ns/op 0 B/op 0 allocs/op BenchmarkDeriveGenerated 8 137342421 ns/op 0 B/op 0 allocs/op BenchmarkXMul 2311 494267 ns/op 1 B/op 0 allocs/op BenchmarkXAdd 2396754 501 ns/op 0 B/op 0 allocs/op BenchmarkXDbl 2072690 571 ns/op 0 B/op 0 allocs/op BenchmarkIsom 78004 15171 ns/op 0 B/op 0 allocs/op BenchmarkFp512Sub 224635152 5.33 ns/op 0 B/op 0 allocs/op BenchmarkFp512Mul 246633255 4.90 ns/op 0 B/op 0 allocs/op BenchmarkCSwap 233228547 5.10 ns/op 0 B/op 0 allocs/op BenchmarkAddRdc 87348240 12.6 ns/op 0 B/op 0 allocs/op BenchmarkSubRdc 95112787 11.7 ns/op 0 B/op 0 allocs/op BenchmarkModExpRdc 25436 46878 ns/op 0 B/op 0 allocs/op BenchmarkMulBmiAsm 19527573 60.1 ns/op 0 B/op 0 allocs/op BenchmarkMulGeneric 7117650 164 ns/op 0 B/op 0 allocs/op * Go code has very similar performance when compared to C implementation. Results from sidh_torturer (4e2996e12d68364761064341cbe1d1b47efafe23) github.com:henrydcase/sidh-torture/csidh | TestName |Go | C | |------------------|----------|----------| |TestSharedSecret | 57.95774 | 57.91092 | |TestKeyGeneration | 62.23614 | 58.12980 | |TestSharedSecret | 55.28988 | 57.23132 | |TestKeyGeneration | 61.68745 | 58.66396 | |TestSharedSecret | 63.19408 | 58.64774 | |TestKeyGeneration | 62.34022 | 61.62539 | |TestSharedSecret | 62.85453 | 68.74503 | |TestKeyGeneration | 52.58518 | 58.40115 | |TestSharedSecret | 50.77081 | 61.91699 | |TestKeyGeneration | 59.91843 | 61.09266 | |TestSharedSecret | 59.97962 | 62.98151 | |TestKeyGeneration | 64.57525 | 56.22863 | |TestSharedSecret | 56.40521 | 55.77447 | |TestKeyGeneration | 67.85850 | 58.52604 | |TestSharedSecret | 60.54290 | 65.14052 | |TestKeyGeneration | 65.45766 | 58.42823 | On average Go implementation is 2% faster.
2019-11-24 03:39:35 +00:00
func (v *fp) equal(in *fp) bool {
var r uint64
for i := range v {
r |= v[i] ^ in[i]
}
return ctIsNonZero64(r) == 0
}