mirror of
https://github.com/golang/go
synced 2024-09-28 23:14:38 -06:00
159b2de442
Benchmark:
name old time/op new time/op delta
Div-8 22.0ns ± 0% 22.0ns ± 0% ~ (all equal)
Div32-8 6.51ns ± 0% 3.00ns ± 0% -53.90% (p=0.000 n=10+8)
Div64-8 22.5ns ± 0% 22.5ns ± 0% ~ (all equal)
Code:
func div32(hi, lo, y uint32) (q, r uint32) {return bits.Div32(hi, lo, y)}
Before:
0x0020 00032 (test.go:24) MOVWU "".y+8(FP), R0
0x0024 00036 ($GOROOT/src/math/bits/bits.go:472) CBZW R0, 132
0x0028 00040 ($GOROOT/src/math/bits/bits.go:472) MOVWU "".hi(FP), R1
0x002c 00044 ($GOROOT/src/math/bits/bits.go:472) CMPW R1, R0
0x0030 00048 ($GOROOT/src/math/bits/bits.go:472) BLS 96
0x0034 00052 ($GOROOT/src/math/bits/bits.go:475) MOVWU "".lo+4(FP), R2
0x0038 00056 ($GOROOT/src/math/bits/bits.go:475) ORR R1<<32, R2, R1
0x003c 00060 ($GOROOT/src/math/bits/bits.go:476) CBZ R0, 140
0x0040 00064 ($GOROOT/src/math/bits/bits.go:476) UDIV R0, R1, R2
0x0044 00068 (test.go:24) MOVW R2, "".q+16(FP)
0x0048 00072 ($GOROOT/src/math/bits/bits.go:476) UREM R0, R1, R0
0x0050 00080 (test.go:24) MOVW R0, "".r+20(FP)
0x0054 00084 (test.go:24) MOVD -8(RSP), R29
0x0058 00088 (test.go:24) MOVD.P 32(RSP), R30
0x005c 00092 (test.go:24) RET (R30)
After:
0x001c 00028 (test.go:24) MOVWU "".y+8(FP), R0
0x0020 00032 (test.go:24) CBZW R0, 92
0x0024 00036 (test.go:24) MOVWU "".hi(FP), R1
0x0028 00040 (test.go:24) CMPW R0, R1
0x002c 00044 (test.go:24) BHS 84
0x0030 00048 (test.go:24) MOVWU "".lo+4(FP), R2
0x0034 00052 (test.go:24) ORR R1<<32, R2, R4
0x0038 00056 (test.go:24) UDIV R0, R4, R3
0x003c 00060 (test.go:24) MSUB R3, R4, R0, R4
0x0040 00064 (test.go:24) MOVW R3, "".q+16(FP)
0x0044 00068 (test.go:24) MOVW R4, "".r+20(FP)
0x0048 00072 (test.go:24) MOVD -8(RSP), R29
0x004c 00076 (test.go:24) MOVD.P 16(RSP), R30
0x0050 00080 (test.go:24) RET (R30)
UREM instruction in the previous assembly code will be converted to UDIV and MSUB instructions
on arm64. However the UDIV instruction in UREM is unnecessary, because it's a duplicate of the
previous UDIV. This CL adds a rule to have this extra UDIV instruction removed by CSE.
Change-Id: Ie2508784320020b2de022806d09f75a7871bb3d7
Reviewed-on: https://go-review.googlesource.com/c/159577
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
488 lines
9.5 KiB
Go
488 lines
9.5 KiB
Go
// asmcheck
|
|
|
|
// Copyright 2018 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 codegen
|
|
|
|
import "math/bits"
|
|
|
|
// ----------------------- //
|
|
// bits.LeadingZeros //
|
|
// ----------------------- //
|
|
|
|
func LeadingZeros(n uint) int {
|
|
// amd64:"BSRQ"
|
|
// s390x:"FLOGR"
|
|
// arm:"CLZ" arm64:"CLZ"
|
|
// mips:"CLZ"
|
|
return bits.LeadingZeros(n)
|
|
}
|
|
|
|
func LeadingZeros64(n uint64) int {
|
|
// amd64:"BSRQ"
|
|
// s390x:"FLOGR"
|
|
// arm:"CLZ" arm64:"CLZ"
|
|
// mips:"CLZ"
|
|
return bits.LeadingZeros64(n)
|
|
}
|
|
|
|
func LeadingZeros32(n uint32) int {
|
|
// amd64:"BSRQ","LEAQ",-"CMOVQEQ"
|
|
// s390x:"FLOGR"
|
|
// arm:"CLZ" arm64:"CLZW"
|
|
// mips:"CLZ"
|
|
return bits.LeadingZeros32(n)
|
|
}
|
|
|
|
func LeadingZeros16(n uint16) int {
|
|
// amd64:"BSRL","LEAL",-"CMOVQEQ"
|
|
// s390x:"FLOGR"
|
|
// arm:"CLZ" arm64:"CLZ"
|
|
// mips:"CLZ"
|
|
return bits.LeadingZeros16(n)
|
|
}
|
|
|
|
func LeadingZeros8(n uint8) int {
|
|
// amd64:"BSRL","LEAL",-"CMOVQEQ"
|
|
// s390x:"FLOGR"
|
|
// arm:"CLZ" arm64:"CLZ"
|
|
// mips:"CLZ"
|
|
return bits.LeadingZeros8(n)
|
|
}
|
|
|
|
// --------------- //
|
|
// bits.Len* //
|
|
// --------------- //
|
|
|
|
func Len(n uint) int {
|
|
// amd64:"BSRQ"
|
|
// s390x:"FLOGR"
|
|
// arm:"CLZ" arm64:"CLZ"
|
|
// mips:"CLZ"
|
|
return bits.Len(n)
|
|
}
|
|
|
|
func Len64(n uint64) int {
|
|
// amd64:"BSRQ"
|
|
// s390x:"FLOGR"
|
|
// arm:"CLZ" arm64:"CLZ"
|
|
// mips:"CLZ"
|
|
return bits.Len64(n)
|
|
}
|
|
|
|
func Len32(n uint32) int {
|
|
// amd64:"BSRQ","LEAQ",-"CMOVQEQ"
|
|
// s390x:"FLOGR"
|
|
// arm:"CLZ" arm64:"CLZ"
|
|
// mips:"CLZ"
|
|
return bits.Len32(n)
|
|
}
|
|
|
|
func Len16(n uint16) int {
|
|
// amd64:"BSRL","LEAL",-"CMOVQEQ"
|
|
// s390x:"FLOGR"
|
|
// arm:"CLZ" arm64:"CLZ"
|
|
// mips:"CLZ"
|
|
return bits.Len16(n)
|
|
}
|
|
|
|
func Len8(n uint8) int {
|
|
// amd64:"BSRL","LEAL",-"CMOVQEQ"
|
|
// s390x:"FLOGR"
|
|
// arm:"CLZ" arm64:"CLZ"
|
|
// mips:"CLZ"
|
|
return bits.Len8(n)
|
|
}
|
|
|
|
// -------------------- //
|
|
// bits.OnesCount //
|
|
// -------------------- //
|
|
|
|
func OnesCount(n uint) int {
|
|
// amd64:"POPCNTQ",".*x86HasPOPCNT"
|
|
// arm64:"VCNT","VUADDLV"
|
|
// s390x:"POPCNT"
|
|
// ppc64:"POPCNTD"
|
|
// ppc64le:"POPCNTD"
|
|
return bits.OnesCount(n)
|
|
}
|
|
|
|
func OnesCount64(n uint64) int {
|
|
// amd64:"POPCNTQ",".*x86HasPOPCNT"
|
|
// arm64:"VCNT","VUADDLV"
|
|
// s390x:"POPCNT"
|
|
// ppc64:"POPCNTD"
|
|
// ppc64le:"POPCNTD"
|
|
return bits.OnesCount64(n)
|
|
}
|
|
|
|
func OnesCount32(n uint32) int {
|
|
// amd64:"POPCNTL",".*x86HasPOPCNT"
|
|
// arm64:"VCNT","VUADDLV"
|
|
// s390x:"POPCNT"
|
|
// ppc64:"POPCNTW"
|
|
// ppc64le:"POPCNTW"
|
|
return bits.OnesCount32(n)
|
|
}
|
|
|
|
func OnesCount16(n uint16) int {
|
|
// amd64:"POPCNTL",".*x86HasPOPCNT"
|
|
// arm64:"VCNT","VUADDLV"
|
|
// s390x:"POPCNT"
|
|
// ppc64:"POPCNTW"
|
|
// ppc64le:"POPCNTW"
|
|
return bits.OnesCount16(n)
|
|
}
|
|
|
|
func OnesCount8(n uint8) int {
|
|
// s390x:"POPCNT"
|
|
// ppc64:"POPCNTB"
|
|
// ppc64le:"POPCNTB"
|
|
return bits.OnesCount8(n)
|
|
}
|
|
|
|
// ----------------------- //
|
|
// bits.ReverseBytes //
|
|
// ----------------------- //
|
|
|
|
func ReverseBytes(n uint) uint {
|
|
// amd64:"BSWAPQ"
|
|
// s390x:"MOVDBR"
|
|
// arm64:"REV"
|
|
return bits.ReverseBytes(n)
|
|
}
|
|
|
|
func ReverseBytes64(n uint64) uint64 {
|
|
// amd64:"BSWAPQ"
|
|
// s390x:"MOVDBR"
|
|
// arm64:"REV"
|
|
return bits.ReverseBytes64(n)
|
|
}
|
|
|
|
func ReverseBytes32(n uint32) uint32 {
|
|
// amd64:"BSWAPL"
|
|
// s390x:"MOVWBR"
|
|
// arm64:"REVW"
|
|
return bits.ReverseBytes32(n)
|
|
}
|
|
|
|
func ReverseBytes16(n uint16) uint16 {
|
|
// amd64:"ROLW"
|
|
// arm64:"REV16W",-"UBFX",-"ORR"
|
|
return bits.ReverseBytes16(n)
|
|
}
|
|
|
|
// --------------------- //
|
|
// bits.RotateLeft //
|
|
// --------------------- //
|
|
|
|
func RotateLeft64(n uint64) uint64 {
|
|
// amd64:"ROLQ"
|
|
// arm64:"ROR"
|
|
// ppc64:"ROTL"
|
|
// ppc64le:"ROTL"
|
|
// s390x:"RLLG"
|
|
return bits.RotateLeft64(n, 37)
|
|
}
|
|
|
|
func RotateLeft32(n uint32) uint32 {
|
|
// amd64:"ROLL" 386:"ROLL"
|
|
// arm64:"RORW"
|
|
// ppc64:"ROTLW"
|
|
// ppc64le:"ROTLW"
|
|
// s390x:"RLL"
|
|
return bits.RotateLeft32(n, 9)
|
|
}
|
|
|
|
func RotateLeft16(n uint16) uint16 {
|
|
// amd64:"ROLW" 386:"ROLW"
|
|
return bits.RotateLeft16(n, 5)
|
|
}
|
|
|
|
func RotateLeft8(n uint8) uint8 {
|
|
// amd64:"ROLB" 386:"ROLB"
|
|
return bits.RotateLeft8(n, 5)
|
|
}
|
|
|
|
func RotateLeftVariable(n uint, m int) uint {
|
|
// amd64:"ROLQ"
|
|
// arm64:"ROR"
|
|
// ppc64:"ROTL"
|
|
// ppc64le:"ROTL"
|
|
// s390x:"RLLG"
|
|
return bits.RotateLeft(n, m)
|
|
}
|
|
|
|
func RotateLeftVariable64(n uint64, m int) uint64 {
|
|
// amd64:"ROLQ"
|
|
// arm64:"ROR"
|
|
// ppc64:"ROTL"
|
|
// ppc64le:"ROTL"
|
|
// s390x:"RLLG"
|
|
return bits.RotateLeft64(n, m)
|
|
}
|
|
|
|
func RotateLeftVariable32(n uint32, m int) uint32 {
|
|
// amd64:"ROLL"
|
|
// arm64:"RORW"
|
|
// ppc64:"ROTLW"
|
|
// ppc64le:"ROTLW"
|
|
// s390x:"RLL"
|
|
return bits.RotateLeft32(n, m)
|
|
}
|
|
|
|
// ------------------------ //
|
|
// bits.TrailingZeros //
|
|
// ------------------------ //
|
|
|
|
func TrailingZeros(n uint) int {
|
|
// amd64:"BSFQ","MOVL\t\\$64","CMOVQEQ"
|
|
// s390x:"FLOGR"
|
|
// ppc64:"ANDN","POPCNTD"
|
|
// ppc64le:"ANDN","POPCNTD"
|
|
return bits.TrailingZeros(n)
|
|
}
|
|
|
|
func TrailingZeros64(n uint64) int {
|
|
// amd64:"BSFQ","MOVL\t\\$64","CMOVQEQ"
|
|
// s390x:"FLOGR"
|
|
// ppc64:"ANDN","POPCNTD"
|
|
// ppc64le:"ANDN","POPCNTD"
|
|
return bits.TrailingZeros64(n)
|
|
}
|
|
|
|
func TrailingZeros32(n uint32) int {
|
|
// amd64:"BTSQ\\t\\$32","BSFQ"
|
|
// s390x:"FLOGR","MOVWZ"
|
|
// ppc64:"ANDN","POPCNTW"
|
|
// ppc64le:"ANDN","POPCNTW"
|
|
return bits.TrailingZeros32(n)
|
|
}
|
|
|
|
func TrailingZeros16(n uint16) int {
|
|
// amd64:"BSFL","BTSL\\t\\$16"
|
|
// s390x:"FLOGR","OR\t\\$65536"
|
|
// ppc64:"POPCNTD","OR\\t\\$65536"
|
|
// ppc64le:"POPCNTD","OR\\t\\$65536"
|
|
return bits.TrailingZeros16(n)
|
|
}
|
|
|
|
func TrailingZeros8(n uint8) int {
|
|
// amd64:"BSFL","BTSL\\t\\$8"
|
|
// s390x:"FLOGR","OR\t\\$256"
|
|
return bits.TrailingZeros8(n)
|
|
}
|
|
|
|
// IterateBitsNN checks special handling of TrailingZerosNN when the input is known to be non-zero.
|
|
|
|
func IterateBits(n uint) int {
|
|
i := 0
|
|
for n != 0 {
|
|
// amd64:"BSFQ",-"CMOVEQ"
|
|
i += bits.TrailingZeros(n)
|
|
n &= n - 1
|
|
}
|
|
return i
|
|
}
|
|
|
|
func IterateBits64(n uint64) int {
|
|
i := 0
|
|
for n != 0 {
|
|
// amd64:"BSFQ",-"CMOVEQ"
|
|
i += bits.TrailingZeros64(n)
|
|
n &= n - 1
|
|
}
|
|
return i
|
|
}
|
|
|
|
func IterateBits32(n uint32) int {
|
|
i := 0
|
|
for n != 0 {
|
|
// amd64:"BSFL",-"BTSQ"
|
|
i += bits.TrailingZeros32(n)
|
|
n &= n - 1
|
|
}
|
|
return i
|
|
}
|
|
|
|
func IterateBits16(n uint16) int {
|
|
i := 0
|
|
for n != 0 {
|
|
// amd64:"BSFL",-"BTSL"
|
|
i += bits.TrailingZeros16(n)
|
|
n &= n - 1
|
|
}
|
|
return i
|
|
}
|
|
|
|
func IterateBits8(n uint8) int {
|
|
i := 0
|
|
for n != 0 {
|
|
// amd64:"BSFL",-"BTSL"
|
|
i += bits.TrailingZeros8(n)
|
|
n &= n - 1
|
|
}
|
|
return i
|
|
}
|
|
|
|
// --------------- //
|
|
// bits.Add* //
|
|
// --------------- //
|
|
|
|
func Add(x, y, ci uint) (r, co uint) {
|
|
// amd64:"NEGL","ADCQ","SBBQ","NEGQ"
|
|
return bits.Add(x, y, ci)
|
|
}
|
|
|
|
func AddC(x, ci uint) (r, co uint) {
|
|
// amd64:"NEGL","ADCQ","SBBQ","NEGQ"
|
|
return bits.Add(x, 7, ci)
|
|
}
|
|
|
|
func AddZ(x, y uint) (r, co uint) {
|
|
// amd64:"ADDQ","SBBQ","NEGQ",-"NEGL",-"ADCQ"
|
|
return bits.Add(x, y, 0)
|
|
}
|
|
|
|
func AddR(x, y, ci uint) uint {
|
|
// amd64:"NEGL","ADCQ",-"SBBQ",-"NEGQ"
|
|
r, _ := bits.Add(x, y, ci)
|
|
return r
|
|
}
|
|
func AddM(p, q, r *[3]uint) {
|
|
var c uint
|
|
r[0], c = bits.Add(p[0], q[0], c)
|
|
// amd64:"ADCQ",-"NEGL",-"SBBQ",-"NEGQ"
|
|
r[1], c = bits.Add(p[1], q[1], c)
|
|
r[2], c = bits.Add(p[2], q[2], c)
|
|
}
|
|
|
|
func Add64(x, y, ci uint64) (r, co uint64) {
|
|
// amd64:"NEGL","ADCQ","SBBQ","NEGQ"
|
|
return bits.Add64(x, y, ci)
|
|
}
|
|
|
|
func Add64C(x, ci uint64) (r, co uint64) {
|
|
// amd64:"NEGL","ADCQ","SBBQ","NEGQ"
|
|
return bits.Add64(x, 7, ci)
|
|
}
|
|
|
|
func Add64Z(x, y uint64) (r, co uint64) {
|
|
// amd64:"ADDQ","SBBQ","NEGQ",-"NEGL",-"ADCQ"
|
|
return bits.Add64(x, y, 0)
|
|
}
|
|
|
|
func Add64R(x, y, ci uint64) uint64 {
|
|
// amd64:"NEGL","ADCQ",-"SBBQ",-"NEGQ"
|
|
r, _ := bits.Add64(x, y, ci)
|
|
return r
|
|
}
|
|
func Add64M(p, q, r *[3]uint64) {
|
|
var c uint64
|
|
r[0], c = bits.Add64(p[0], q[0], c)
|
|
// amd64:"ADCQ",-"NEGL",-"SBBQ",-"NEGQ"
|
|
r[1], c = bits.Add64(p[1], q[1], c)
|
|
r[2], c = bits.Add64(p[2], q[2], c)
|
|
}
|
|
|
|
// --------------- //
|
|
// bits.Sub* //
|
|
// --------------- //
|
|
|
|
func Sub(x, y, ci uint) (r, co uint) {
|
|
// amd64:"NEGL","SBBQ","NEGQ"
|
|
return bits.Sub(x, y, ci)
|
|
}
|
|
|
|
func SubC(x, ci uint) (r, co uint) {
|
|
// amd64:"NEGL","SBBQ","NEGQ"
|
|
return bits.Sub(x, 7, ci)
|
|
}
|
|
|
|
func SubZ(x, y uint) (r, co uint) {
|
|
// amd64:"SUBQ","SBBQ","NEGQ",-"NEGL"
|
|
return bits.Sub(x, y, 0)
|
|
}
|
|
|
|
func SubR(x, y, ci uint) uint {
|
|
// amd64:"NEGL","SBBQ",-"NEGQ"
|
|
r, _ := bits.Sub(x, y, ci)
|
|
return r
|
|
}
|
|
func SubM(p, q, r *[3]uint) {
|
|
var c uint
|
|
r[0], c = bits.Sub(p[0], q[0], c)
|
|
// amd64:"SBBQ",-"NEGL",-"NEGQ"
|
|
r[1], c = bits.Sub(p[1], q[1], c)
|
|
r[2], c = bits.Sub(p[2], q[2], c)
|
|
}
|
|
|
|
func Sub64(x, y, ci uint64) (r, co uint64) {
|
|
// amd64:"NEGL","SBBQ","NEGQ"
|
|
return bits.Sub64(x, y, ci)
|
|
}
|
|
|
|
func Sub64C(x, ci uint64) (r, co uint64) {
|
|
// amd64:"NEGL","SBBQ","NEGQ"
|
|
return bits.Sub64(x, 7, ci)
|
|
}
|
|
|
|
func Sub64Z(x, y uint64) (r, co uint64) {
|
|
// amd64:"SUBQ","SBBQ","NEGQ",-"NEGL"
|
|
return bits.Sub64(x, y, 0)
|
|
}
|
|
|
|
func Sub64R(x, y, ci uint64) uint64 {
|
|
// amd64:"NEGL","SBBQ",-"NEGQ"
|
|
r, _ := bits.Sub64(x, y, ci)
|
|
return r
|
|
}
|
|
func Sub64M(p, q, r *[3]uint64) {
|
|
var c uint64
|
|
r[0], c = bits.Sub64(p[0], q[0], c)
|
|
// amd64:"SBBQ",-"NEGL",-"NEGQ"
|
|
r[1], c = bits.Sub64(p[1], q[1], c)
|
|
r[2], c = bits.Sub64(p[2], q[2], c)
|
|
}
|
|
|
|
// --------------- //
|
|
// bits.Mul* //
|
|
// --------------- //
|
|
|
|
func Mul(x, y uint) (hi, lo uint) {
|
|
// amd64:"MULQ"
|
|
// arm64:"UMULH","MUL"
|
|
// ppc64:"MULHDU","MULLD"
|
|
// ppc64le:"MULHDU","MULLD"
|
|
return bits.Mul(x, y)
|
|
}
|
|
|
|
func Mul64(x, y uint64) (hi, lo uint64) {
|
|
// amd64:"MULQ"
|
|
// arm64:"UMULH","MUL"
|
|
// ppc64:"MULHDU","MULLD"
|
|
// ppc64le:"MULHDU","MULLD"
|
|
return bits.Mul64(x, y)
|
|
}
|
|
|
|
// --------------- //
|
|
// bits.Div* //
|
|
// --------------- //
|
|
|
|
func Div(hi, lo, x uint) (q, r uint) {
|
|
// amd64:"DIVQ"
|
|
return bits.Div(hi, lo, x)
|
|
}
|
|
|
|
func Div32(hi, lo, x uint32) (q, r uint32) {
|
|
// arm64:"ORR","UDIV","MSUB",-"UREM"
|
|
return bits.Div32(hi, lo, x)
|
|
}
|
|
|
|
func Div64(hi, lo, x uint64) (q, r uint64) {
|
|
// amd64:"DIVQ"
|
|
return bits.Div64(hi, lo, x)
|
|
}
|