diff options
| author | Wladimir J. van der Laan <[email protected]> | 2018-03-01 12:12:55 +0100 |
|---|---|---|
| committer | Wladimir J. van der Laan <[email protected]> | 2018-03-01 12:13:08 +0100 |
| commit | 32987d5aebc4e0fb757ad77dc727e5d7ee9d394b (patch) | |
| tree | c40fa26e4d0350f6609b3557030c800ea9921631 /src/bench | |
| parent | Merge #12556: [Trivial] fix version typo in getpeerinfo RPC call help (diff) | |
| parent | Use memset() to optimize prevector::resize() (diff) | |
| download | discoin-32987d5aebc4e0fb757ad77dc727e5d7ee9d394b.tar.xz discoin-32987d5aebc4e0fb757ad77dc727e5d7ee9d394b.zip | |
Merge #12549: Make prevector::resize() and other prevector operations much faster
5aad635 Use memset() to optimize prevector::resize() (Evan Klitzke)
e46be25 Reduce redundant code of prevector and speed it up (Akio Nakamura)
f0e7aa7 Add new prevector benchmarks. (Evan Klitzke)
Pull request description:
This branch optimizes various `prevector` operations, especially resizing vectors. While profiling the `loadblk` thread I noticed that a lot of time was being spent in `prevector::resize()` which led to this work. I have some data here indicating that it takes up **37%** of the time in `ReadBlockFromDisk()`: https://monad.io/readblockfromdisk.svg
This branch improves things significantly. For trivial types, the new results for the prevector benchmark are:
* `PrevectorClearTrivial` which tests `prevector::clear()` becomes 24.6x faster
* `PrevectorDestructorTrivial` which tests `prevector::~prevector()` becomes 20.5x faster
* `PrevectorResizeTrivial` which tests `prevector::resize()` becomes 20.3x faster
Note that in practice it looks like the prevector is only used to contain `unsigned char` types, which is a trivial type. The benchmarks are testing a bit of an extreme case, but the changes here are motivated by the profiling data for `ReadBlockFromDisk()` I linked to above.
The pull request here consists of a series of three commits:
* The first adds new benchmarks but does not change the prevector code.
* The second is from @AkioNak , and merges some prevector optimizations he submitted in #11988
* The third optimizes `prevector::resize()` to use `memset()` when the prevector contains trivially constructible types
Tree-SHA512: 28f7cbb91a19f9f43b6a5942781d7eb2e3197389186b666f086b69df12bee37773140f765426d715bfb8ebff79cb27a5f1206d0325b54b4aa65598b50fb18368
Diffstat (limited to 'src/bench')
| -rw-r--r-- | src/bench/prevector.cpp | 77 | ||||
| -rw-r--r-- | src/bench/prevector_destructor.cpp | 36 |
2 files changed, 77 insertions, 36 deletions
diff --git a/src/bench/prevector.cpp b/src/bench/prevector.cpp new file mode 100644 index 000000000..d0f28d1a3 --- /dev/null +++ b/src/bench/prevector.cpp @@ -0,0 +1,77 @@ +// Copyright (c) 2015-2017 The Bitcoin Core developers +// Distributed under the MIT software license, see the accompanying +// file COPYING or http://www.opensource.org/licenses/mit-license.php. + +#include <compat.h> +#include <prevector.h> + +#include <bench/bench.h> + +struct nontrivial_t { + int x; + nontrivial_t() :x(-1) {} +}; +static_assert(!IS_TRIVIALLY_CONSTRUCTIBLE<nontrivial_t>::value, + "expected nontrivial_t to not be trivially constructible"); + +typedef unsigned char trivial_t; +static_assert(IS_TRIVIALLY_CONSTRUCTIBLE<trivial_t>::value, + "expected trivial_t to be trivially constructible"); + +template <typename T> +static void PrevectorDestructor(benchmark::State& state) +{ + while (state.KeepRunning()) { + for (auto x = 0; x < 1000; ++x) { + prevector<28, T> t0; + prevector<28, T> t1; + t0.resize(28); + t1.resize(29); + } + } +} + +template <typename T> +static void PrevectorClear(benchmark::State& state) +{ + + while (state.KeepRunning()) { + for (auto x = 0; x < 1000; ++x) { + prevector<28, T> t0; + prevector<28, T> t1; + t0.resize(28); + t0.clear(); + t1.resize(29); + t0.clear(); + } + } +} + +template <typename T> +void PrevectorResize(benchmark::State& state) +{ + while (state.KeepRunning()) { + prevector<28, T> t0; + prevector<28, T> t1; + for (auto x = 0; x < 1000; ++x) { + t0.resize(28); + t0.resize(0); + t1.resize(29); + t1.resize(0); + } + } +} + +#define PREVECTOR_TEST(name, nontrivops, trivops) \ + static void Prevector ## name ## Nontrivial(benchmark::State& state) { \ + PrevectorResize<nontrivial_t>(state); \ + } \ + BENCHMARK(Prevector ## name ## Nontrivial, nontrivops); \ + static void Prevector ## name ## Trivial(benchmark::State& state) { \ + PrevectorResize<trivial_t>(state); \ + } \ + BENCHMARK(Prevector ## name ## Trivial, trivops); + +PREVECTOR_TEST(Clear, 28300, 88600) +PREVECTOR_TEST(Destructor, 28800, 88900) +PREVECTOR_TEST(Resize, 28900, 90300) diff --git a/src/bench/prevector_destructor.cpp b/src/bench/prevector_destructor.cpp deleted file mode 100644 index 39d0ee5eb..000000000 --- a/src/bench/prevector_destructor.cpp +++ /dev/null @@ -1,36 +0,0 @@ -// Copyright (c) 2015-2017 The Bitcoin Core developers -// Distributed under the MIT software license, see the accompanying -// file COPYING or http://www.opensource.org/licenses/mit-license.php. - -#include <bench/bench.h> -#include <prevector.h> - -static void PrevectorDestructor(benchmark::State& state) -{ - while (state.KeepRunning()) { - for (auto x = 0; x < 1000; ++x) { - prevector<28, unsigned char> t0; - prevector<28, unsigned char> t1; - t0.resize(28); - t1.resize(29); - } - } -} - -static void PrevectorClear(benchmark::State& state) -{ - - while (state.KeepRunning()) { - for (auto x = 0; x < 1000; ++x) { - prevector<28, unsigned char> t0; - prevector<28, unsigned char> t1; - t0.resize(28); - t0.clear(); - t1.resize(29); - t0.clear(); - } - } -} - -BENCHMARK(PrevectorDestructor, 5700); -BENCHMARK(PrevectorClear, 5600); |