Add pool based memory resource #25325
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
martinus
force-pushed
the
2022-06-very-not-scary-NodePoolResource
branch
2 times, most recently
from
6de57f7 to
d3c437f
Compare
|
The following sections might be updated with supplementary metadata relevant to reviewers and maintainers. ReviewsSee the guideline for information on the review process.
If your review is incorrectly listed, please react with 👎 to this comment and the bot will ignore it on the next update. ConflictsNo conflicts as of last run. |
This was referenced Jun 10, 2022
martinus
force-pushed
the
2022-06-very-not-scary-NodePoolResource
branch
3 times, most recently
from
850fe38 to
ba38016
Compare
Riahiamirreza
approved these changes
Jun 11, 2022
martinus
force-pushed
the
2022-06-very-not-scary-NodePoolResource
branch
5 times, most recently
from
6932541 to
b56c223
Compare
9 tasks
martinus
force-pushed
the
2022-06-very-not-scary-NodePoolResource
branch
4 times, most recently
from
98bb268 to
d196da1
Compare
martinus
changed the title
martinus
force-pushed
the
2022-06-very-not-scary-NodePoolResource
branch
from
d196da1 to
9205b60
Compare
|
rebased to 9205b60 with minor fixes in |
|
Posthumous utACK 9f947fc |
|
Wohoo 🎉 Thanks everyone for making this happen! |
sidhujag
pushed a commit
to syscoin/syscoin
that referenced
this pull request
Apr 21, 2023
1 task
This was referenced Nov 17, 2023
Fabcien
pushed a commit
to Bitcoin-ABC/bitcoin-abc
that referenced
this pull request
May 16, 2024
Summary: as many of the unit tests don't use this code This is a backport of [[bitcoin/bitcoin#26940 | core#26940]] (partial) and [[bitcoin/bitcoin#27425 | core#27425]] bitcoin/bitcoin@81f5ade This is a depency of [[bitcoin/bitcoin#25325 | core#25325]] Test Plan: `ninja check` Reviewers: #bitcoin_abc, Fabien Reviewed By: #bitcoin_abc, Fabien Differential Revision: https://reviews.bitcoinabc.org/D16155
Fabcien
pushed a commit
to Bitcoin-ABC/bitcoin-abc
that referenced
this pull request
May 20, 2024
Summary: A memory resource similar to std::pmr::unsynchronized_pool_resource, but optimized for node-based containers. Co-Authored-By: Pieter Wuille <pieter@wuille.net> This is a partial backport of [[bitcoin/bitcoin#25325 | core#25325]] bitcoin/bitcoin@b8401c3 with ARM fixes from bitcoin/bitcoin@ce881bf Test Plan: `ninja all check-all bench-bitcoin` ``` | ns/op | op/s | err% | total | benchmark |--------------------:|--------------------:|--------:|----------:|:---------- | 26.04 | 38,408,682.21 | 1.3% | 0.02 | `PoolAllocator_StdUnorderedMap` | 10.29 | 97,207,783.17 | 0.3% | 0.01 | `PoolAllocator_StdUnorderedMapWithPoolResource` ``` Reviewers: #bitcoin_abc, Fabien Reviewed By: #bitcoin_abc, Fabien Differential Revision: https://reviews.bitcoinabc.org/D16175
Fabcien
pushed a commit
to Bitcoin-ABC/bitcoin-abc
that referenced
this pull request
May 20, 2024
…cator Summary: Extracts the resource from a PoolAllocator and uses it for calculation of the node's memory usage. This is a partial backport of [[bitcoin/bitcoin#25325 | core#25325]] bitcoin/bitcoin@e19943f with an ARM fix and additional test to make sure the pool is actually used by the nodes from [[bitcoin/bitcoin#28913 | core#28913]] Depends on D16175 Test Plan: `ninja check` Reviewers: #bitcoin_abc, Fabien Reviewed By: #bitcoin_abc, Fabien Differential Revision: https://reviews.bitcoinabc.org/D16176
Fabcien
pushed a commit
to Bitcoin-ABC/bitcoin-abc
that referenced
this pull request
May 20, 2024
Summary: This frees up all associated memory with the map, not only the nodes. This is necessary in preparation for using the PoolAllocator for CCoinsMap, which does not actually free any memory on clear(). This is a partial backport of [[bitcoin/bitcoin#25325 | core#25325]] bitcoin/bitcoin@5e4ac5a Depends on D16176 Test Plan: `ninja all check-all` Reviewers: #bitcoin_abc, Fabien Reviewed By: #bitcoin_abc, Fabien Differential Revision: https://reviews.bitcoinabc.org/D16177
Fabcien
pushed a commit
to Bitcoin-ABC/bitcoin-abc
that referenced
this pull request
May 20, 2024
Summary: > In my benchmarks, using this pool allocator for CCoinsMap gives about 20% faster `-reindex-chainstate` with -dbcache=5000 with practically the same memory usage. The change in max RSS changed was 0.3%. > > The `validation_flush_tests` tests need to be updated because memory allocation is now done in large pools instead of one node at a time, so the limits need to be updated accordingly. This is a partial backport of [[bitcoin/bitcoin#25325 | core#25325]] bitcoin/bitcoin@9f947fc with a fix from [[bitcoin/bitcoin#28913 | core#28913]] Depends on D16177 Benchmark on a 12 cores AMD CPU, with the data on a spinning harddrive, using the default dbcache (1024 MB): ``` time src/bitcoind -reindex-chainstate -stopatheight=400000 -assumevalid=000000000000000004ec466ce4732fe6f1ed1cddc2ed4b328fff5224276e3f6f ``` Before: ``` real 96m2s user 27m27s sys 2min58s ``` After: ``` real 74m22s user 23m39s sys 2min23s ``` Test Plan: `ninja all check-all` Reviewers: #bitcoin_abc, Fabien Reviewed By: #bitcoin_abc, Fabien Subscribers: Fabien Differential Revision: https://reviews.bitcoinabc.org/D16178
kwvg
added a commit
to kwvg/dash
that referenced
this pull request
Aug 13, 2024
kwvg
added a commit
to kwvg/dash
that referenced
this pull request
Aug 26, 2024
kwvg
added a commit
to kwvg/dash
that referenced
this pull request
Aug 27, 2024
kwvg
added a commit
to kwvg/dash
that referenced
this pull request
Aug 27, 2024
kwvg
added a commit
to kwvg/dash
that referenced
this pull request
Aug 31, 2024
kwvg
added a commit
to kwvg/dash
that referenced
this pull request
Sep 4, 2024
PastaPastaPasta
added a commit
to dashpay/dash
that referenced
this pull request
Sep 4, 2024
, bitcoin#24021, bitcoin#24543, bitcoin#26844, bitcoin#25325, bitcoin#28165, partial bitcoin#20524, bitcoin#26036, bitcoin#27981 (networking backports: part 7) 76a458e fmt: apply formatting suggestions from `clang-format-diff.py` (Kittywhiskers Van Gogh) 63962ec merge bitcoin#28165: transport abstraction (Kittywhiskers Van Gogh) c6b9186 merge bitcoin#25325: Add pool based memory resource (Kittywhiskers Van Gogh) 8c986d6 partial bitcoin#27981: Fix potential network stalling bug (Kittywhiskers Van Gogh) 13f6dc1 merge bitcoin#26844: Pass MSG_MORE flag when sending non-final network messages (Kittywhiskers Van Gogh) caaa0fd net: use `std::deque` for `vSendMsg` instead of `std::list` (Kittywhiskers Van Gogh) 2ecba6b partial bitcoin#26036: add NetEventsInterface::g_msgproc_mutex (Kittywhiskers Van Gogh) f6c9439 merge bitcoin#24543: Move remaining globals into PeerManagerImpl (Kittywhiskers Van Gogh) dbe41ea refactor: move object request logic to `PeerManagerImpl` (Kittywhiskers Van Gogh) 112c4e0 merge bitcoin#24021: Rename and move PoissonNextSend functions (Kittywhiskers Van Gogh) 6d690ed merge bitcoin#23970: Remove pointless and confusing shift in RelayAddress (Kittywhiskers Van Gogh) 87205f2 merge bitcoin#21327: ignore transactions while in IBD (Kittywhiskers Van Gogh) 51ad8e4 merge bitcoin#21148: Split orphan handling from net_processing into txorphanage (Kittywhiskers Van Gogh) cbff29a partial bitcoin#20524: Move MIN_VERSION_SUPPORTED to p2p.py (Kittywhiskers Van Gogh) Pull request description: ## Additional Information * Dependent on #6098 * Dependent on #6233 * `p2p_ibd_txrelay.py` was first introduced in [bitcoin#19423](bitcoin#19423) to test feefilter logic but on account of Dash not having feefilter capabilities, that backport was skipped over but on account of the tests introduced in [bitcoin#21327](bitcoin#21327) that test capabilities present in Dash, a minimal version of `p2p_ibd_txrelay.py` has been committed in. * `vSendMsg` is originally a `std::deque` and as an optimization, was changed to a `std::list` in 027a852 ([dash#3398](#3398)) but this renders us unable to backport [bitcoin#26844](bitcoin#26844) as it introduces build failures. The optimization has been reverted to make way for the backport. <details> <summary>Compile failure:</summary> ``` net.cpp:959:20: error: invalid operands to binary expression ('iterator' (aka '_List_iterator<std::vector<unsigned char, std::allocator<unsigned char>>>') and 'int') if (it + 1 != node.vSendMsg.end()) { ~~ ^ ~ /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/stl_bvector.h:303:3: note: candidate function not viable: no known conversion from 'iterator' (aka '_List_iterator<std::vector<unsigned char, std::allocator<unsigned char>>>') to 'ptrdiff_t' (aka 'long') for 1st argument operator+(ptrdiff_t __n, const _Bit_iterator& __x) [...] 1 error generated. make[2]: *** [Makefile:11296: libbitcoin_server_a-net.o] Error 1 make[2]: *** Waiting for unfinished jobs.... make[2]: Leaving directory '/src/dash/src' make[1]: *** [Makefile:19171: all-recursive] Error 1 make[1]: Leaving directory '/src/dash/src' make: *** [Makefile:799: all-recursive] Error 1 ``` </details> * The collection of `CNode` pointers in `CConnman::SocketHandlerConnected` has been changed to a `std::set` to allow for us to erase elements from `vReceivableNodes` if the node is _also_ in the set of sendable nodes and the send hasn't entirely succeeded to avoid a deadlock (i.e. backport [bitcoin#27981](bitcoin#27981)) * When backporting [bitcoin#28165](bitcoin#28165), `denialofservice_tests` has been modified to still check with `vSendMsg` instead of `Transport::GetBytesToSend()` as changes in networking code to support LT-SEMs (level-triggered socket events mode) mean that the message doesn't get shifted from `vSendMsg` to `m_message_to_send`, as the test expects. * Specifically, the changes made for LT-SEM support result in the function responsible for making that shift (`Transport::SetMessageToSend()` through `CConnman::SocketSendData()`), not being called during the test runtime. * As checking `vSendMsg` (directly or through `nSendMsgSize`) isn't enough to determine if the queue is empty, we now also check with `to_send` from `Transport::GetBytesToSend()` to help us make that determination. This mirrors the change present in the upstream backport ([source](https://github.com/bitcoin/bitcoin/pull/28165/files#diff-00021eed586a482abdb09d6cdada1d90115abe988a91421851960e26658bed02R1324-R1327)). ## Breaking Changes * `bandwidth.message.*.bytesSent` will no longer include overhead and will now only report message size as specifics that let us calculate the overhead have been abstracted away. ## Checklist: - [x] I have performed a self-review of my own code - [x] I have commented my code, particularly in hard-to-understand areas **(note: N/A)** - [x] I have added or updated relevant unit/integration/functional/e2e tests - [x] I have made corresponding changes to the documentation **(note: N/A)** - [x] I have assigned this pull request to a milestone _(for repository code-owners and collaborators only)_ ACKs for top commit: PastaPastaPasta: utACK 76a458e Tree-SHA512: 2e47c207c1f854cfbd5b28c07dd78e12765ddb919abcd7710325df5d253cd0ba4bc30aa21545d88519e8acfe65638a57db4ca66853aca82fc355542210f4b394
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
A memory resource similar to
std::pmr::unsynchronized_pool_resource, but optimized for node-based containers. The goal is to be able to cache more coins with the same memory usage, and allocate/deallocate faster.This is a reimplementation of #22702. The goal was to implement it in a way that is simpler to review & test
There is now a generic
PoolResourcefor allocating/deallocating memory. This has practically the same API asstd::pmr::memory_resource. (Unfortunately I cannot use std::pmr because libc++ simply doesn't implement that API).Thanks to sipa there is now a fuzzer for PoolResource! On a fast machine I ran it for ~770 million executions without finding any issue.
The estimation of the correct node size is now gone, PoolResource now has multiple pools and just needs to be created large enough to have space for the unordered_map nodes.
I ran benchmarks with #22702, mergebase, and this PR. Frequency locked Intel i7-8700, clang++ 13.0.1 to reindex up to block 690000.
The performance is practically identical with #22702, just 0.4% slower. It's ~21% faster than master:
Note that on cache drops mergebase's memory doesnt go so far down because it does not free the
CCoinsMapbucket array.