mirror of
https://github.com/golang/go
synced 2024-11-18 17:34:51 -07:00
f41451e7eb
On modern 64bit CPUs a SHR, SHL or AND instruction take 1 cycle to execute. A pair of shifts that operate on the same register will take 2 cycles and needs to wait for the input register value to be available. Large constants used to mask the high bits of a register with an AND instruction can not be encoded as an immediate in the AND instruction on amd64 and therefore need to be loaded into a register with a MOV instruction. However that MOV instruction is not dependent on the output register and on many CPUs does not compete with the AND or shift instructions for execution ports. Using a pair of shifts to mask high bits instead of an AND to mask high bits of a register has a shorter encoding and uses one less general purpose register but is slower due to taking one clock cycle longer if there is no register pressure that would make the AND variant need to generate a spill. For example the instructions emitted for (x & 1 << 63) before this CL are: 48c1ea3f SHRQ $0x3f, DX 48c1e23f SHLQ $0x3f, DX after this CL the instructions are the same as GCC and LLVM use: 48b80000000000000080 MOVQ $0x8000000000000000, AX 4821d0 ANDQ DX, AX Some platforms such as arm64 already have SSA optimization rules to fuse two shift instructions back into an AND. Removing the general rule to rewrite AND to SHR+SHL speeds up this benchmark: var GlobalU uint func BenchmarkAndHighBits(b *testing.B) { x := uint(0) for i := 0; i < b.N; i++ { x &= 1 << 63 } GlobalU = x } amd64/darwin on Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz: name old time/op new time/op delta AndHighBits-4 0.61ns ± 6% 0.42ns ± 6% -31.42% (p=0.000 n=25+25): 'go run run.go -all_codegen -v codegen' passes with following adjustments: ARM64: The BFXIL pattern ((x << lc) >> rc | y & ac) needed adjustment since ORshiftRL generation fusing '>> rc' and '|' interferes with matching ((x << lc) >> rc) to generate UBFX. Previously ORshiftLL was created first using the shifts generated for (y & ac). S390X: Add rules for abs and copysign to match use of AND instead of SHIFTs. Updates #33826 Updates #32781 Change-Id: I5a59f6239660d53c029cd22dfb44ddf39f93a56c Reviewed-on: https://go-review.googlesource.com/c/go/+/196810 Run-TryBot: Martin Möhrmann <moehrmann@google.com> Reviewed-by: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
179 lines
4.4 KiB
Go
179 lines
4.4 KiB
Go
// asmcheck
|
|
|
|
// Copyright 2018 The Go Authors. All rights reserved.
|
|
// Use of this source code is governed by a BSD-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
package codegen
|
|
|
|
import "math"
|
|
|
|
var sink64 [8]float64
|
|
|
|
func approx(x float64) {
|
|
// s390x:"FIDBR\t[$]6"
|
|
// arm64:"FRINTPD"
|
|
// ppc64:"FRIP"
|
|
// ppc64le:"FRIP"
|
|
// wasm:"F64Ceil"
|
|
sink64[0] = math.Ceil(x)
|
|
|
|
// s390x:"FIDBR\t[$]7"
|
|
// arm64:"FRINTMD"
|
|
// ppc64:"FRIM"
|
|
// ppc64le:"FRIM"
|
|
// wasm:"F64Floor"
|
|
sink64[1] = math.Floor(x)
|
|
|
|
// s390x:"FIDBR\t[$]1"
|
|
// arm64:"FRINTAD"
|
|
// ppc64:"FRIN"
|
|
// ppc64le:"FRIN"
|
|
sink64[2] = math.Round(x)
|
|
|
|
// s390x:"FIDBR\t[$]5"
|
|
// arm64:"FRINTZD"
|
|
// ppc64:"FRIZ"
|
|
// ppc64le:"FRIZ"
|
|
// wasm:"F64Trunc"
|
|
sink64[3] = math.Trunc(x)
|
|
|
|
// s390x:"FIDBR\t[$]4"
|
|
// arm64:"FRINTND"
|
|
// wasm:"F64Nearest"
|
|
sink64[4] = math.RoundToEven(x)
|
|
}
|
|
|
|
func sqrt(x float64) float64 {
|
|
// amd64:"SQRTSD"
|
|
// 386/387:"FSQRT" 386/sse2:"SQRTSD"
|
|
// arm64:"FSQRTD"
|
|
// arm/7:"SQRTD"
|
|
// mips/hardfloat:"SQRTD" mips/softfloat:-"SQRTD"
|
|
// mips64/hardfloat:"SQRTD" mips64/softfloat:-"SQRTD"
|
|
// wasm:"F64Sqrt"
|
|
return math.Sqrt(x)
|
|
}
|
|
|
|
// Check that it's using integer registers
|
|
func abs(x, y float64) {
|
|
// amd64:"BTRQ\t[$]63"
|
|
// arm64:"FABSD\t"
|
|
// s390x:"LPDFR\t",-"MOVD\t" (no integer load/store)
|
|
// ppc64:"FABS\t"
|
|
// ppc64le:"FABS\t"
|
|
// wasm:"F64Abs"
|
|
// arm/6:"ABSD\t"
|
|
sink64[0] = math.Abs(x)
|
|
|
|
// amd64:"BTRQ\t[$]63","PXOR" (TODO: this should be BTSQ)
|
|
// s390x:"LNDFR\t",-"MOVD\t" (no integer load/store)
|
|
// ppc64:"FNABS\t"
|
|
// ppc64le:"FNABS\t"
|
|
sink64[1] = -math.Abs(y)
|
|
}
|
|
|
|
// Check that it's using integer registers
|
|
func abs32(x float32) float32 {
|
|
// s390x:"LPDFR",-"LDEBR",-"LEDBR" (no float64 conversion)
|
|
return float32(math.Abs(float64(x)))
|
|
}
|
|
|
|
// Check that it's using integer registers
|
|
func copysign(a, b, c float64) {
|
|
// amd64:"BTRQ\t[$]63","ANDQ","ORQ"
|
|
// s390x:"CPSDR",-"MOVD" (no integer load/store)
|
|
// ppc64:"FCPSGN"
|
|
// ppc64le:"FCPSGN"
|
|
// wasm:"F64Copysign"
|
|
sink64[0] = math.Copysign(a, b)
|
|
|
|
// amd64:"BTSQ\t[$]63"
|
|
// s390x:"LNDFR\t",-"MOVD\t" (no integer load/store)
|
|
// ppc64:"FCPSGN"
|
|
// ppc64le:"FCPSGN"
|
|
// arm64:"ORR", -"AND"
|
|
sink64[1] = math.Copysign(c, -1)
|
|
|
|
// Like math.Copysign(c, -1), but with integer operations. Useful
|
|
// for platforms that have a copysign opcode to see if it's detected.
|
|
// s390x:"LNDFR\t",-"MOVD\t" (no integer load/store)
|
|
sink64[2] = math.Float64frombits(math.Float64bits(a) | 1<<63)
|
|
|
|
// amd64:"ANDQ","ORQ"
|
|
// s390x:"CPSDR\t",-"MOVD\t" (no integer load/store)
|
|
// ppc64:"FCPSGN"
|
|
// ppc64le:"FCPSGN"
|
|
sink64[3] = math.Copysign(-1, c)
|
|
}
|
|
|
|
func fromFloat64(f64 float64) uint64 {
|
|
// amd64:"MOVQ\tX.*, [^X].*"
|
|
// arm64:"FMOVD\tF.*, R.*"
|
|
// ppc64:"MFVSRD"
|
|
// ppc64le:"MFVSRD"
|
|
return math.Float64bits(f64+1) + 1
|
|
}
|
|
|
|
func fromFloat32(f32 float32) uint32 {
|
|
// amd64:"MOVL\tX.*, [^X].*"
|
|
// arm64:"FMOVS\tF.*, R.*"
|
|
return math.Float32bits(f32+1) + 1
|
|
}
|
|
|
|
func toFloat64(u64 uint64) float64 {
|
|
// amd64:"MOVQ\t[^X].*, X.*"
|
|
// arm64:"FMOVD\tR.*, F.*"
|
|
// ppc64:"MTVSRD"
|
|
// ppc64le:"MTVSRD"
|
|
return math.Float64frombits(u64+1) + 1
|
|
}
|
|
|
|
func toFloat32(u32 uint32) float32 {
|
|
// amd64:"MOVL\t[^X].*, X.*"
|
|
// arm64:"FMOVS\tR.*, F.*"
|
|
return math.Float32frombits(u32+1) + 1
|
|
}
|
|
|
|
// Test that comparisons with constants converted to float
|
|
// are evaluated at compile-time
|
|
|
|
func constantCheck64() bool {
|
|
// amd64:"MOVB\t[$]0",-"FCMP",-"MOVB\t[$]1"
|
|
// s390x:"MOV(B|BZ|D)\t[$]0,",-"FCMPU",-"MOV(B|BZ|D)\t[$]1,"
|
|
return 0.5 == float64(uint32(1)) || 1.5 > float64(uint64(1<<63)) || math.NaN() == math.NaN()
|
|
}
|
|
|
|
func constantCheck32() bool {
|
|
// amd64:"MOVB\t[$]1",-"FCMP",-"MOVB\t[$]0"
|
|
// s390x:"MOV(B|BZ|D)\t[$]1,",-"FCMPU",-"MOV(B|BZ|D)\t[$]0,"
|
|
return float32(0.5) <= float32(int64(1)) && float32(1.5) >= float32(int32(-1<<31)) && float32(math.NaN()) != float32(math.NaN())
|
|
}
|
|
|
|
// Test that integer constants are converted to floating point constants
|
|
// at compile-time
|
|
|
|
func constantConvert32(x float32) float32 {
|
|
// amd64:"MOVSS\t[$]f32.3f800000\\(SB\\)"
|
|
// s390x:"FMOVS\t[$]f32.3f800000\\(SB\\)"
|
|
// ppc64:"FMOVS\t[$]f32.3f800000\\(SB\\)"
|
|
// ppc64le:"FMOVS\t[$]f32.3f800000\\(SB\\)"
|
|
// arm64:"FMOVS\t[$]\\(1.0\\)"
|
|
if x > math.Float32frombits(0x3f800000) {
|
|
return -x
|
|
}
|
|
return x
|
|
}
|
|
|
|
func constantConvertInt32(x uint32) uint32 {
|
|
// amd64:-"MOVSS"
|
|
// s390x:-"FMOVS"
|
|
// ppc64:-"FMOVS"
|
|
// ppc64le:-"FMOVS"
|
|
// arm64:-"FMOVS"
|
|
if x > math.Float32bits(1) {
|
|
return -x
|
|
}
|
|
return x
|
|
}
|