core/vm: use uint256.Bytes32 and builtin copy to make MSTORE faster #637
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
In commit f791124 ("core/vm: optimize the mstore opcode with loop unrolling"), we optimize the loop that copies each byte by manually unrolling the loop as it seems like Go cannot do that at this time. This makes the code quite ugly and might increase the number of unique instructions executed, creates more pressure to the instruction cache. This commit instead follows the go-ethereum commit e0a1fd5 ("core/vm: optimize Memory.Set32") by using uint256.Bytes32 and builtin copy. The uint256.Bytes32 is inlined and is compiled into fewer instructions 4x (load, bswap, store). The builtin copy can copy 32 bytes by just 2 load-store pairs using 128-bit (16-byte) xmm register. goos: linux goarch: amd64 pkg: github.com/ethereum/go-ethereum/core/vm cpu: 11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz │ old.txt │ new.txt │ │ sec/op │ sec/op vs base │ EvmInsertionSort-8 108.7m ± 6% 104.6m ± 26% ~ (p=0.631 n=10) EvmQuickSort-8 6.848m ± 14% 6.633m ± 3% ~ (p=0.089 n=10) EvmSignatureValidation-8 15.96µ ± 3% 15.48µ ± 3% ~ (p=0.052 n=10) EvmMulticallErcTransfer-8 6.503m ± 15% 6.562m ± 6% ~ (p=0.912 n=10) EvmRedBlackTree-8 302.5m ± 4% 305.0m ± 2% ~ (p=0.684 n=10) OpMstore-8 33.47n ± 10% 30.09n ± 6% -10.07% (p=0.000 n=10) geomean 959.9µ 930.0µ -3.11%
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In commit f791124 ("core/vm: optimize the mstore opcode with loop unrolling"), we optimize the loop that copies each byte by manually unrolling the loop as it seems like Go cannot do that at this time. This makes the code quite ugly and might increase the number of unique instructions executed, creates more pressure to the instruction cache.
This commit instead follows the go-ethereum commit e0a1fd5 ("core/vm: optimize Memory.Set32") by using uint256.Bytes32 and builtin copy. The uint256.Bytes32 is inlined and is compiled into fewer instructions 4x (load, bswap, store). The builtin copy can copy 32 bytes by just 2 load-store pairs using 128-bit (16-byte) xmm register.