golang/go
0
1
Fork 0
mirror of https://github.com/golang/go synced 2025-04-29 02:27:46 +00:00
Code Packages Activity
Files
.github
api
doc
lib
misc
src
test
abi
alias3.dir
asmhdr.dir
bench
chan
closure3.dir
closure5.dir
codegen
README
addrcalc.go
alloc.go
arithmetic.go
bitfield.go
bits.go
bmi.go
bool.go
clobberdead.go
clobberdeadreg.go
compare_and_branch.go
comparisons.go
condmove.go
copy.go
floats.go
fuse.go
issue22703.go
issue25378.go
issue31618.go
issue33580.go
issue38554.go
issue42610.go
issue48054.go
issue52635.go
logic.go
mapaccess.go
maps.go
math.go
mathbits.go
memcombine.go
memops.go
noextend.go
race.go
regabi_regalloc.go
retpoline.go
rotate.go
select.go
shift.go
shortcircuit.go
slices.go
smallintiface.go
spectre.go
stack.go
strings.go
structs.go
switch.go
zerosize.go
ddd2.dir
dwarf
fixedbugs
import2.dir
import4.dir
interface
intrinsic.dir
ken
linkname.dir
method4.dir
retjmp.dir
runtime
stress
syntax
typeparam
uintptrescapes.dir
235.go
64bit.go
README.md
alg.go
alias.go
alias1.go
alias2.go
alias3.go
align.go
append.go
append1.go
args.go
armimm.go
asmhdr.go
assign.go
assign1.go
atomicload.go
bigalg.go
bigmap.go
blank.go
blank1.go
bom.go
bombad.go
bounds.go
cannotassign.go
chancap.go
chanlinear.go
char_lit.go
char_lit1.go
checkbce.go
clearfat.go
closedchan.go
closure.go
closure1.go
closure2.go
closure3.go
closure4.go
closure5.go
closure6.go
closure7.go
cmp.go
cmp6.go
cmplx.go
cmplxdivide.c
cmplxdivide.go
cmplxdivide1.go
complit.go
complit1.go
compos.go
const.go
const1.go
const2.go
const3.go
const4.go
const5.go
const6.go
const7.go
const8.go
convT2X.go
convert.go
convert1.go
convert2.go
convert3.go
convert4.go
convlit.go
convlit1.go
copy.go
copy1.go
crlf.go
ddd.go
ddd1.go
ddd2.go
decl.go
declbad.go
defer.go
defererrcheck.go
deferfin.go
defernil.go
deferprint.go
deferprint.out
devirt.go
directive.go
directive2.go
divide.go
divmod.go
embedfunc.go
embedvers.go
empty.go
env.go
eof.go
eof1.go
escape.go
escape2.go
escape2n.go
escape3.go
escape4.go
escape5.go
escape_array.go
escape_calls.go
escape_closure.go
escape_field.go
escape_goto.go
escape_hash_maphash.go
escape_iface.go
escape_indir.go
escape_level.go
escape_map.go
escape_param.go
escape_runtime_atomic.go
escape_selfassign.go
escape_slice.go
escape_struct_param1.go
escape_struct_param2.go
escape_struct_return.go
escape_sync_atomic.go
escape_unsafe.go
fibo.go
finprofiled.go
float_lit.go
float_lit2.go
float_lit3.go
floatcmp.go
for.go
func.go
func1.go
func2.go
func3.go
func4.go
func5.go
func6.go
func7.go
func8.go
funcdup.go
funcdup2.go
fuse.go
gc.go
gc1.go
gc2.go
gcgort.go
gcstring.go
goprint.go
goprint.out
goto.go
heapsampling.go
helloworld.go
helloworld.out
if.go
import.go
import1.go
import2.go
import4.go
import5.go
import6.go
index.go
index0.go
index1.go
index2.go
indirect.go
indirect1.go
init.go
init1.go
initcomma.go
initempty.go
initexp.go
initialize.go
initializerr.go
initloop.go
inline.go
inline_big.go
inline_caller.go
inline_callers.go
inline_endian.go
inline_literal.go
inline_math_bits_rotate.go
inline_sync.go
inline_variadic.go
int_lit.go
intcvt.go
intrinsic.go
intrinsic_atomic.go
iota.go
label.go
label1.go
linkmain.go
linkmain_run.go
linkname.go
linkname3.go
linkobj.go
linkx.go
linkx_run.go
literal.go
literal2.go
live.go
live1.go
live2.go
live_regabi.go
live_uintptrkeepalive.go
loopbce.go
mainsig.go
makechan.go
makemap.go
makenew.go
makeslice.go
mallocfin.go
map.go
map1.go
mapclear.go
maplinear.go
maymorestack.go
mergemul.go
method.go
method1.go
method2.go
method3.go
method4.go
method5.go
method6.go
method7.go
named.go
named1.go
nil.go
nilcheck.go
nilptr.go
nilptr2.go
nilptr3.go
nilptr4.go
nilptr5.go
nilptr5_aix.go
nilptr5_wasm.go
nilptr_aix.go
nosplit.go
notinheap.go
notinheap2.go
notinheap3.go
nowritebarrier.go
nul1.go
opt_branchlikely.go
parentype.go
peano.go
phiopt.go
print.go
print.out
printbig.go
printbig.out
prove.go
prove_constant_folding.go
range.go
recover.go
recover1.go
recover2.go
recover3.go
recover4.go
recover5.go
reflectmethod1.go
reflectmethod2.go
reflectmethod3.go
reflectmethod4.go
reflectmethod5.go
reflectmethod6.go
reflectmethod7.go
reflectmethod8.go
rename.go
rename1.go
reorder.go
reorder2.go
retjmp.go
return.go
rotate.go
rotate0.go
rotate1.go
rotate2.go
rotate3.go
run.go
rune.go
runtime.go
shift1.go
shift2.go
shift3.go
sieve.go
sigchld.go
sigchld.out
simassign.go
sinit.go
sinit_run.go
sizeof.go
slice3.go
slice3err.go
slicecap.go
sliceopt.go
solitaire.go
stack.go
stackobj.go
stackobj2.go
stackobj3.go
strcopy.go
strength.go
string_lit.go
stringrange.go
struct0.go
switch.go
switch2.go
switch3.go
switch4.go
switch5.go
switch6.go
switch7.go
tinyfin.go
torture.go
turing.go
typecheck.go
typecheckloop.go
typeswitch.go
typeswitch1.go
typeswitch2.go
typeswitch2b.go
typeswitch3.go
uintptrescapes.go
uintptrescapes2.go
uintptrescapes3.go
uintptrkeepalive.go
undef.go
unsafebuiltins.go
used.go
utf.go
varerr.go
varinit.go
winbatch.go
writebarrier.go
zerodivide.go
.gitattributes
.gitignore
CONTRIBUTING.md
LICENSE
PATENTS
README.md
SECURITY.md
VERSION
codereview.cfg
go/test/codegen/arithmetic.go
Archana R 4056934e48 test/codegen: updated arithmetic tests to verify on ppc64,ppc64le 
Updated multiple tests in test/codegen/arithmetic.go to verify
on ppc64/ppc64le as well

Change-Id: I79ca9f87017ea31147a4ba16f5d42ba0fcae64e1
Reviewed-on: https://go-review.googlesource.com/c/go/+/358546
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
2021-11-01 13:12:37 +00:00

614 lines
15 KiB
Go
Raw Permalink Blame History

// 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
// This file contains codegen tests related to arithmetic
// simplifications and optimizations on integer types.
// For codegen tests on float types, see floats.go.
// ----------------- //
// Subtraction //
// ----------------- //
var ef int
func SubMem(arr []int, b, c, d int) int {
// 386:`SUBL\s[A-Z]+,\s8\([A-Z]+\)`
// amd64:`SUBQ\s[A-Z]+,\s16\([A-Z]+\)`
arr[2] -= b
// 386:`SUBL\s[A-Z]+,\s12\([A-Z]+\)`
// amd64:`SUBQ\s[A-Z]+,\s24\([A-Z]+\)`
arr[3] -= b
// 386:`DECL\s16\([A-Z]+\)`
arr[4]--
// 386:`ADDL\s[$]-20,\s20\([A-Z]+\)`
arr[5] -= 20
// 386:`SUBL\s\([A-Z]+\)\([A-Z]+\*4\),\s[A-Z]+`
ef -= arr[b]
// 386:`SUBL\s[A-Z]+,\s\([A-Z]+\)\([A-Z]+\*4\)`
arr[c] -= b
// 386:`ADDL\s[$]-15,\s\([A-Z]+\)\([A-Z]+\*4\)`
arr[d] -= 15
// 386:`DECL\s\([A-Z]+\)\([A-Z]+\*4\)`
arr[b]--
// amd64:`DECQ\s64\([A-Z]+\)`
arr[8]--
// 386:"SUBL\t4"
// amd64:"SUBQ\t8"
return arr[0] - arr[1]
}
func SubFromConst(a int) int {
// ppc64le: `SUBC\tR[0-9]+,\s[$]40,\sR`
// ppc64: `SUBC\tR[0-9]+,\s[$]40,\sR`
b := 40 - a
return b
}
func SubFromConstNeg(a int) int {
// ppc64le: `ADD\t[$]40,\sR[0-9]+,\sR`
// ppc64: `ADD\t[$]40,\sR[0-9]+,\sR`
c := 40 - (-a)
return c
}
func SubSubFromConst(a int) int {
// ppc64le: `ADD\t[$]20,\sR[0-9]+,\sR`
// ppc64: `ADD\t[$]20,\sR[0-9]+,\sR`
c := 40 - (20 - a)
return c
}
func AddSubFromConst(a int) int {
// ppc64le: `SUBC\tR[0-9]+,\s[$]60,\sR`
// ppc64: `SUBC\tR[0-9]+,\s[$]60,\sR`
c := 40 + (20 - a)
return c
}
func NegSubFromConst(a int) int {
// ppc64le: `ADD\t[$]-20,\sR[0-9]+,\sR`
// ppc64: `ADD\t[$]-20,\sR[0-9]+,\sR`
c := -(20 - a)
return c
}
func NegAddFromConstNeg(a int) int {
// ppc64le: `SUBC\tR[0-9]+,\s[$]40,\sR`
// ppc64: `SUBC\tR[0-9]+,\s[$]40,\sR`
c := -(-40 + a)
return c
}
func SubSubNegSimplify(a, b int) int {
// amd64:"NEGQ"
// ppc64:"NEG"
// ppc64le:"NEG"
r := (a - b) - a
return r
}
func SubAddSimplify(a, b int) int {
// amd64:-"SUBQ",-"ADDQ"
// ppc64:-"SUB",-"ADD"
// ppc64le:-"SUB",-"ADD"
r := a + (b - a)
return r
}
func SubAddNegSimplify(a, b int) int {
// amd64:"NEGQ",-"ADDQ",-"SUBQ"
// ppc64:"NEG",-"ADD",-"SUB"
// ppc64le:"NEG",-"ADD",-"SUB"
r := a - (b + a)
return r
}
func AddAddSubSimplify(a, b, c int) int {
// amd64:-"SUBQ"
// ppc64:-"SUB"
// ppc64le:-"SUB"
r := a + (b + (c - a))
return r
}
// -------------------- //
// Multiplication //
// -------------------- //
func Pow2Muls(n1, n2 int) (int, int) {
// amd64:"SHLQ\t[$]5",-"IMULQ"
// 386:"SHLL\t[$]5",-"IMULL"
// arm:"SLL\t[$]5",-"MUL"
// arm64:"LSL\t[$]5",-"MUL"
// ppc64:"SLD\t[$]5",-"MUL"
// ppc64le:"SLD\t[$]5",-"MUL"
a := n1 * 32
// amd64:"SHLQ\t[$]6",-"IMULQ"
// 386:"SHLL\t[$]6",-"IMULL"
// arm:"SLL\t[$]6",-"MUL"
// arm64:`NEG\sR[0-9]+<<6,\sR[0-9]+`,-`LSL`,-`MUL`
// ppc64:"SLD\t[$]6","NEG\\sR[0-9]+,\\sR[0-9]+",-"MUL"
// ppc64le:"SLD\t[$]6","NEG\\sR[0-9]+,\\sR[0-9]+",-"MUL"
b := -64 * n2
return a, b
}
func Mul_96(n int) int {
// amd64:`SHLQ\t[$]5`,`LEAQ\t\(.*\)\(.*\*2\),`,-`IMULQ`
// 386:`SHLL\t[$]5`,`LEAL\t\(.*\)\(.*\*2\),`,-`IMULL`
// arm64:`LSL\t[$]5`,`ADD\sR[0-9]+<<1,\sR[0-9]+`,-`MUL`
// arm:`SLL\t[$]5`,`ADD\sR[0-9]+<<1,\sR[0-9]+`,-`MUL`
// s390x:`SLD\t[$]5`,`SLD\t[$]6`,-`MULLD`
return n * 96
}
func Mul_n120(n int) int {
// s390x:`SLD\t[$]3`,`SLD\t[$]7`,-`MULLD`
return n * -120
}
func MulMemSrc(a []uint32, b []float32) {
// 386:`IMULL\s4\([A-Z]+\),\s[A-Z]+`
a[0] *= a[1]
// 386/sse2:`MULSS\s4\([A-Z]+\),\sX[0-9]+`
// amd64:`MULSS\s4\([A-Z]+\),\sX[0-9]+`
b[0] *= b[1]
}
// Multiplications merging tests
func MergeMuls1(n int) int {
// amd64:"IMUL3Q\t[$]46"
// 386:"IMUL3L\t[$]46"
// ppc64le:"MULLD\t[$]46"
// ppc64:"MULLD\t[$]46"
return 15*n + 31*n // 46n
}
func MergeMuls2(n int) int {
// amd64:"IMUL3Q\t[$]23","(ADDQ\t[$]29)|(LEAQ\t29)"
// 386:"IMUL3L\t[$]23","ADDL\t[$]29"
// ppc64le/power9:"MADDLD",-"MULLD\t[$]23",-"ADD\t[$]29"
// ppc64le/power8:"MULLD\t[$]23","ADD\t[$]29"
return 5*n + 7*(n+1) + 11*(n+2) // 23n + 29
}
func MergeMuls3(a, n int) int {
// amd64:"ADDQ\t[$]19",-"IMULQ\t[$]19"
// 386:"ADDL\t[$]19",-"IMULL\t[$]19"
// ppc64:"ADD\t[$]19",-"MULLD\t[$]19"
// ppc64le:"ADD\t[$]19",-"MULLD\t[$]19"
return a*n + 19*n // (a+19)n
}
func MergeMuls4(n int) int {
// amd64:"IMUL3Q\t[$]14"
// 386:"IMUL3L\t[$]14"
// ppc64:"MULLD\t[$]14"
// ppc64le:"MULLD\t[$]14"
return 23*n - 9*n // 14n
}
func MergeMuls5(a, n int) int {
// amd64:"ADDQ\t[$]-19",-"IMULQ\t[$]19"
// 386:"ADDL\t[$]-19",-"IMULL\t[$]19"
// ppc64:"ADD\t[$]-19",-"MULLD\t[$]19"
// ppc64le:"ADD\t[$]-19",-"MULLD\t[$]19"
return a*n - 19*n // (a-19)n
}
// -------------- //
// Division //
// -------------- //
func DivMemSrc(a []float64) {
// 386/sse2:`DIVSD\s8\([A-Z]+\),\sX[0-9]+`
// amd64:`DIVSD\s8\([A-Z]+\),\sX[0-9]+`
a[0] /= a[1]
}
func Pow2Divs(n1 uint, n2 int) (uint, int) {
// 386:"SHRL\t[$]5",-"DIVL"
// amd64:"SHRQ\t[$]5",-"DIVQ"
// arm:"SRL\t[$]5",-".*udiv"
// arm64:"LSR\t[$]5",-"UDIV"
// ppc64:"SRD"
// ppc64le:"SRD"
a := n1 / 32 // unsigned
// amd64:"SARQ\t[$]6",-"IDIVQ"
// 386:"SARL\t[$]6",-"IDIVL"
// arm:"SRA\t[$]6",-".*udiv"
// arm64:"ASR\t[$]6",-"SDIV"
// ppc64:"SRAD"
// ppc64le:"SRAD"
b := n2 / 64 // signed
return a, b
}
// Check that constant divisions get turned into MULs
func ConstDivs(n1 uint, n2 int) (uint, int) {
// amd64:"MOVQ\t[$]-1085102592571150095","MULQ",-"DIVQ"
// 386:"MOVL\t[$]-252645135","MULL",-"DIVL"
// arm64:`MOVD`,`UMULH`,-`DIV`
// arm:`MOVW`,`MUL`,-`.*udiv`
a := n1 / 17 // unsigned
// amd64:"MOVQ\t[$]-1085102592571150095","IMULQ",-"IDIVQ"
// 386:"MOVL\t[$]-252645135","IMULL",-"IDIVL"
// arm64:`SMULH`,-`DIV`
// arm:`MOVW`,`MUL`,-`.*udiv`
b := n2 / 17 // signed
return a, b
}
func FloatDivs(a []float32) float32 {
// amd64:`DIVSS\s8\([A-Z]+\),\sX[0-9]+`
// 386/sse2:`DIVSS\s8\([A-Z]+\),\sX[0-9]+`
return a[1] / a[2]
}
func Pow2Mods(n1 uint, n2 int) (uint, int) {
// 386:"ANDL\t[$]31",-"DIVL"
// amd64:"ANDL\t[$]31",-"DIVQ"
// arm:"AND\t[$]31",-".*udiv"
// arm64:"AND\t[$]31",-"UDIV"
// ppc64:"ANDCC\t[$]31"
// ppc64le:"ANDCC\t[$]31"
a := n1 % 32 // unsigned
// 386:"SHRL",-"IDIVL"
// amd64:"SHRQ",-"IDIVQ"
// arm:"SRA",-".*udiv"
// arm64:"ASR",-"REM"
// ppc64:"SRAD"
// ppc64le:"SRAD"
b := n2 % 64 // signed
return a, b
}
// Check that signed divisibility checks get converted to AND on low bits
func Pow2DivisibleSigned(n1, n2 int) (bool, bool) {
// 386:"TESTL\t[$]63",-"DIVL",-"SHRL"
// amd64:"TESTQ\t[$]63",-"DIVQ",-"SHRQ"
// arm:"AND\t[$]63",-".*udiv",-"SRA"
// arm64:"AND\t[$]63",-"UDIV",-"ASR"
// ppc64:"ANDCC\t[$]63",-"SRAD"
// ppc64le:"ANDCC\t[$]63",-"SRAD"
a := n1%64 == 0 // signed divisible
// 386:"TESTL\t[$]63",-"DIVL",-"SHRL"
// amd64:"TESTQ\t[$]63",-"DIVQ",-"SHRQ"
// arm:"AND\t[$]63",-".*udiv",-"SRA"
// arm64:"AND\t[$]63",-"UDIV",-"ASR"
// ppc64:"ANDCC\t[$]63",-"SRAD"
// ppc64le:"ANDCC\t[$]63",-"SRAD"
b := n2%64 != 0 // signed indivisible
return a, b
}
// Check that constant modulo divs get turned into MULs
func ConstMods(n1 uint, n2 int) (uint, int) {
// amd64:"MOVQ\t[$]-1085102592571150095","MULQ",-"DIVQ"
// 386:"MOVL\t[$]-252645135","MULL",-"DIVL"
// arm64:`MOVD`,`UMULH`,-`DIV`
// arm:`MOVW`,`MUL`,-`.*udiv`
a := n1 % 17 // unsigned
// amd64:"MOVQ\t[$]-1085102592571150095","IMULQ",-"IDIVQ"
// 386:"MOVL\t[$]-252645135","IMULL",-"IDIVL"
// arm64:`SMULH`,-`DIV`
// arm:`MOVW`,`MUL`,-`.*udiv`
b := n2 % 17 // signed
return a, b
}
// Check that divisibility checks x%c==0 are converted to MULs and rotates
func Divisible(n1 uint, n2 int) (bool, bool, bool, bool) {
// amd64:"MOVQ\t[$]-6148914691236517205","IMULQ","ROLQ\t[$]63",-"DIVQ"
// 386:"IMUL3L\t[$]-1431655765","ROLL\t[$]31",-"DIVQ"
// arm64:"MOVD\t[$]-6148914691236517205","MUL","ROR",-"DIV"
// arm:"MUL","CMP\t[$]715827882",-".*udiv"
// ppc64:"MULLD","ROTL\t[$]63"
// ppc64le:"MULLD","ROTL\t[$]63"
evenU := n1%6 == 0
// amd64:"MOVQ\t[$]-8737931403336103397","IMULQ",-"ROLQ",-"DIVQ"
// 386:"IMUL3L\t[$]678152731",-"ROLL",-"DIVQ"
// arm64:"MOVD\t[$]-8737931403336103397","MUL",-"ROR",-"DIV"
// arm:"MUL","CMP\t[$]226050910",-".*udiv"
// ppc64:"MULLD",-"ROTL"
// ppc64le:"MULLD",-"ROTL"
oddU := n1%19 == 0
// amd64:"IMULQ","ADD","ROLQ\t[$]63",-"DIVQ"
// 386:"IMUL3L\t[$]-1431655765","ADDL\t[$]715827882","ROLL\t[$]31",-"DIVQ"
// arm64:"MUL","ADD\t[$]3074457345618258602","ROR",-"DIV"
// arm:"MUL","ADD\t[$]715827882",-".*udiv"
// ppc64/power8:"MULLD","ADD","ROTL\t[$]63"
// ppc64le/power8:"MULLD","ADD","ROTL\t[$]63"
// ppc64/power9:"MADDLD","ROTL\t[$]63"
// ppc64le/power9:"MADDLD","ROTL\t[$]63"
evenS := n2%6 == 0
// amd64:"IMULQ","ADD",-"ROLQ",-"DIVQ"
// 386:"IMUL3L\t[$]678152731","ADDL\t[$]113025455",-"ROLL",-"DIVQ"
// arm64:"MUL","ADD\t[$]485440633518672410",-"ROR",-"DIV"
// arm:"MUL","ADD\t[$]113025455",-".*udiv"
// ppc64/power8:"MULLD","ADD",-"ROTL"
// ppc64/power9:"MADDLD",-"ROTL"
// ppc64le/power8:"MULLD","ADD",-"ROTL"
// ppc64le/power9:"MADDLD",-"ROTL"
oddS := n2%19 == 0
return evenU, oddU, evenS, oddS
}
// Check that fix-up code is not generated for divisions where it has been proven that
// that the divisor is not -1 or that the dividend is > MinIntNN.
func NoFix64A(divr int64) (int64, int64) {
var d int64 = 42
var e int64 = 84
if divr > 5 {
d /= divr // amd64:-"JMP"
e %= divr // amd64:-"JMP"
// The following statement is to avoid conflict between the above check
// and the normal JMP generated at the end of the block.
d += e
}
return d, e
}
func NoFix64B(divd int64) (int64, int64) {
var d int64
var e int64
var divr int64 = -1
if divd > -9223372036854775808 {
d = divd / divr // amd64:-"JMP"
e = divd % divr // amd64:-"JMP"
d += e
}
return d, e
}
func NoFix32A(divr int32) (int32, int32) {
var d int32 = 42
var e int32 = 84
if divr > 5 {
// amd64:-"JMP"
// 386:-"JMP"
d /= divr
// amd64:-"JMP"
// 386:-"JMP"
e %= divr
d += e
}
return d, e
}
func NoFix32B(divd int32) (int32, int32) {
var d int32
var e int32
var divr int32 = -1
if divd > -2147483648 {
// amd64:-"JMP"
// 386:-"JMP"
d = divd / divr
// amd64:-"JMP"
// 386:-"JMP"
e = divd % divr
d += e
}
return d, e
}
func NoFix16A(divr int16) (int16, int16) {
var d int16 = 42
var e int16 = 84
if divr > 5 {
// amd64:-"JMP"
// 386:-"JMP"
d /= divr
// amd64:-"JMP"
// 386:-"JMP"
e %= divr
d += e
}
return d, e
}
func NoFix16B(divd int16) (int16, int16) {
var d int16
var e int16
var divr int16 = -1
if divd > -32768 {
// amd64:-"JMP"
// 386:-"JMP"
d = divd / divr
// amd64:-"JMP"
// 386:-"JMP"
e = divd % divr
d += e
}
return d, e
}
// Check that len() and cap() calls divided by powers of two are
// optimized into shifts and ands
func LenDiv1(a []int) int {
// 386:"SHRL\t[$]10"
// amd64:"SHRQ\t[$]10"
// arm64:"LSR\t[$]10",-"SDIV"
// arm:"SRL\t[$]10",-".*udiv"
// ppc64:"SRD"\t[$]10"
// ppc64le:"SRD"\t[$]10"
return len(a) / 1024
}
func LenDiv2(s string) int {
// 386:"SHRL\t[$]11"
// amd64:"SHRQ\t[$]11"
// arm64:"LSR\t[$]11",-"SDIV"
// arm:"SRL\t[$]11",-".*udiv"
// ppc64:"SRD\t[$]11"
// ppc64le:"SRD\t[$]11"
return len(s) / (4097 >> 1)
}
func LenMod1(a []int) int {
// 386:"ANDL\t[$]1023"
// amd64:"ANDL\t[$]1023"
// arm64:"AND\t[$]1023",-"SDIV"
// arm/6:"AND",-".*udiv"
// arm/7:"BFC",-".*udiv",-"AND"
// ppc64:"ANDCC\t[$]1023"
// ppc64le:"ANDCC\t[$]1023"
return len(a) % 1024
}
func LenMod2(s string) int {
// 386:"ANDL\t[$]2047"
// amd64:"ANDL\t[$]2047"
// arm64:"AND\t[$]2047",-"SDIV"
// arm/6:"AND",-".*udiv"
// arm/7:"BFC",-".*udiv",-"AND"
// ppc64:"ANDCC\t[$]2047"
// ppc64le:"ANDCC\t[$]2047"
return len(s) % (4097 >> 1)
}
func CapDiv(a []int) int {
// 386:"SHRL\t[$]12"
// amd64:"SHRQ\t[$]12"
// arm64:"LSR\t[$]12",-"SDIV"
// arm:"SRL\t[$]12",-".*udiv"
// ppc64:"SRD\t[$]12"
// ppc64le:"SRD\t[$]12"
return cap(a) / ((1 << 11) + 2048)
}
func CapMod(a []int) int {
// 386:"ANDL\t[$]4095"
// amd64:"ANDL\t[$]4095"
// arm64:"AND\t[$]4095",-"SDIV"
// arm/6:"AND",-".*udiv"
// arm/7:"BFC",-".*udiv",-"AND"
// ppc64:"ANDCC\t[$]4095"
// ppc64le:"ANDCC\t[$]4095"
return cap(a) % ((1 << 11) + 2048)
}
func AddMul(x int) int {
// amd64:"LEAQ\t1"
return 2*x + 1
}
func MULA(a, b, c uint32) (uint32, uint32, uint32) {
// arm:`MULA`,-`MUL\s`
// arm64:`MADDW`,-`MULW`
r0 := a*b + c
// arm:`MULA`,-`MUL\s`
// arm64:`MADDW`,-`MULW`
r1 := c*79 + a
// arm:`ADD`,-`MULA`,-`MUL\s`
// arm64:`ADD`,-`MADD`,-`MULW`
// ppc64:`ADD`,-`MULLD`
// ppc64le:`ADD`,-`MULLD`
r2 := b*64 + c
return r0, r1, r2
}
func MULS(a, b, c uint32) (uint32, uint32, uint32) {
// arm/7:`MULS`,-`MUL\s`
// arm/6:`SUB`,`MUL\s`,-`MULS`
// arm64:`MSUBW`,-`MULW`
r0 := c - a*b
// arm/7:`MULS`,-`MUL\s`
// arm/6:`SUB`,`MUL\s`,-`MULS`
// arm64:`MSUBW`,-`MULW`
r1 := a - c*79
// arm/7:`SUB`,-`MULS`,-`MUL\s`
// arm64:`SUB`,-`MSUBW`,-`MULW`
// ppc64:`SUB`,-`MULLD`
// ppc64le:`SUB`,-`MULLD`
r2 := c - b*64
return r0, r1, r2
}
func addSpecial(a, b, c uint32) (uint32, uint32, uint32) {
// amd64:`INCL`
a++
// amd64:`DECL`
b--
// amd64:`SUBL.*-128`
c += 128
return a, b, c
}
// Divide -> shift rules usually require fixup for negative inputs.
// If the input is non-negative, make sure the fixup is eliminated.
func divInt(v int64) int64 {
if v < 0 {
return 0
}
// amd64:-`.*SARQ.*63,`, -".*SHRQ", ".*SARQ.*[$]9,"
return v / 512
}
// The reassociate rules "x - (z + C) -> (x - z) - C" and
// "(z + C) -x -> C + (z - x)" can optimize the following cases.
func constantFold1(i0, j0, i1, j1, i2, j2, i3, j3 int) (int, int, int, int) {
// arm64:"SUB","ADD\t[$]2"
// ppc64:"SUB","ADD\t[$]2"
// ppc64le:"SUB","ADD\t[$]2"
r0 := (i0 + 3) - (j0 + 1)
// arm64:"SUB","SUB\t[$]4"
// ppc64:"SUB","ADD\t[$]-4"
// ppc64le:"SUB","ADD\t[$]-4"
r1 := (i1 - 3) - (j1 + 1)
// arm64:"SUB","ADD\t[$]4"
// ppc64:"SUB","ADD\t[$]4"
// ppc64le:"SUB","ADD\t[$]4"
r2 := (i2 + 3) - (j2 - 1)
// arm64:"SUB","SUB\t[$]2"
// ppc64:"SUB","ADD\t[$]-2"
// ppc64le:"SUB","ADD\t[$]-2"
r3 := (i3 - 3) - (j3 - 1)
return r0, r1, r2, r3
}
// The reassociate rules "x - (z + C) -> (x - z) - C" and
// "(C - z) - x -> C - (z + x)" can optimize the following cases.
func constantFold2(i0, j0, i1, j1 int) (int, int) {
// arm64:"ADD","MOVD\t[$]2","SUB"
// ppc64le: `SUBC\tR[0-9]+,\s[$]2,\sR`
// ppc64: `SUBC\tR[0-9]+,\s[$]2,\sR`
r0 := (3 - i0) - (j0 + 1)
// arm64:"ADD","MOVD\t[$]4","SUB"
// ppc64le: `SUBC\tR[0-9]+,\s[$]4,\sR`
// ppc64: `SUBC\tR[0-9]+,\s[$]4,\sR`
r1 := (3 - i1) - (j1 - 1)
return r0, r1
}
func constantFold3(i, j int) int {
// arm64: "MOVD\t[$]30","MUL",-"ADD",-"LSL"
// ppc64:"MULLD\t[$]30","MULLD"
// ppc64le:"MULLD\t[$]30","MULLD"
r := (5 * i) * (6 * j)
return r
}
View Git Blame Copy Permalink