mirror of
https://github.com/golang/go
synced 2024-11-17 21:24:55 -07:00
0301c6c351
Use a small python script to consolidate duplicate
ppc64/ppc64le tests into a single ppc64x codegen test.
This makes small assumption that anytime two tests with
for different arch/variant combos exists, those tests
can be combined into a single ppc64x test.
E.x:
// ppc64le: foo
// ppc64le/power9: foo
into
// ppc64x: foo
or
// ppc64: foo
// ppc64le: foo
into
// ppc64x: foo
import glob
import re
files = glob.glob("codegen/*.go")
for file in files:
with open(file) as f:
text = [l for l in f]
i = 0
while i < len(text):
first = re.match("\s*// ?ppc64(le)?(/power[89])?:(.*)", text[i])
if first:
j = i+1
while j < len(text):
second = re.match("\s*// ?ppc64(le)?(/power[89])?:(.*)", text[j])
if not second:
break
if (not first.group(2) or first.group(2) == second.group(2)) and first.group(3) == second.group(3):
text[i] = re.sub(" ?ppc64(le|x)?"," ppc64x",text[i])
text=text[:j] + (text[j+1:])
else:
j += 1
i+=1
with open(file, 'w') as f:
f.write("".join(text))
Change-Id: Ic6b009b54eacaadc5a23db9c5a3bf7331b595821
Reviewed-on: https://go-review.googlesource.com/c/go/+/463220
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Bryan Mills <bcmills@google.com>
Run-TryBot: Paul Murphy <murp@ibm.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
260 lines
5.3 KiB
Go
260 lines
5.3 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"
|
|
|
|
// ------------------- //
|
|
// const rotates //
|
|
// ------------------- //
|
|
|
|
func rot64(x uint64) uint64 {
|
|
var a uint64
|
|
|
|
// amd64:"ROLQ\t[$]7"
|
|
// ppc64x:"ROTL\t[$]7"
|
|
// loong64: "ROTRV\t[$]57"
|
|
a += x<<7 | x>>57
|
|
|
|
// amd64:"ROLQ\t[$]8"
|
|
// arm64:"ROR\t[$]56"
|
|
// s390x:"RISBGZ\t[$]0, [$]63, [$]8, "
|
|
// ppc64x:"ROTL\t[$]8"
|
|
// loong64: "ROTRV\t[$]56"
|
|
a += x<<8 + x>>56
|
|
|
|
// amd64:"ROLQ\t[$]9"
|
|
// arm64:"ROR\t[$]55"
|
|
// s390x:"RISBGZ\t[$]0, [$]63, [$]9, "
|
|
// ppc64x:"ROTL\t[$]9"
|
|
// loong64: "ROTRV\t[$]55"
|
|
a += x<<9 ^ x>>55
|
|
|
|
// amd64:"ROLQ\t[$]10"
|
|
// arm64:"ROR\t[$]54"
|
|
// s390x:"RISBGZ\t[$]0, [$]63, [$]10, "
|
|
// ppc64x:"ROTL\t[$]10"
|
|
// arm64:"ROR\t[$]54"
|
|
// s390x:"RISBGZ\t[$]0, [$]63, [$]10, "
|
|
// loong64: "ROTRV\t[$]54"
|
|
a += bits.RotateLeft64(x, 10)
|
|
|
|
return a
|
|
}
|
|
|
|
func rot32(x uint32) uint32 {
|
|
var a uint32
|
|
|
|
// amd64:"ROLL\t[$]7"
|
|
// arm:"MOVW\tR\\d+@>25"
|
|
// ppc64x:"ROTLW\t[$]7"
|
|
// loong64: "ROTR\t[$]25"
|
|
a += x<<7 | x>>25
|
|
|
|
// amd64:`ROLL\t[$]8`
|
|
// arm:"MOVW\tR\\d+@>24"
|
|
// arm64:"RORW\t[$]24"
|
|
// s390x:"RLL\t[$]8"
|
|
// ppc64x:"ROTLW\t[$]8"
|
|
// loong64: "ROTR\t[$]24"
|
|
a += x<<8 + x>>24
|
|
|
|
// amd64:"ROLL\t[$]9"
|
|
// arm:"MOVW\tR\\d+@>23"
|
|
// arm64:"RORW\t[$]23"
|
|
// s390x:"RLL\t[$]9"
|
|
// ppc64x:"ROTLW\t[$]9"
|
|
// loong64: "ROTR\t[$]23"
|
|
a += x<<9 ^ x>>23
|
|
|
|
// amd64:"ROLL\t[$]10"
|
|
// arm:"MOVW\tR\\d+@>22"
|
|
// arm64:"RORW\t[$]22"
|
|
// s390x:"RLL\t[$]10"
|
|
// ppc64x:"ROTLW\t[$]10"
|
|
// arm64:"RORW\t[$]22"
|
|
// s390x:"RLL\t[$]10"
|
|
// loong64: "ROTR\t[$]22"
|
|
a += bits.RotateLeft32(x, 10)
|
|
|
|
return a
|
|
}
|
|
|
|
func rot16(x uint16) uint16 {
|
|
var a uint16
|
|
|
|
// amd64:"ROLW\t[$]7"
|
|
a += x<<7 | x>>9
|
|
|
|
// amd64:`ROLW\t[$]8`
|
|
a += x<<8 + x>>8
|
|
|
|
// amd64:"ROLW\t[$]9"
|
|
a += x<<9 ^ x>>7
|
|
|
|
return a
|
|
}
|
|
|
|
func rot8(x uint8) uint8 {
|
|
var a uint8
|
|
|
|
// amd64:"ROLB\t[$]5"
|
|
a += x<<5 | x>>3
|
|
|
|
// amd64:`ROLB\t[$]6`
|
|
a += x<<6 + x>>2
|
|
|
|
// amd64:"ROLB\t[$]7"
|
|
a += x<<7 ^ x>>1
|
|
|
|
return a
|
|
}
|
|
|
|
// ----------------------- //
|
|
// non-const rotates //
|
|
// ----------------------- //
|
|
|
|
func rot64nc(x uint64, z uint) uint64 {
|
|
var a uint64
|
|
|
|
z &= 63
|
|
|
|
// amd64:"ROLQ",-"AND"
|
|
// arm64:"ROR","NEG",-"AND"
|
|
// ppc64x:"ROTL",-"NEG",-"AND"
|
|
// loong64: "ROTRV", -"AND"
|
|
a += x<<z | x>>(64-z)
|
|
|
|
// amd64:"RORQ",-"AND"
|
|
// arm64:"ROR",-"NEG",-"AND"
|
|
// ppc64x:"ROTL","NEG",-"AND"
|
|
// loong64: "ROTRV", -"AND"
|
|
a += x>>z | x<<(64-z)
|
|
|
|
return a
|
|
}
|
|
|
|
func rot32nc(x uint32, z uint) uint32 {
|
|
var a uint32
|
|
|
|
z &= 31
|
|
|
|
// amd64:"ROLL",-"AND"
|
|
// arm64:"ROR","NEG",-"AND"
|
|
// ppc64x:"ROTLW",-"NEG",-"AND"
|
|
// loong64: "ROTR", -"AND"
|
|
a += x<<z | x>>(32-z)
|
|
|
|
// amd64:"RORL",-"AND"
|
|
// arm64:"ROR",-"NEG",-"AND"
|
|
// ppc64x:"ROTLW","NEG",-"AND"
|
|
// loong64: "ROTR", -"AND"
|
|
a += x>>z | x<<(32-z)
|
|
|
|
return a
|
|
}
|
|
|
|
func rot16nc(x uint16, z uint) uint16 {
|
|
var a uint16
|
|
|
|
z &= 15
|
|
|
|
// amd64:"ROLW",-"ANDQ"
|
|
a += x<<z | x>>(16-z)
|
|
|
|
// amd64:"RORW",-"ANDQ"
|
|
a += x>>z | x<<(16-z)
|
|
|
|
return a
|
|
}
|
|
|
|
func rot8nc(x uint8, z uint) uint8 {
|
|
var a uint8
|
|
|
|
z &= 7
|
|
|
|
// amd64:"ROLB",-"ANDQ"
|
|
a += x<<z | x>>(8-z)
|
|
|
|
// amd64:"RORB",-"ANDQ"
|
|
a += x>>z | x<<(8-z)
|
|
|
|
return a
|
|
}
|
|
|
|
// Issue 18254: rotate after inlining
|
|
func f32(x uint32) uint32 {
|
|
// amd64:"ROLL\t[$]7"
|
|
return rot32nc(x, 7)
|
|
}
|
|
|
|
func doubleRotate(x uint64) uint64 {
|
|
x = (x << 5) | (x >> 59)
|
|
// amd64:"ROLQ\t[$]15"
|
|
// arm64:"ROR\t[$]49"
|
|
x = (x << 10) | (x >> 54)
|
|
return x
|
|
}
|
|
|
|
// --------------------------------------- //
|
|
// Combined Rotate + Masking operations //
|
|
// --------------------------------------- //
|
|
|
|
func checkMaskedRotate32(a []uint32, r int) {
|
|
i := 0
|
|
|
|
// ppc64x: "RLWNM\t[$]16, R[0-9]+, [$]8, [$]15, R[0-9]+"
|
|
a[i] = bits.RotateLeft32(a[i], 16) & 0xFF0000
|
|
i++
|
|
// ppc64x: "RLWNM\t[$]16, R[0-9]+, [$]8, [$]15, R[0-9]+"
|
|
a[i] = bits.RotateLeft32(a[i]&0xFF, 16)
|
|
i++
|
|
// ppc64x: "RLWNM\t[$]4, R[0-9]+, [$]20, [$]27, R[0-9]+"
|
|
a[i] = bits.RotateLeft32(a[i], 4) & 0xFF0
|
|
i++
|
|
// ppc64x: "RLWNM\t[$]16, R[0-9]+, [$]24, [$]31, R[0-9]+"
|
|
a[i] = bits.RotateLeft32(a[i]&0xFF0000, 16)
|
|
i++
|
|
|
|
// ppc64x: "RLWNM\tR[0-9]+, R[0-9]+, [$]8, [$]15, R[0-9]+"
|
|
a[i] = bits.RotateLeft32(a[i], r) & 0xFF0000
|
|
i++
|
|
// ppc64x: "RLWNM\tR[0-9]+, R[0-9]+, [$]16, [$]23, R[0-9]+"
|
|
a[i] = bits.RotateLeft32(a[i], r) & 0xFF00
|
|
i++
|
|
|
|
// ppc64x: "RLWNM\tR[0-9]+, R[0-9]+, [$]20, [$]11, R[0-9]+"
|
|
a[i] = bits.RotateLeft32(a[i], r) & 0xFFF00FFF
|
|
i++
|
|
// ppc64x: "RLWNM\t[$]4, R[0-9]+, [$]20, [$]11, R[0-9]+"
|
|
a[i] = bits.RotateLeft32(a[i], 4) & 0xFFF00FFF
|
|
i++
|
|
}
|
|
|
|
// combined arithmetic and rotate on arm64
|
|
func checkArithmeticWithRotate(a *[1000]uint64) {
|
|
// arm64: "AND\tR[0-9]+@>51, R[0-9]+, R[0-9]+"
|
|
a[2] = a[1] & bits.RotateLeft64(a[0], 13)
|
|
// arm64: "ORR\tR[0-9]+@>51, R[0-9]+, R[0-9]+"
|
|
a[5] = a[4] | bits.RotateLeft64(a[3], 13)
|
|
// arm64: "EOR\tR[0-9]+@>51, R[0-9]+, R[0-9]+"
|
|
a[8] = a[7] ^ bits.RotateLeft64(a[6], 13)
|
|
// arm64: "MVN\tR[0-9]+@>51, R[0-9]+"
|
|
a[10] = ^bits.RotateLeft64(a[9], 13)
|
|
// arm64: "BIC\tR[0-9]+@>51, R[0-9]+, R[0-9]+"
|
|
a[13] = a[12] &^ bits.RotateLeft64(a[11], 13)
|
|
// arm64: "EON\tR[0-9]+@>51, R[0-9]+, R[0-9]+"
|
|
a[16] = a[15] ^ ^bits.RotateLeft64(a[14], 13)
|
|
// arm64: "ORN\tR[0-9]+@>51, R[0-9]+, R[0-9]+"
|
|
a[19] = a[18] | ^bits.RotateLeft64(a[17], 13)
|
|
// arm64: "TST\tR[0-9]+@>51, R[0-9]+"
|
|
if a[18]&bits.RotateLeft64(a[19], 13) == 0 {
|
|
a[20] = 1
|
|
}
|
|
|
|
}
|