Skip to content

transaction pool optimizations#21328

Merged
holiman merged 3 commits into
ethereum:masterfrom
holiman:txfixes
Jul 14, 2020
Merged

transaction pool optimizations#21328
holiman merged 3 commits into
ethereum:masterfrom
holiman:txfixes

Conversation

@holiman
Copy link
Copy Markdown
Contributor

@holiman holiman commented Jul 13, 2020

This PR contains some of the commits from the (now reverted) #21232 , leaving out the erroneous conversions to uint64

[user@work core]$ benchstat pool.before pool.after 
name                         old time/op    new time/op    delta
PoolBatchInsert100-6           8.52ms ± 2%    8.50ms ± 2%     ~     (p=1.000 n=5+5)
PoolBatchInsert1000-6          99.2ms ± 1%   101.3ms ± 2%     ~     (p=0.190 n=4+5)
PoolBatchInsert10000-6          1.17s ± 9%     1.17s ±12%     ~     (p=1.000 n=5+5)
PoolBatchLocalInsert100-6       350ms ± 5%      31ms ± 3%  -91.20%  (p=0.008 n=5+5)
PoolBatchLocalInsert1000-6      5.15s ± 8%     0.40s ± 2%  -92.19%  (p=0.016 n=5+4)
PoolBatchLocalInsert10000-6     36.5s ± 2%      3.0s ± 4%  -91.84%  (p=0.036 n=3+5)

name                         old alloc/op   new alloc/op   delta
PoolBatchInsert100-6            293kB ± 0%     264kB ± 0%   -9.71%  (p=0.008 n=5+5)
PoolBatchInsert1000-6          2.63MB ± 0%    2.60MB ± 0%   -1.31%  (p=0.016 n=5+4)
PoolBatchInsert10000-6         22.7MB ± 0%    22.7MB ± 0%   -0.13%  (p=0.008 n=5+5)
PoolBatchLocalInsert100-6      14.7MB ± 0%     0.3MB ± 0%  -97.80%  (p=0.008 n=5+5)
PoolBatchLocalInsert1000-6      205MB ± 8%       3MB ± 0%  -98.71%  (p=0.016 n=5+4)
PoolBatchLocalInsert10000-6    1.46GB ± 0%    0.03GB ± 0%  -98.17%  (p=0.036 n=3+5)

name                         old allocs/op  new allocs/op  delta
PoolBatchInsert100-6            3.78k ± 0%     3.78k ± 0%     ~     (p=0.190 n=5+5)
PoolBatchInsert1000-6           37.8k ± 0%     37.8k ± 0%   -0.16%  (p=0.000 n=5+4)
PoolBatchInsert10000-6           339k ± 0%      339k ± 0%     ~     (p=0.548 n=5+5)
PoolBatchLocalInsert100-6       4.54k ± 0%     3.90k ± 0%  -14.15%  (p=0.016 n=5+4)
PoolBatchLocalInsert1000-6      46.2k ± 1%     38.2k ± 0%  -17.19%  (p=0.008 n=5+5)
PoolBatchLocalInsert10000-6      445k ± 0%      381k ± 0%  -14.36%  (p=0.036 n=3+5)

@holiman holiman changed the title Txfixes transaction pool optimizations Jul 13, 2020
Copy link
Copy Markdown
Member

@MariusVanDerWijden MariusVanDerWijden left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

Comment thread core/tx_pool.go
txs := list.Flatten() // Heavy but will be cached and is needed by the miner anyway
pool.pendingNonces.set(addr, txs[len(txs)-1].Nonce()+1)
highestPending := list.LastElement()
pool.pendingNonces.set(addr, highestPending.Nonce()+1)
Copy link
Copy Markdown
Contributor

@ucwong ucwong Jul 14, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why not use one line

pool.pendingNonces.set(addr, list.LastElement().Nonce()+1)

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, that would be fine too. As it is, it's more obvious for a reader that we're getting the highest nonce:d transaction, so it's a bit simpler to understand what's happening. I don't see any compelling reason to change it now

@holiman holiman added this to the 1.9.17 milestone Jul 14, 2020
@holiman holiman merged commit 6c9f040 into ethereum:master Jul 14, 2020
enriquefynn pushed a commit to enriquefynn/go-ethereum that referenced this pull request Mar 10, 2021
* core: added local tx pool test case

* core, crypto: various allocation savings regarding tx handling

* core/txlist, txpool: save a reheap operation, avoid some bigint allocs

Co-authored-by: Marius van der Wijden <m.vanderwijden@live.de>
@gzliudan gzliudan mentioned this pull request May 13, 2024
20 tasks
gzliudan pushed a commit to gzliudan/XDPoSChain that referenced this pull request May 13, 2024
* core: added local tx pool test case

* core, crypto: various allocation savings regarding tx handling

* core/txlist, txpool: save a reheap operation, avoid some bigint allocs

Co-authored-by: Marius van der Wijden <m.vanderwijden@live.de>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants