diff options
author | Keith Randall <khr@golang.org> | 2020-04-13 09:59:20 -0700 |
---|---|---|
committer | Keith Randall <khr@golang.org> | 2020-04-30 17:21:31 +0000 |
commit | 9ed0fb42e34ff64dd6447572d8bc9dbb05ddb552 (patch) | |
tree | e59e3546819c8202237f0b911b5adad7f48040cb /test | |
parent | 882ec701d2bc5df807637418d726f644ed2e0ef6 (diff) | |
download | go-git-9ed0fb42e34ff64dd6447572d8bc9dbb05ddb552.tar.gz |
cmd/compile: add indexed memory modification ops to amd64
name old time/op new time/op delta
Modify-16 404ns ± 1% 365ns ± 1% -9.73% (p=0.000 n=10+10)
ConstModify-16 407ns ± 0% 385ns ± 2% -5.56% (p=0.000 n=9+10)
Seems to generally help generated code.
Binary size change is in the noise.
Change-Id: I57891bfaf0f7dfc5d143bb9f7ebafc7079d2614f
Reviewed-on: https://go-review.googlesource.com/c/go/+/228098
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Diffstat (limited to 'test')
-rw-r--r-- | test/codegen/memops.go | 33 |
1 files changed, 32 insertions, 1 deletions
diff --git a/test/codegen/memops.go b/test/codegen/memops.go index 701735f875..cd35910c12 100644 --- a/test/codegen/memops.go +++ b/test/codegen/memops.go @@ -242,25 +242,56 @@ func idxLoadPlusOp64(x []int64, i int) int64 { return s } -func idxStorePlusOp(x []int32, i int, v int32) { +func idxStorePlusOp32(x []int32, i int, v int32) { // 386: `ADDL\t[A-Z]+, 4\([A-Z]+\)\([A-Z]+\*4\)` + // amd64: `ADDL\t[A-Z]+[0-9]*, 4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)` x[i+1] += v // 386: `SUBL\t[A-Z]+, 8\([A-Z]+\)\([A-Z]+\*4\)` + // amd64: `SUBL\t[A-Z]+[0-9]*, 8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)` x[i+2] -= v // 386: `ANDL\t[A-Z]+, 12\([A-Z]+\)\([A-Z]+\*4\)` + // amd64: `ANDL\t[A-Z]+[0-9]*, 12\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)` x[i+3] &= v // 386: `ORL\t[A-Z]+, 16\([A-Z]+\)\([A-Z]+\*4\)` + // amd64: `ORL\t[A-Z]+[0-9]*, 16\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)` x[i+4] |= v // 386: `XORL\t[A-Z]+, 20\([A-Z]+\)\([A-Z]+\*4\)` + // amd64: `XORL\t[A-Z]+[0-9]*, 20\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)` x[i+5] ^= v // 386: `ADDL\t[$]77, 24\([A-Z]+\)\([A-Z]+\*4\)` + // amd64: `ADDL\t[$]77, 24\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)` x[i+6] += 77 // 386: `ANDL\t[$]77, 28\([A-Z]+\)\([A-Z]+\*4\)` + // amd64: `ANDL\t[$]77, 28\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)` x[i+7] &= 77 // 386: `ORL\t[$]77, 32\([A-Z]+\)\([A-Z]+\*4\)` + // amd64: `ORL\t[$]77, 32\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)` x[i+8] |= 77 // 386: `XORL\t[$]77, 36\([A-Z]+\)\([A-Z]+\*4\)` + // amd64: `XORL\t[$]77, 36\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)` + x[i+9] ^= 77 +} + +func idxStorePlusOp64(x []int64, i int, v int64) { + // amd64: `ADDQ\t[A-Z]+[0-9]*, 8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)` + x[i+1] += v + // amd64: `SUBQ\t[A-Z]+[0-9]*, 16\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)` + x[i+2] -= v + // amd64: `ANDQ\t[A-Z]+[0-9]*, 24\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)` + x[i+3] &= v + // amd64: `ORQ\t[A-Z]+[0-9]*, 32\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)` + x[i+4] |= v + // amd64: `XORQ\t[A-Z]+[0-9]*, 40\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)` + x[i+5] ^= v + + // amd64: `ADDQ\t[$]77, 48\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)` + x[i+6] += 77 + // amd64: `ANDQ\t[$]77, 56\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)` + x[i+7] &= 77 + // amd64: `ORQ\t[$]77, 64\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)` + x[i+8] |= 77 + // amd64: `XORQ\t[$]77, 72\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)` x[i+9] ^= 77 } |