[go] crypto/md5: performance optimization
Gopher Robot (Gerrit)
Congratulations on opening your first change. Thank you for your contribution!
Next steps:
A maintainer will review your change and provide feedback. See
https://go.dev/doc/contribute#review for more info and tips to get your
patch through code review.
Most changes in the Go project go through a few rounds of revision. This can be
surprising to people new to the project. The careful, iterative review process
is our way of helping mentor contributors and ensuring that their contributions
have a lasting impact.
During May-July and Nov-Jan the Go project is in a code freeze, during which
little code gets reviewed or merged. If a reviewer responds with a comment like
R=go1.11 or adds a tag like "wait-release", it means that this CL will be
reviewed as part of the next development cycle. See https://go.dev/s/release
for more details.
To view, visit change 433475. To unsubscribe, or for help writing mail filters, visit settings.
Priya N Vaidya (Gerrit)
Priya N Vaidya has uploaded this change for review.
crypto/md5: performance optimization
optimized g function
removed lea instruction
Change-Id: I16d032e199ab7313d9a11b84656dfb30e34b2131
---
M src/crypto/md5/md5block_amd64.s
1 file changed, 25 insertions(+), 8 deletions(-)
diff --git a/src/crypto/md5/md5block_amd64.s b/src/crypto/md5/md5block_amd64.s
index 7c7d92d..ec935bb 100644
--- a/src/crypto/md5/md5block_amd64.s
+++ b/src/crypto/md5/md5block_amd64.s
@@ -40,10 +40,11 @@
#define ROUND1(a, b, c, d, index, const, shift) \
XORL c, R9; \
- LEAL const(a)(R8*1), a; \
+ ADDL $const, a; \
ANDL b, R9; \
+ ADDL R8, a; \
XORL d, R9; \
- MOVL (index*4)(SI), R8; \
+ MOVL (index*4)(SI), R8; \
ADDL R9, a; \
ROLL $shift, a; \
MOVL c, R9; \
@@ -72,13 +73,15 @@
#define ROUND2(a, b, c, d, index, const, shift) \
NOTL R9; \
- LEAL const(a)(R8*1),a; \
+ ADDL $const, a; \
+ ANDL c, R9; \
+ ADDL R8, a; \
+ MOVL d, R10; \
+ ADDL R9, a ; \
ANDL b, R10; \
- ANDL c, R9; \
+ ADDL R10, a; \
MOVL (index*4)(SI),R8; \
- ORL R9, R10; \
MOVL c, R9; \
- ADDL R10, a; \
MOVL c, R10; \
ROLL $shift, a; \
ADDL b, a
@@ -104,7 +107,8 @@
MOVL CX, R9
#define ROUND3(a, b, c, d, index, const, shift) \
- LEAL const(a)(R8*1),a; \
+ ADDL $const,a; \
+ ADDL R8, a; \
MOVL (index*4)(SI),R8; \
XORL d, R9; \
XORL b, R9; \
@@ -135,7 +139,8 @@
XORL DX, R9
#define ROUND4(a, b, c, d, index, const, shift) \
- LEAL const(a)(R8*1),a; \
+ ADDL $const, a; \
+ ADDL R8, a; \
ORL b, R9; \
XORL c, R9; \
ADDL R9, a; \
To view, visit change 433475. To unsubscribe, or for help writing mail filters, visit settings.
Dan Kortschak (Gerrit)
Attention is currently required from: Filippo Valsorda, Priya N Vaidya.
2 comments:
Commit Message:
Can you provide benchmarking support for this change?
File src/crypto/md5/md5block_amd64.s:
Patch Set #1, Line 43:
ADDL $const, a; \These additions indent with spaces rather than tab as is used on the existing code. Can you make the changes agree with what is here?
To view, visit change 433475. To unsubscribe, or for help writing mail filters, visit settings.
Priya N Vaidya (Gerrit)
Attention is currently required from: Dan Kortschak, Filippo Valsorda.
2 comments:
Commit Message:
Can you provide benchmarking support for this change?
@d...@kortschak.io - I have ran the standard go benchmarks within the md5 directory on aws instances, m6ix, m5ix etc. In order for me to publish data here i would need my company approval and will take some more time. But if you run the md5 benchmarks in go tree you can see the performance improvement
File src/crypto/md5/md5block_amd64.s:
Patch Set #1, Line 43:
ADDL $const, a; \
These additions indent with spaces rather than tab as is used on the existing code. […]
will do.
To view, visit change 433475. To unsubscribe, or for help writing mail filters, visit settings.
Dan Kortschak (Gerrit)
Attention is currently required from: Filippo Valsorda, Priya N Vaidya.
1 comment:
Commit Message:
@dan@kortschak. […]
I have run the benchmarks locally comparing this change with its parent. You could use that output for the commit message.
```
name old time/op new time/op delta
Hash8Bytes-8 122ns ± 1% 115ns ± 2% -6.39% (p=0.000 n=9+10)
Hash64-8 205ns ± 0% 190ns ± 1% -7.31% (p=0.000 n=8+10)
Hash128-8 298ns ± 1% 275ns ± 1% -7.82% (p=0.000 n=9+9)
Hash256-8 483ns ± 1% 450ns ± 2% -6.85% (p=0.000 n=8+10)
Hash512-8 855ns ± 1% 785ns ± 1% -8.16% (p=0.000 n=10+10)
Hash1K-8 1.61µs ± 2% 1.47µs ± 0% -8.64% (p=0.000 n=9+9)
Hash8K-8 12.1µs ± 1% 11.1µs ± 1% -8.58% (p=0.000 n=10+10)
Hash1M-8 1.53ms ± 2% 1.41ms ± 2% -8.17% (p=0.000 n=10+10)
Hash8M-8 12.2ms ± 0% 11.2ms ± 2% -7.68% (p=0.000 n=8+10)
Hash8BytesUnaligned-8 122ns ± 1% 114ns ± 1% -6.69% (p=0.000 n=9+9)
Hash1KUnaligned-8 1.61µs ± 1% 1.47µs ± 1% -8.72% (p=0.000 n=10+8)
Hash8KUnaligned-8 12.1µs ± 2% 13.1µs ±16% ~ (p=0.138 n=10+10)name old speed new speed delta
Hash8Bytes-8 65.4MB/s ± 1% 69.8MB/s ± 2% +6.83% (p=0.000 n=9+10)
Hash64-8 313MB/s ± 0% 337MB/s ± 1% +7.91% (p=0.000 n=8+10)
Hash128-8 429MB/s ± 1% 466MB/s ± 1% +8.49% (p=0.000 n=9+9)
Hash256-8 529MB/s ± 1% 569MB/s ± 2% +7.53% (p=0.000 n=9+10)
Hash512-8 599MB/s ± 1% 652MB/s ± 1% +8.88% (p=0.000 n=10+10)
Hash1K-8 637MB/s ± 2% 697MB/s ± 1% +9.43% (p=0.000 n=9+9)
Hash8K-8 677MB/s ± 1% 741MB/s ± 1% +9.39% (p=0.000 n=10+10)
Hash1M-8 684MB/s ± 2% 745MB/s ± 2% +8.90% (p=0.000 n=10+10)
Hash8M-8 689MB/s ± 0% 747MB/s ± 2% +8.32% (p=0.000 n=8+10)
Hash8BytesUnaligned-8 65.6MB/s ± 1% 70.3MB/s ± 1% +7.15% (p=0.000 n=9+9)
Hash1KUnaligned-8 635MB/s ± 1% 696MB/s ± 1% +9.54% (p=0.000 n=10+8)
Hash8KUnaligned-8 677MB/s ± 2% 633MB/s ±18% ~ (p=0.143 n=10+10)
```
To view, visit change 433475. To unsubscribe, or for help writing mail filters, visit settings.
Priya N Vaidya (Gerrit)
Attention is currently required from: Filippo Valsorda, Priya N Vaidya.
Priya N Vaidya uploaded patch set #2 to this change.
crypto/md5: performance optimization
optimized g function
removed lea instruction
@d...@kortschak.io have run the benchmarks locally comparing this change with its parent.
Here are the performance results.
Change-Id: I16d032e199ab7313d9a11b84656dfb30e34b2131
---
M src/crypto/md5/md5block_amd64.s
1 file changed, 56 insertions(+), 8 deletions(-)
To view, visit change 433475. To unsubscribe, or for help writing mail filters, visit settings.
Joel Sing (Gerrit)
Attention is currently required from: Filippo Valsorda, Priya N Vaidya.
1 comment:
Patchset:
It would appear that https://go-review.googlesource.com/c/go/+/283538 has a more extensive set of optimisations and out performs this change.
To view, visit change 433475. To unsubscribe, or for help writing mail filters, visit settings.