rlp, trie: faster trie node encoding (#24126) #606
Merged
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
huyngopt1994
force-pushed
the
path-base-implementing
branch
from
69b0c7f
to
e42b5b8
Compare
|
This PR is LGTM, Could u add description related to some minor refactors in rlp package and help to resolve conflicts before merging too ? |
minh-bq
force-pushed
the
faster-node-encoding
branch
from
58545c7
to
34727c3
Compare
commit ethereum/go-ethereum@65ed1a6. This change speeds up trie hashing and all other activities that require RLP encoding of trie nodes by approximately 20%. The speedup is achieved by avoiding reflection overhead during node encoding. The interface type trie.node now contains a method 'encode' that works with rlp.EncoderBuffer. Management of EncoderBuffers is left to calling code. trie.hasher, which is pooled to avoid allocations, now maintains an EncoderBuffer. This means memory resources related to trie node encoding are tied to the hasher pool. This also refactors some functions in rlp package. goos: linux goarch: amd64 cpu: 11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz │ old.txt │ new.txt │ │ sec/op │ sec/op vs base │ DeriveSha200/std_trie-8 725.1µ ± 31% 613.8µ ± 37% ~ (p=0.481 n=10) DeriveSha200/stack_trie-8 572.3µ ± 10% 493.1µ ± 13% -13.85% (p=0.005 n=10) geomean 644.2µ 550.1µ -14.61% │ old.txt │ new.txt │ │ B/op │ B/op vs base │ DeriveSha200/std_trie-8 287.4Ki ± 0% 283.0Ki ± 0% -1.53% (p=0.000 n=10) DeriveSha200/stack_trie-8 56.34Ki ± 0% 42.43Ki ± 0% -24.69% (p=0.000 n=10) geomean 127.2Ki 109.6Ki -13.88% │ old.txt │ new.txt │ │ allocs/op │ allocs/op vs base │ DeriveSha200/std_trie-8 2.931k ± 0% 2.917k ± 0% -0.46% (p=0.000 n=10) DeriveSha200/stack_trie-8 1.462k ± 0% 1.246k ± 0% -14.77% (p=0.000 n=10) geomean 2.070k 1.907k -7.90% │ old.txt │ new.txt │ │ sec/op │ sec/op vs base │ Prove-8 664.0µ ± 21% 450.2µ ± 27% -32.20% (p=0.000 n=10) VerifyProof-8 8.643µ ± 18% 9.009µ ± 33% ~ (p=0.684 n=10) VerifyRangeProof10-8 99.18µ ± 25% 67.60µ ± 67% ~ (p=0.089 n=10) VerifyRangeProof100-8 496.3µ ± 20% 487.0µ ± 33% ~ (p=0.739 n=10) VerifyRangeProof1000-8 5.149m ± 32% 4.095m ± 49% ~ (p=0.971 n=10) VerifyRangeProof5000-8 19.79m ± 60% 19.16m ± 28% ~ (p=0.631 n=10) VerifyRangeNoProof10-8 499.0µ ± 15% 422.8µ ± 29% -15.25% (p=0.035 n=10) VerifyRangeNoProof500-8 1.747m ± 30% 1.417m ± 24% -18.91% (p=0.023 n=10) VerifyRangeNoProof1000-8 3.025m ± 29% 2.239m ± 33% -25.98% (p=0.009 n=10) geomean 750.9µ 622.6µ -17.09% │ old.txt │ new.txt │ │ sec/op │ sec/op vs base │ HashFixedSize/10-8 60.30µ ± 19% 44.84µ ± 17% -25.64% (p=0.000 n=10) HashFixedSize/100-8 205.9µ ± 32% 145.2µ ± 19% -29.48% (p=0.000 n=10) HashFixedSize/1K-8 1326.5µ ± 23% 939.2µ ± 25% -29.20% (p=0.002 n=10) HashFixedSize/10K-8 14.77m ± 25% 12.74m ± 19% ~ (p=0.075 n=10) HashFixedSize/100K-8 135.2m ± 19% 104.1m ± 18% -23.03% (p=0.003 n=10) geomean 2.011m 1.520m -24.43% │ old.txt │ new.txt │ │ B/op │ B/op vs base │ HashFixedSize/10-8 11.729Ki ± 0% 9.752Ki ± 0% -16.85% (p=0.000 n=10) HashFixedSize/100-8 58.56Ki ± 0% 49.23Ki ± 0% -15.93% (p=0.000 n=10) HashFixedSize/1K-8 578.1Ki ± 0% 481.5Ki ± 0% -16.72% (p=0.000 n=10) HashFixedSize/10K-8 6.019Mi ± 0% 4.985Mi ± 0% -17.18% (p=0.000 n=10) HashFixedSize/100K-8 59.53Mi ± 0% 49.29Mi ± 0% -17.20% (p=0.000 n=10) geomean 683.5Ki 568.8Ki -16.78% │ old.txt │ new.txt │ │ allocs/op │ allocs/op vs base │ HashFixedSize/10-8 149.0 ± 0% 142.0 ± 0% -4.70% (p=0.000 n=10) HashFixedSize/100-8 772.0 ± 0% 739.0 ± 0% -4.27% (p=0.000 n=10) HashFixedSize/1K-8 7.443k ± 0% 7.099k ± 0% -4.62% (p=0.000 n=10) HashFixedSize/10K-8 77.09k ± 0% 73.32k ± 0% -4.89% (p=0.000 n=10) HashFixedSize/100K-8 767.8k ± 0% 730.5k ± 0% -4.86% (p=0.000 n=10) geomean 8.729k 8.321k -4.67% Co-authored-by: Felix Lange <fjl@twurst.com>
minh-bq
force-pushed
the
faster-node-encoding
branch
from
34727c3
to
b67bcf7
Compare
huyngopt1994
approved these changes
Oct 29, 2024
huyngopt1994
merged commit 1615821
into
axieinfinity:path-base-implementing
1 check passed
huyngopt1994
pushed a commit
that referenced
this pull request
Nov 21, 2024
commit ethereum/go-ethereum@65ed1a6. This change speeds up trie hashing and all other activities that require RLP encoding of trie nodes by approximately 20%. The speedup is achieved by avoiding reflection overhead during node encoding. The interface type trie.node now contains a method 'encode' that works with rlp.EncoderBuffer. Management of EncoderBuffers is left to calling code. trie.hasher, which is pooled to avoid allocations, now maintains an EncoderBuffer. This means memory resources related to trie node encoding are tied to the hasher pool. This also refactors some functions in rlp package. goos: linux goarch: amd64 cpu: 11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz │ old.txt │ new.txt │ │ sec/op │ sec/op vs base │ DeriveSha200/std_trie-8 725.1µ ± 31% 613.8µ ± 37% ~ (p=0.481 n=10) DeriveSha200/stack_trie-8 572.3µ ± 10% 493.1µ ± 13% -13.85% (p=0.005 n=10) geomean 644.2µ 550.1µ -14.61% │ old.txt │ new.txt │ │ B/op │ B/op vs base │ DeriveSha200/std_trie-8 287.4Ki ± 0% 283.0Ki ± 0% -1.53% (p=0.000 n=10) DeriveSha200/stack_trie-8 56.34Ki ± 0% 42.43Ki ± 0% -24.69% (p=0.000 n=10) geomean 127.2Ki 109.6Ki -13.88% │ old.txt │ new.txt │ │ allocs/op │ allocs/op vs base │ DeriveSha200/std_trie-8 2.931k ± 0% 2.917k ± 0% -0.46% (p=0.000 n=10) DeriveSha200/stack_trie-8 1.462k ± 0% 1.246k ± 0% -14.77% (p=0.000 n=10) geomean 2.070k 1.907k -7.90% │ old.txt │ new.txt │ │ sec/op │ sec/op vs base │ Prove-8 664.0µ ± 21% 450.2µ ± 27% -32.20% (p=0.000 n=10) VerifyProof-8 8.643µ ± 18% 9.009µ ± 33% ~ (p=0.684 n=10) VerifyRangeProof10-8 99.18µ ± 25% 67.60µ ± 67% ~ (p=0.089 n=10) VerifyRangeProof100-8 496.3µ ± 20% 487.0µ ± 33% ~ (p=0.739 n=10) VerifyRangeProof1000-8 5.149m ± 32% 4.095m ± 49% ~ (p=0.971 n=10) VerifyRangeProof5000-8 19.79m ± 60% 19.16m ± 28% ~ (p=0.631 n=10) VerifyRangeNoProof10-8 499.0µ ± 15% 422.8µ ± 29% -15.25% (p=0.035 n=10) VerifyRangeNoProof500-8 1.747m ± 30% 1.417m ± 24% -18.91% (p=0.023 n=10) VerifyRangeNoProof1000-8 3.025m ± 29% 2.239m ± 33% -25.98% (p=0.009 n=10) geomean 750.9µ 622.6µ -17.09% │ old.txt │ new.txt │ │ sec/op │ sec/op vs base │ HashFixedSize/10-8 60.30µ ± 19% 44.84µ ± 17% -25.64% (p=0.000 n=10) HashFixedSize/100-8 205.9µ ± 32% 145.2µ ± 19% -29.48% (p=0.000 n=10) HashFixedSize/1K-8 1326.5µ ± 23% 939.2µ ± 25% -29.20% (p=0.002 n=10) HashFixedSize/10K-8 14.77m ± 25% 12.74m ± 19% ~ (p=0.075 n=10) HashFixedSize/100K-8 135.2m ± 19% 104.1m ± 18% -23.03% (p=0.003 n=10) geomean 2.011m 1.520m -24.43% │ old.txt │ new.txt │ │ B/op │ B/op vs base │ HashFixedSize/10-8 11.729Ki ± 0% 9.752Ki ± 0% -16.85% (p=0.000 n=10) HashFixedSize/100-8 58.56Ki ± 0% 49.23Ki ± 0% -15.93% (p=0.000 n=10) HashFixedSize/1K-8 578.1Ki ± 0% 481.5Ki ± 0% -16.72% (p=0.000 n=10) HashFixedSize/10K-8 6.019Mi ± 0% 4.985Mi ± 0% -17.18% (p=0.000 n=10) HashFixedSize/100K-8 59.53Mi ± 0% 49.29Mi ± 0% -17.20% (p=0.000 n=10) geomean 683.5Ki 568.8Ki -16.78% │ old.txt │ new.txt │ │ allocs/op │ allocs/op vs base │ HashFixedSize/10-8 149.0 ± 0% 142.0 ± 0% -4.70% (p=0.000 n=10) HashFixedSize/100-8 772.0 ± 0% 739.0 ± 0% -4.27% (p=0.000 n=10) HashFixedSize/1K-8 7.443k ± 0% 7.099k ± 0% -4.62% (p=0.000 n=10) HashFixedSize/10K-8 77.09k ± 0% 73.32k ± 0% -4.89% (p=0.000 n=10) HashFixedSize/100K-8 767.8k ± 0% 730.5k ± 0% -4.86% (p=0.000 n=10) geomean 8.729k 8.321k -4.67% Co-authored-by: Qian Bin <cola.tin.com@gmail.com> Co-authored-by: Felix Lange <fjl@twurst.com>
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.
commit ethereum/go-ethereum@65ed1a6.
This change speeds up trie hashing and all other activities that require RLP encoding of trie nodes by approximately 20%. The speedup is achieved by avoiding reflection overhead during node encoding.
The interface type trie.node now contains a method 'encode' that works with rlp.EncoderBuffer. Management of EncoderBuffers is left to calling code. trie.hasher, which is pooled to avoid allocations, now maintains an EncoderBuffer. This means memory resources related to trie node encoding are tied to the hasher pool.
This also refactors some functions in rlp package.