math/big: optimize addVV and subVV on arm64 (c4f3fe95) · Commits · Kirill Smelkov / go

Commit c4f3fe95 authored Sep 25, 2017 by

erifan01 Committed by Cherry Zhang Mar 06, 2018

math/big: optimize addVV and subVV on arm64

The biggest hot spot of the existing implementation is "load" operations, which lead to poor performance.
By unrolling the cycle 4x and 2x, and using "LDP", "STP" instructions, this CL can reduce the "load" cost and improve performance.

Benchmarks:

name old time/op new time/op delta
AddVV/1-8 21.5ns ± 0% 11.5ns ± 0% -46.51% (p=0.008 n=5+5)
AddVV/2-8 13.5ns ± 0% 12.0ns ± 0% -11.11% (p=0.008 n=5+5)
AddVV/3-8 15.5ns ± 0% 13.0ns ± 0% -16.13% (p=0.008 n=5+5)
AddVV/4-8 17.5ns ± 0% 13.5ns ± 0% -22.86% (p=0.008 n=5+5)
AddVV/5-8 19.5ns ± 0% 14.5ns ± 0% -25.64% (p=0.008 n=5+5)
AddVV/10-8 29.5ns ± 0% 18.0ns ± 0% -38.98% (p=0.008 n=5+5)
AddVV/100-8 217ns ± 0% 94ns ± 0% -56.64% (p=0.008 n=5+5)
AddVV/1000-8 2.02µs ± 0% 1.03µs ± 0% -48.85% (p=0.008 n=5+5)
AddVV/10000-8 20.5µs ± 0% 11.3µs ± 0% -44.70% (p=0.008 n=5+5)
AddVV/100000-8 247µs ± 3% 154µs ± 0% -37.52% (p=0.008 n=5+5)
SubVV/1-8 21.5ns ± 0% 11.5ns ± 0% ~ (p=0.079 n=4+5)
SubVV/2-8 13.5ns ± 0% 12.0ns ± 0% -11.11% (p=0.008 n=5+5)
SubVV/3-8 15.5ns ± 0% 13.0ns ± 0% -16.13% (p=0.008 n=5+5)
SubVV/4-8 17.5ns ± 0% 13.5ns ± 0% -22.86% (p=0.008 n=5+5)
SubVV/5-8 19.5ns ± 0% 14.5ns ± 0% -25.64% (p=0.008 n=5+5)
SubVV/10-8 29.5ns ± 0% 18.0ns ± 0% -38.98% (p=0.008 n=5+5)
SubVV/100-8 217ns ± 0% 94ns ± 0% -56.64% (p=0.008 n=5+5)
SubVV/1000-8 2.02µs ± 0% 0.80µs ± 0% -60.50% (p=0.008 n=5+5)
SubVV/10000-8 20.5µs ± 0% 11.3µs ± 0% -44.99% (p=0.008 n=5+5)
SubVV/100000-8 221µs ±11% 223µs ±16% ~ (p=0.690 n=5+5)
AddVW/1-8 9.32ns ± 0% 9.32ns ± 0% ~ (all equal)
AddVW/2-8 19.7ns ± 1% 19.7ns ± 0% ~ (p=0.381 n=5+4)
AddVW/3-8 11.5ns ± 0% 11.5ns ± 0% ~ (all equal)
AddVW/4-8 13.0ns ± 0% 13.0ns ± 0% ~ (all equal)
AddVW/5-8 14.5ns ± 0% 14.5ns ± 0% ~ (all equal)
AddVW/10-8 22.0ns ± 0% 22.0ns ± 0% ~ (all equal)
AddVW/100-8 167ns ± 0% 167ns ± 0% ~ (all equal)
AddVW/1000-8 1.52µs ± 0% 1.52µs ± 0% +0.40% (p=0.008 n=5+5)
AddVW/10000-8 15.1µs ± 0% 15.1µs ± 0% ~ (p=0.556 n=5+4)
AddVW/100000-8 152µs ± 1% 152µs ± 1% ~ (p=0.690 n=5+5)
AddMulVVW/1-8 33.3ns ± 0% 32.7ns ± 1% -1.86% (p=0.008 n=5+5)
AddMulVVW/2-8 59.3ns ± 1% 56.9ns ± 1% -4.15% (p=0.008 n=5+5)
AddMulVVW/3-8 80.5ns ± 1% 85.4ns ± 3% +6.19% (p=0.008 n=5+5)
AddMulVVW/4-8 127ns ± 0% 111ns ± 1% -13.19% (p=0.008 n=5+5)
AddMulVVW/5-8 144ns ± 0% 149ns ± 0% +3.47% (p=0.016 n=4+5)
AddMulVVW/10-8 298ns ± 1% 283ns ± 0% -4.77% (p=0.008 n=5+5)
AddMulVVW/100-8 3.06µs ± 0% 2.99µs ± 0% -2.21% (p=0.008 n=5+5)
AddMulVVW/1000-8 31.3µs ± 0% 26.9µs ± 0% -14.17% (p=0.008 n=5+5)
AddMulVVW/10000-8 316µs ± 0% 305µs ± 0% -3.51% (p=0.008 n=5+5)
AddMulVVW/100000-8 3.17ms ± 0% 3.17ms ± 1% ~ (p=0.690 n=5+5)
DecimalConversion-8 316µs ± 1% 313µs ± 2% ~ (p=0.095 n=5+5)
FloatString/100-8 2.53µs ± 1% 2.56µs ± 2% ~ (p=0.222 n=5+5)
FloatString/1000-8 58.4µs ± 0% 58.5µs ± 0% ~ (p=0.206 n=5+5)
FloatString/10000-8 4.59ms ± 0% 4.58ms ± 0% -0.31% (p=0.008 n=5+5)
FloatString/100000-8 446ms ± 0% 444ms ± 0% -0.31% (p=0.008 n=5+5)
FloatAdd/10-8 184ns ± 0% 172ns ± 0% -6.30% (p=0.008 n=5+5)
FloatAdd/100-8 189ns ± 2% 191ns ± 4% ~ (p=0.381 n=5+5)
FloatAdd/1000-8 371ns ± 0% 347ns ± 1% -6.42% (p=0.008 n=5+5)
FloatAdd/10000-8 1.87µs ± 0% 1.68µs ± 0% -10.16% (p=0.008 n=5+5)
FloatAdd/100000-8 17.1µs ± 0% 15.6µs ± 0% -8.74% (p=0.016 n=5+4)
FloatSub/10-8 152ns ± 0% 138ns ± 0% -9.47% (p=0.000 n=4+5)
FloatSub/100-8 148ns ± 0% 142ns ± 0% -4.05% (p=0.000 n=5+4)
FloatSub/1000-8 245ns ± 1% 217ns ± 0% -11.28% (p=0.000 n=5+4)
FloatSub/10000-8 1.07µs ± 0% 0.88µs ± 1% -18.14% (p=0.008 n=5+5)
FloatSub/100000-8 9.58µs ± 0% 7.96µs ± 0% -16.84% (p=0.008 n=5+5)
ParseFloatSmallExp-8 28.8µs ± 1% 29.0µs ± 1% ~ (p=0.095 n=5+5)
ParseFloatLargeExp-8 126µs ± 1% 126µs ± 1% ~ (p=0.841 n=5+5)
GCD10x10/WithoutXY-8 277ns ± 2% 281ns ± 4% ~ (p=0.746 n=5+5)
GCD10x10/WithXY-8 2.10µs ± 1% 2.12µs ± 3% ~ (p=0.548 n=5+5)
GCD10x100/WithoutXY-8 615ns ± 3% 607ns ± 2% ~ (p=0.135 n=5+5)
GCD10x100/WithXY-8 3.50µs ± 2% 3.62µs ± 5% ~ (p=0.151 n=5+5)
GCD10x1000/WithoutXY-8 1.39µs ± 2% 1.39µs ± 3% ~ (p=0.690 n=5+5)
GCD10x1000/WithXY-8 7.39µs ± 1% 7.34µs ± 2% ~ (p=0.135 n=5+5)
GCD10x10000/WithoutXY-8 8.66µs ± 1% 8.68µs ± 1% ~ (p=0.421 n=5+5)
GCD10x10000/WithXY-8 28.1µs ± 2% 27.0µs ± 2% -3.81% (p=0.008 n=5+5)
GCD10x100000/WithoutXY-8 79.3µs ± 1% 79.3µs ± 1% ~ (p=0.841 n=5+5)
GCD10x100000/WithXY-8 238µs ± 0% 227µs ± 1% -4.74% (p=0.008 n=5+5)
GCD100x100/WithoutXY-8 1.89µs ± 1% 1.88µs ± 2% ~ (p=0.968 n=5+5)
GCD100x100/WithXY-8 26.7µs ± 1% 27.0µs ± 1% +1.44% (p=0.032 n=5+5)
GCD100x1000/WithoutXY-8 4.48µs ± 1% 4.45µs ± 2% ~ (p=0.341 n=5+5)
GCD100x1000/WithXY-8 36.3µs ± 1% 35.1µs ± 1% -3.27% (p=0.008 n=5+5)
GCD100x10000/WithoutXY-8 22.8µs ± 0% 22.7µs ± 1% ~ (p=0.056 n=5+5)
GCD100x10000/WithXY-8 145µs ± 1% 133µs ± 1% -8.33% (p=0.008 n=5+5)
GCD100x100000/WithoutXY-8 198µs ± 0% 195µs ± 0% -1.56% (p=0.008 n=5+5)
GCD100x100000/WithXY-8 1.11ms ± 0% 1.00ms ± 0% -10.04% (p=0.008 n=5+5)
GCD1000x1000/WithoutXY-8 25.2µs ± 1% 24.8µs ± 1% -1.63% (p=0.016 n=5+5)
GCD1000x1000/WithXY-8 513µs ± 0% 517µs ± 2% ~ (p=0.421 n=5+5)
GCD1000x10000/WithoutXY-8 57.0µs ± 0% 52.7µs ± 1% -7.56% (p=0.008 n=5+5)
GCD1000x10000/WithXY-8 1.20ms ± 0% 1.10ms ± 0% -8.70% (p=0.008 n=5+5)
GCD1000x100000/WithoutXY-8 358µs ± 0% 318µs ± 1% -11.03% (p=0.008 n=5+5)
GCD1000x100000/WithXY-8 8.71ms ± 0% 7.65ms ± 0% -12.19% (p=0.008 n=5+5)
GCD10000x10000/WithoutXY-8 690µs ± 0% 630µs ± 0% -8.71% (p=0.008 n=5+5)
GCD10000x10000/WithXY-8 16.0ms ± 1% 14.9ms ± 0% -6.85% (p=0.008 n=5+5)
GCD10000x100000/WithoutXY-8 2.09ms ± 0% 1.75ms ± 0% -16.09% (p=0.016 n=5+4)
GCD10000x100000/WithXY-8 86.8ms ± 0% 76.3ms ± 0% -12.09% (p=0.008 n=5+5)
GCD100000x100000/WithoutXY-8 51.1ms ± 0% 46.0ms ± 0% -9.97% (p=0.008 n=5+5)
GCD100000x100000/WithXY-8 1.25s ± 0% 1.15s ± 0% -7.92% (p=0.008 n=5+5)
Hilbert-8 2.45ms ± 1% 2.49ms ± 1% +1.99% (p=0.008 n=5+5)
Binomial-8 4.98µs ± 3% 4.90µs ± 2% ~ (p=0.421 n=5+5)
QuoRem-8 7.10µs ± 0% 6.21µs ± 0% -12.55% (p=0.016 n=5+4)
Exp-8 161ms ± 0% 161ms ± 0% ~ (p=0.421 n=5+5)
Exp2-8 161ms ± 0% 161ms ± 0% ~ (p=0.151 n=5+5)
Bitset-8 40.4ns ± 0% 40.3ns ± 0% ~ (p=0.190 n=5+5)
BitsetNeg-8 163ns ± 3% 137ns ± 2% -15.91% (p=0.008 n=5+5)
BitsetOrig-8 377ns ± 1% 372ns ± 1% -1.22% (p=0.024 n=5+5)
BitsetNegOrig-8 631ns ± 1% 605ns ± 1% -4.09% (p=0.008 n=5+5)
ModSqrt225_Tonelli-8 7.26ms ± 0% 7.26ms ± 0% ~ (p=0.548 n=5+5)
ModSqrt224_3Mod4-8 2.24ms ± 0% 2.24ms ± 0% ~ (p=1.000 n=5+5)
ModSqrt5430_Tonelli-8 62.4s ± 0% 62.4s ± 0% ~ (p=0.841 n=5+5)
ModSqrt5430_3Mod4-8 20.8s ± 0% 20.7s ± 0% ~ (p=0.056 n=5+5)
Sqrt-8 101µs ± 0% 89µs ± 0% -12.17% (p=0.008 n=5+5)
IntSqr/1-8 32.5ns ± 1% 32.7ns ± 1% ~ (p=0.056 n=5+5)
IntSqr/2-8 160ns ± 5% 158ns ± 0% ~ (p=0.397 n=5+4)
IntSqr/3-8 298ns ± 4% 296ns ± 4% ~ (p=0.667 n=5+5)
IntSqr/5-8 737ns ± 5% 761ns ± 3% +3.34% (p=0.016 n=5+5)
IntSqr/8-8 1.87µs ± 4% 1.90µs ± 3% ~ (p=0.222 n=5+5)
IntSqr/10-8 2.96µs ± 4% 2.92µs ± 6% ~ (p=0.310 n=5+5)
IntSqr/20-8 6.28µs ± 3% 6.21µs ± 2% ~ (p=0.310 n=5+5)
IntSqr/30-8 14.0µs ± 2% 13.9µs ± 2% ~ (p=0.548 n=5+5)
IntSqr/50-8 37.7µs ± 3% 38.3µs ± 2% ~ (p=0.095 n=5+5)
IntSqr/80-8 95.9µs ± 2% 95.1µs ± 1% ~ (p=0.310 n=5+5)
IntSqr/100-8 148µs ± 1% 148µs ± 1% ~ (p=0.841 n=5+5)
IntSqr/200-8 586µs ± 1% 587µs ± 1% ~ (p=1.000 n=5+5)
IntSqr/300-8 1.32ms ± 0% 1.31ms ± 1% -0.73% (p=0.032 n=5+5)
IntSqr/500-8 2.48ms ± 0% 2.45ms ± 0% -1.15% (p=0.008 n=5+5)
IntSqr/800-8 4.68ms ± 0% 4.62ms ± 0% -1.23% (p=0.008 n=5+5)
IntSqr/1000-8 7.57ms ± 0% 7.50ms ± 0% -0.84% (p=0.008 n=5+5)
Mul-8 311ms ± 0% 308ms ± 0% -0.81% (p=0.008 n=5+5)
Exp3Power/0x10-8 574ns ± 1% 578ns ± 2% ~ (p=0.500 n=5+5)
Exp3Power/0x40-8 640ns ± 1% 646ns ± 0% ~ (p=0.056 n=5+5)
Exp3Power/0x100-8 1.42µs ± 1% 1.42µs ± 1% ~ (p=0.246 n=5+5)
Exp3Power/0x400-8 8.30µs ± 1% 8.29µs ± 1% ~ (p=0.802 n=5+5)
Exp3Power/0x1000-8 60.0µs ± 0% 59.9µs ± 0% -0.24% (p=0.016 n=5+5)
Exp3Power/0x4000-8 817µs ± 0% 816µs ± 0% -0.17% (p=0.008 n=5+5)
Exp3Power/0x10000-8 7.80ms ± 1% 7.70ms ± 0% -1.23% (p=0.008 n=5+5)
Exp3Power/0x40000-8 73.4ms ± 0% 72.5ms ± 0% -1.28% (p=0.008 n=5+5)
Exp3Power/0x100000-8 665ms ± 0% 656ms ± 0% -1.34% (p=0.008 n=5+5)
Exp3Power/0x400000-8 5.99s ± 0% 5.90s ± 0% -1.40% (p=0.008 n=5+5)
Fibo-8 116ms ± 0% 50ms ± 0% -57.09% (p=0.008 n=5+5)
NatSqr/1-8 112ns ± 4% 112ns ± 2% ~ (p=0.968 n=5+5)
NatSqr/2-8 251ns ± 2% 250ns ± 1% ~ (p=0.571 n=5+5)
NatSqr/3-8 378ns ± 2% 379ns ± 2% ~ (p=0.794 n=5+5)
NatSqr/5-8 829ns ± 3% 827ns ± 2% ~ (p=1.000 n=5+5)
NatSqr/8-8 1.97µs ± 2% 1.95µs ± 2% ~ (p=0.310 n=5+5)
NatSqr/10-8 3.02µs ± 2% 2.99µs ± 2% ~ (p=0.421 n=5+5)
NatSqr/20-8 6.51µs ± 2% 6.49µs ± 1% ~ (p=0.841 n=5+5)
NatSqr/30-8 14.1µs ± 2% 14.0µs ± 2% ~ (p=0.841 n=5+5)
NatSqr/50-8 38.1µs ± 2% 38.3µs ± 3% ~ (p=0.690 n=5+5)
NatSqr/80-8 95.5µs ± 2% 96.0µs ± 1% ~ (p=0.421 n=5+5)
NatSqr/100-8 150µs ± 1% 148µs ± 2% ~ (p=0.095 n=5+5)
NatSqr/200-8 588µs ± 1% 590µs ± 1% ~ (p=0.421 n=5+5)
NatSqr/300-8 1.32ms ± 1% 1.31ms ± 1% ~ (p=0.841 n=5+5)
NatSqr/500-8 2.50ms ± 0% 2.47ms ± 0% -1.03% (p=0.008 n=5+5)
NatSqr/800-8 4.70ms ± 0% 4.64ms ± 0% -1.31% (p=0.008 n=5+5)
NatSqr/1000-8 7.60ms ± 0% 7.52ms ± 0% -1.01% (p=0.008 n=5+5)
ScanPi-8 326µs ± 0% 326µs ± 0% ~ (p=0.841 n=5+5)
StringPiParallel-8 70.3µs ± 5% 63.8µs ±10% ~ (p=0.056 n=5+5)
Scan/10/Base2-8 1.09µs ± 0% 1.09µs ± 0% ~ (p=0.317 n=5+5)
Scan/100/Base2-8 7.79µs ± 0% 7.78µs ± 0% ~ (p=0.063 n=5+5)
Scan/1000/Base2-8 79.0µs ± 0% 78.9µs ± 0% -0.18% (p=0.008 n=5+5)
Scan/10000/Base2-8 1.22ms ± 0% 1.22ms ± 0% -0.15% (p=0.008 n=5+5)
Scan/100000/Base2-8 55.1ms ± 0% 55.2ms ± 0% +0.20% (p=0.008 n=5+5)
Scan/10/Base8-8 512ns ± 0% 512ns ± 1% ~ (p=0.810 n=5+5)
Scan/100/Base8-8 2.89µs ± 0% 2.89µs ± 0% ~ (p=0.810 n=5+5)
Scan/1000/Base8-8 31.0µs ± 0% 31.0µs ± 0% ~ (p=0.151 n=5+5)
Scan/10000/Base8-8 740µs ± 0% 741µs ± 0% +0.10% (p=0.008 n=5+5)
Scan/100000/Base8-8 50.6ms ± 0% 50.6ms ± 0% +0.08% (p=0.008 n=5+5)
Scan/10/Base10-8 487ns ± 0% 487ns ± 0% ~ (p=0.571 n=5+5)
Scan/100/Base10-8 2.67µs ± 0% 2.67µs ± 0% ~ (p=0.810 n=5+5)
Scan/1000/Base10-8 28.7µs ± 0% 28.7µs ± 0% +0.06% (p=0.008 n=5+5)
Scan/10000/Base10-8 716µs ± 0% 717µs ± 0% ~ (p=0.222 n=5+5)
Scan/100000/Base10-8 50.3ms ± 0% 50.3ms ± 0% +0.10% (p=0.008 n=5+5)
Scan/10/Base16-8 438ns ± 0% 437ns ± 1% ~ (p=0.786 n=5+5)
Scan/100/Base16-8 2.47µs ± 0% 2.47µs ± 0% -0.19% (p=0.048 n=5+5)
Scan/1000/Base16-8 27.2µs ± 0% 27.3µs ± 0% ~ (p=0.087 n=5+5)
Scan/10000/Base16-8 722µs ± 0% 722µs ± 0% +0.11% (p=0.008 n=5+5)
Scan/100000/Base16-8 52.6ms ± 0% 52.7ms ± 0% +0.15% (p=0.008 n=5+5)
String/10/Base2-8 247ns ± 2% 248ns ± 1% ~ (p=0.437 n=5+5)
String/100/Base2-8 1.51µs ± 0% 1.51µs ± 0% -0.37% (p=0.024 n=5+5)
String/1000/Base2-8 13.6µs ± 1% 13.5µs ± 0% ~ (p=0.095 n=5+5)
String/10000/Base2-8 135µs ± 0% 135µs ± 1% ~ (p=0.841 n=5+5)
String/100000/Base2-8 1.32ms ± 1% 1.32ms ± 1% ~ (p=0.690 n=5+5)
String/10/Base8-8 169ns ± 1% 169ns ± 1% ~ (p=1.000 n=5+5)
String/100/Base8-8 636ns ± 0% 634ns ± 1% ~ (p=0.413 n=5+5)
String/1000/Base8-8 5.33µs ± 1% 5.32µs ± 0% ~ (p=0.222 n=5+5)
String/10000/Base8-8 50.9µs ± 1% 50.7µs ± 0% ~ (p=0.151 n=5+5)
String/100000/Base8-8 500µs ± 1% 497µs ± 0% ~ (p=0.421 n=5+5)
String/10/Base10-8 516ns ± 1% 513ns ± 0% -0.62% (p=0.016 n=5+4)
String/100/Base10-8 1.97µs ± 0% 1.96µs ± 0% ~ (p=0.667 n=4+5)
String/1000/Base10-8 12.5µs ± 0% 11.5µs ± 0% -7.92% (p=0.008 n=5+5)
String/10000/Base10-8 57.7µs ± 0% 52.5µs ± 0% -8.93% (p=0.008 n=5+5)
String/100000/Base10-8 25.6ms ± 0% 21.6ms ± 0% -15.94% (p=0.008 n=5+5)
String/10/Base16-8 150ns ± 1% 149ns ± 0% ~ (p=0.413 n=5+4)
String/100/Base16-8 514ns ± 1% 514ns ± 1% ~ (p=0.849 n=5+5)
String/1000/Base16-8 4.01µs ± 0% 4.01µs ± 0% ~ (p=0.421 n=5+5)
String/10000/Base16-8 37.8µs ± 1% 37.8µs ± 1% ~ (p=0.841 n=5+5)
String/100000/Base16-8 373µs ± 2% 373µs ± 0% ~ (p=0.421 n=5+5)
LeafSize/0-8 6.63ms ± 0% 6.63ms ± 0% ~ (p=0.730 n=4+5)
LeafSize/1-8 74.0µs ± 0% 67.7µs ± 1% -8.53% (p=0.008 n=5+5)
LeafSize/2-8 74.2µs ± 0% 68.3µs ± 1% -7.99% (p=0.008 n=5+5)
LeafSize/3-8 379µs ± 0% 309µs ± 0% -18.52% (p=0.008 n=5+5)
LeafSize/4-8 72.7µs ± 1% 66.7µs ± 0% -8.37% (p=0.008 n=5+5)
LeafSize/5-8 471µs ± 0% 384µs ± 0% -18.55% (p=0.008 n=5+5)
LeafSize/6-8 378µs ± 0% 308µs ± 0% -18.59% (p=0.008 n=5+5)
LeafSize/7-8 245µs ± 0% 204µs ± 1% -16.75% (p=0.008 n=5+5)
LeafSize/8-8 73.4µs ± 0% 66.9µs ± 1% -8.79% (p=0.008 n=5+5)
LeafSize/9-8 538µs ± 0% 437µs ± 0% -18.75% (p=0.008 n=5+5)
LeafSize/10-8 472µs ± 0% 396µs ± 1% -16.01% (p=0.008 n=5+5)
LeafSize/11-8 460µs ± 0% 374µs ± 0% -18.58% (p=0.008 n=5+5)
LeafSize/12-8 378µs ± 0% 308µs ± 0% -18.38% (p=0.008 n=5+5)
LeafSize/13-8 343µs ± 0% 284µs ± 0% -17.30% (p=0.008 n=5+5)
LeafSize/14-8 248µs ± 0% 206µs ± 0% -16.94% (p=0.008 n=5+5)
LeafSize/15-8 169µs ± 0% 144µs ± 0% -14.69% (p=0.008 n=5+5)
LeafSize/16-8 72.9µs ± 0% 66.8µs ± 1% -8.27% (p=0.008 n=5+5)
LeafSize/32-8 82.5µs ± 0% 76.7µs ± 0% -7.04% (p=0.008 n=5+5)
LeafSize/64-8 134µs ± 0% 129µs ± 0% -3.80% (p=0.008 n=5+5)
ProbablyPrime/n=0-8 44.2ms ± 0% 43.4ms ± 0% -1.95% (p=0.008 n=5+5)
ProbablyPrime/n=1-8 64.9ms ± 0% 64.0ms ± 0% -1.27% (p=0.008 n=5+5)
ProbablyPrime/n=5-8 147ms ± 0% 146ms ± 0% -0.58% (p=0.008 n=5+5)
ProbablyPrime/n=10-8 250ms ± 0% 249ms ± 0% -0.35% (p=0.008 n=5+5)
ProbablyPrime/n=20-8 456ms ± 0% 455ms ± 0% -0.18% (p=0.008 n=5+5)
ProbablyPrime/Lucas-8 23.6ms ± 0% 22.7ms ± 0% -3.74% (p=0.008 n=5+5)
ProbablyPrime/MillerRabinBase2-8 20.7ms ± 0% 20.6ms ± 0% ~ (p=0.421 n=5+5)
FloatSqrt/64-8 2.25µs ± 1% 2.29µs ± 0% +1.48% (p=0.008 n=5+5)
FloatSqrt/128-8 4.86µs ± 1% 4.92µs ± 1% +1.21% (p=0.032 n=5+5)
FloatSqrt/256-8 13.6µs ± 0% 13.7µs ± 1% +1.31% (p=0.032 n=5+5)
FloatSqrt/1000-8 70.0µs ± 1% 70.1µs ± 0% ~ (p=0.690 n=5+5)
FloatSqrt/10000-8 1.92ms ± 0% 1.90ms ± 0% -0.59% (p=0.008 n=5+5)
FloatSqrt/100000-8 55.3ms ± 0% 54.8ms ± 0% -1.01% (p=0.008 n=5+5)
FloatSqrt/1000000-8 4.56s ± 0% 4.50s ± 0% -1.28% (p=0.008 n=5+5)

name old speed new speed delta
AddVV/1-8 2.97GB/s ± 0% 5.56GB/s ± 0% +86.85% (p=0.008 n=5+5)
AddVV/2-8 9.47GB/s ± 0% 10.66GB/s ± 0% +12.50% (p=0.008 n=5+5)
AddVV/3-8 12.4GB/s ± 0% 14.7GB/s ± 0% +19.10% (p=0.008 n=5+5)
AddVV/4-8 14.6GB/s ± 0% 18.9GB/s ± 0% +29.63% (p=0.016 n=4+5)
AddVV/5-8 16.4GB/s ± 0% 22.0GB/s ± 0% +34.47% (p=0.016 n=5+4)
AddVV/10-8 21.7GB/s ± 0% 35.5GB/s ± 0% +63.89% (p=0.008 n=5+5)
AddVV/100-8 29.4GB/s ± 0% 68.0GB/s ± 0% +131.38% (p=0.008 n=5+5)
AddVV/1000-8 31.7GB/s ± 0% 61.9GB/s ± 0% +95.43% (p=0.008 n=5+5)
AddVV/10000-8 31.2GB/s ± 0% 56.4GB/s ± 0% +80.83% (p=0.008 n=5+5)
AddVV/100000-8 25.9GB/s ± 3% 41.4GB/s ± 0% +59.98% (p=0.008 n=5+5)
SubVV/1-8 2.97GB/s ± 0% 5.56GB/s ± 0% +86.97% (p=0.016 n=4+5)
SubVV/2-8 9.47GB/s ± 0% 10.66GB/s ± 0% +12.51% (p=0.008 n=5+5)
SubVV/3-8 12.4GB/s ± 0% 14.8GB/s ± 0% +19.23% (p=0.016 n=4+5)
SubVV/4-8 14.6GB/s ± 0% 18.9GB/s ± 0% +29.56% (p=0.008 n=5+5)
SubVV/5-8 16.4GB/s ± 0% 22.0GB/s ± 0% +34.47% (p=0.016 n=4+5)
SubVV/10-8 21.7GB/s ± 0% 35.5GB/s ± 0% +63.89% (p=0.008 n=5+5)
SubVV/100-8 29.4GB/s ± 0% 68.0GB/s ± 0% +131.38% (p=0.008 n=5+5)
SubVV/1000-8 31.6GB/s ± 0% 80.1GB/s ± 0% +153.08% (p=0.008 n=5+5)
SubVV/10000-8 31.2GB/s ± 0% 56.7GB/s ± 0% +81.79% (p=0.008 n=5+5)
SubVV/100000-8 29.1GB/s ±10% 29.0GB/s ±18% ~ (p=0.690 n=5+5)
AddVW/1-8 859MB/s ± 0% 859MB/s ± 0% -0.01% (p=0.008 n=5+5)
AddVW/2-8 811MB/s ± 1% 814MB/s ± 0% ~ (p=0.413 n=5+4)
AddVW/3-8 2.08GB/s ± 0% 2.08GB/s ± 0% ~ (p=0.206 n=5+5)
AddVW/4-8 2.46GB/s ± 0% 2.46GB/s ± 0% ~ (p=0.056 n=5+5)
AddVW/5-8 2.75GB/s ± 0% 2.75GB/s ± 0% ~ (p=0.508 n=5+5)
AddVW/10-8 3.63GB/s ± 0% 3.63GB/s ± 0% ~ (p=0.214 n=5+5)
AddVW/100-8 4.79GB/s ± 0% 4.79GB/s ± 0% ~ (p=0.500 n=5+5)
AddVW/1000-8 5.27GB/s ± 0% 5.25GB/s ± 0% -0.43% (p=0.008 n=5+5)
AddVW/10000-8 5.30GB/s ± 0% 5.30GB/s ± 0% ~ (p=0.397 n=5+5)
AddVW/100000-8 5.27GB/s ± 1% 5.25GB/s ± 1% ~ (p=0.690 n=5+5)
AddMulVVW/1-8 1.92GB/s ± 0% 1.96GB/s ± 1% +1.95% (p=0.008 n=5+5)
AddMulVVW/2-8 2.16GB/s ± 1% 2.25GB/s ± 1% +4.32% (p=0.008 n=5+5)
AddMulVVW/3-8 2.39GB/s ± 1% 2.25GB/s ± 3% -5.79% (p=0.008 n=5+5)
AddMulVVW/4-8 2.00GB/s ± 0% 2.31GB/s ± 1% +15.31% (p=0.008 n=5+5)
AddMulVVW/5-8 2.22GB/s ± 0% 2.14GB/s ± 0% -3.86% (p=0.008 n=5+5)
AddMulVVW/10-8 2.15GB/s ± 1% 2.25GB/s ± 0% +5.03% (p=0.008 n=5+5)
AddMulVVW/100-8 2.09GB/s ± 0% 2.14GB/s ± 0% +2.25% (p=0.008 n=5+5)
AddMulVVW/1000-8 2.04GB/s ± 0% 2.38GB/s ± 0% +16.52% (p=0.008 n=5+5)
AddMulVVW/10000-8 2.03GB/s ± 0% 2.10GB/s ± 0% +3.64% (p=0.008 n=5+5)
AddMulVVW/100000-8 2.02GB/s ± 0% 2.02GB/s ± 1% ~ (p=0.690 n=5+5)

Change-Id: Ie482d67a7dbb5af6f5d81af2b3d9d14bd66336db
Reviewed-on: https://go-review.googlesource.com/77831Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

parent adcf2d59

Hide whitespace changes

Inline Side-by-side

View file @ c4f3fe95

@@ -30,42 +30,82 @@ TEXT ·divWW(SB),NOSPLIT,$0
 // func addVV(z, x, y []Word) (c Word)
 TEXT ·addVV(SB),NOSPLIT,$0
 	MOVD	z+0(FP), R3
 	MOVD	z_len+8(FP), R0
 	MOVD	x+24(FP), R1
 	MOVD	y+48(FP), R2
 	ADDS	$0, R0 // clear carry flag
 loop:
 	CBZ	R0, done // careful not to touch the carry flag
 	MOVD.P	8(R1), R4
 	MOVD.P	8(R2), R5
 	ADCS	R4, R5
 	MOVD.P	R5, 8(R3)
 	MOVD	x+24(FP), R8
 	MOVD	y+48(FP), R9
 	MOVD	z+0(FP), R10
 	ADDS	$0, R0		// clear carry flag
 	TBZ	$0, R0, two
 	MOVD.P	8(R8), R11
 	MOVD.P	8(R9), R15
 	ADCS	R15, R11
 	MOVD.P	R11, 8(R10)
 	SUB	$1, R0
 two:
 	TBZ	$1, R0, loop
 	LDP.P	16(R8), (R11, R12)
 	LDP.P	16(R9), (R15, R16)
 	ADCS	R15, R11
 	ADCS	R16, R12
 	STP.P	(R11, R12), 16(R10)
 	SUB	$2, R0
 loop:
 	CBZ	R0, done	// careful not to touch the carry flag
 	LDP.P	32(R8), (R11, R12)
 	LDP	-16(R8), (R13, R14)
 	LDP.P	32(R9), (R15, R16)
 	LDP	-16(R9), (R17, R18)
 	ADCS	R15, R11
 	ADCS	R16, R12
 	ADCS	R17, R13
 	ADCS	R18, R14
 	STP.P	(R11, R12), 32(R10)
 	STP	(R13, R14), -16(R10)
 	SUB	$4, R0
 	B	loop
 done:
 	CSET	HS, R0 // extract carry flag
 	CSET	HS, R0		// extract carry flag
 	MOVD	R0, c+72(FP)
 	RET
 // func subVV(z, x, y []Word) (c Word)
 TEXT ·subVV(SB),NOSPLIT,$0
 	MOVD	z+0(FP), R3
 	MOVD	z_len+8(FP), R0
 	MOVD	x+24(FP), R1
 	MOVD	y+48(FP), R2
 	CMP	R0, R0 // set carry flag
 loop:
 	CBZ	R0, done // careful not to touch the carry flag
 	MOVD.P	8(R1), R4
 	MOVD.P	8(R2), R5
 	SBCS	R5, R4
 	MOVD.P	R4, 8(R3)
 	MOVD	x+24(FP), R8
 	MOVD	y+48(FP), R9
 	MOVD	z+0(FP), R10
 	CMP	R0, R0		// set carry flag
 	TBZ	$0, R0, two
 	MOVD.P	8(R8), R11
 	MOVD.P	8(R9), R15
 	SBCS	R15, R11
 	MOVD.P	R11, 8(R10)
 	SUB	$1, R0
 two:
 	TBZ	$1, R0, loop
 	LDP.P	16(R8), (R11, R12)
 	LDP.P	16(R9), (R15, R16)
 	SBCS	R15, R11
 	SBCS	R16, R12
 	STP.P	(R11, R12), 16(R10)
 	SUB	$2, R0
 loop:
 	CBZ	R0, done	// careful not to touch the carry flag
 	LDP.P	32(R8), (R11, R12)
 	LDP	-16(R8), (R13, R14)
 	LDP.P	32(R9), (R15, R16)
 	LDP	-16(R9), (R17, R18)
 	SBCS	R15, R11
 	SBCS	R16, R12
 	SBCS	R17, R13
 	SBCS	R18, R14
 	STP.P	(R11, R12), 32(R10)
 	STP	(R13, R14), -16(R10)
 	SUB	$4, R0
 	B	loop
 done:
 	CSET	LO, R0 // extract carry flag
 	CSET	LO, R0		// extract carry flag
 	MOVD	R0, c+72(FP)
 	RET
-...

View file @ c4f3fe95

@@ -142,6 +142,23 @@ func BenchmarkAddVV(b *testing.B) {
+	}
+}
 func BenchmarkSubVV(b *testing.B) {
 	for _, n := range benchSizes {
 		if isRaceBuilder && n > 1e3 {
 			continue
+		}
 		x := rndV(n)
 		y := rndV(n)
 		z := make([]Word, n)
 		b.Run(fmt.Sprint(n), func(b *testing.B) {
 			b.SetBytes(int64(n * _W))
 			for i := 0; i < b.N; i++ {
 				subVV(z, x, y)
+			}
 		})
+	}
+}
 type funVW func(z, x []Word, y Word) (c Word)
 type argVW struct {
 	z, x nat
-...

Please register or to comment