Add asymptotes for benchmarking framework #17375

JeremyRubin · 2019-11-05T00:30:31Z

It can be pretty convenient for certain classes of benchmark to be able to run it with dynamic parameters, to generate curves of magnitudes runtimes.

This PR adds this functionality and shows an example using it.

This functionality is a bit useless unless you also combine it with a filter to ensure the params are meaningful for the target asymptotic benchmark.

laanwj · 2019-11-06T09:45:16Z

src/bench/bench_bitcoin.cpp

@@ -25,6 +26,7 @@ static void SetupBenchArgs()
    gArgs.AddArg("-evals=<n>", strprintf("Number of measurement evaluations to perform. (default: %u)", DEFAULT_BENCH_EVALUATIONS), ArgsManager::ALLOW_ANY, OptionsCategory::OPTIONS);
    gArgs.AddArg("-filter=<regex>", strprintf("Regular expression filter to select benchmark by name (default: %s)", DEFAULT_BENCH_FILTER), ArgsManager::ALLOW_ANY, OptionsCategory::OPTIONS);
    gArgs.AddArg("-scaling=<n>", strprintf("Scaling factor for benchmark's runtime (default: %u)", DEFAULT_BENCH_SCALING), ArgsManager::ALLOW_ANY, OptionsCategory::OPTIONS);
+    gArgs.AddArg("-asymptote=n", strprintf("Asymptotic factors for benchmark's runtime, positional (default: \"\")"), ArgsManager::ALLOW_ANY, OptionsCategory::OPTIONS);


Can you please add some short summary, as to what 'asymptotic factors' mean in this context? I don't think it's self-explanatory, this argument help gives me zero context on how to use it in practice 😄

I squash-pushed improved documentation for this parameter, which includes an example of how to use it.

maflcko · 2019-11-07T15:48:30Z

Not sure how useful the general framework is. A smaller patch would be to just copy past the BENCHMARK(ComplexMemPool, 1); line and hardcode the "asymptotes". Or if you like, make the benchmark settings dynamic, so that an asymptote can be created by running bench_bitcoin -filter=... -scale_asymptotes_factor=5 to multiply all dynamic settings in the selected/filtered bench by that factor.

JeremyRubin · 2019-11-07T23:18:08Z

@MarcoFalke that's insufficient for the example given. In that benchmark, all different sorts of asymptotic behaviors can be triggered with different growth factors.

I typically run it something like this:

for x in 10 100 1000 10000 100000 1000000; do ./src/bench/bench_bitcoin -asymptote=100 -asymptote=100 -asymptote=$x -asymptote=10 --evals=1; done | grep Com --line-buffered | cut -d, -f6

Scaling all the factors doesn't let you express nuances.

The other issue is that if you just copy the existing Benchmark stuff you lose the 'runs in one second" property.

DrahtBot · 2019-12-20T18:37:06Z

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

Conflicts

Reviewers, this pull request conflicts with the following ones:

#17781 (rpc: Remove mempool global from miner by MarcoFalke)

If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.

DrahtBot · 2020-01-02T23:22:12Z

Needs rebase

JeremyRubin · 2020-01-28T20:12:23Z

closed in favor of #18011

This adds support to calculate asymptotic complexity of a benchmark. This is similar to bitcoin#17375, but currently only one asymptote is supported, and I have added support in the benchmark `ComplexMemPool` as an example. Usage is e.g. like this: ``` ./bench_bitcoin -filter=ComplexMemPool -asymptote=25,50,100,200,400,600,800 ``` This runs the benchmark `ComplexMemPool` several times but with different complexityN settings. The benchmark can extract that number and use it accordingly. Here, it's used for `childTxs`. The output is this: | complexityN | ns/op | op/s | err% | ins/op | cyc/op | IPC | total | benchmark |------------:|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|----------:|:---------- | 25 | 1,064,241.00 | 939.64 | 1.4% | 3,960,279.00 | 2,829,708.00 | 1.400 | 0.01 | `ComplexMemPool` | 50 | 1,579,530.00 | 633.10 | 1.0% | 6,231,810.00 | 4,412,674.00 | 1.412 | 0.02 | `ComplexMemPool` | 100 | 4,022,774.00 | 248.58 | 0.6% | 16,544,406.00 | 11,889,535.00 | 1.392 | 0.04 | `ComplexMemPool` | 200 | 15,390,986.00 | 64.97 | 0.2% | 63,904,254.00 | 47,731,705.00 | 1.339 | 0.17 | `ComplexMemPool` | 400 | 69,394,711.00 | 14.41 | 0.1% | 272,602,461.00 | 219,014,691.00 | 1.245 | 0.76 | `ComplexMemPool` | 600 | 168,977,165.00 | 5.92 | 0.1% | 639,108,082.00 | 535,316,887.00 | 1.194 | 1.86 | `ComplexMemPool` | 800 | 310,109,077.00 | 3.22 | 0.1% |1,149,134,246.00 | 984,620,812.00 | 1.167 | 3.41 | `ComplexMemPool` | coefficient | err% | complexity |--------------:|-------:|------------ | 4.78486e-07 | 4.5% | O(n^2) | 6.38557e-10 | 21.7% | O(n^3) | 3.42338e-05 | 38.0% | O(n log n) | 0.000313914 | 46.9% | O(n) | 0.0129823 | 114.4% | O(log n) | 0.0815055 | 133.8% | O(1) The best fitting curve is O(n^2), so the algorithm seems to scale quadratic with `childTxs` in the range 25 to 800.

This replaces the current benchmarking framework with nanobench [1], an MIT licensed single-header benchmarking library, of which I am the autor. This has in my opinion several advantages, especially on Linux: * fast: Running all benchmarks takes ~6 seconds instead of 4m13s on an Intel i7-8700 CPU @ 3.20GHz. * accurate: I ran e.g. the benchmark for SipHash_32b 10 times and calculate standard deviation / mean = coefficient of variation: * 0.57% CV for old benchmarking framework * 0.20% CV for nanobench So the benchmark results with nanobench seem to vary less than with the old framework. * It automatically determines runtime based on clock precision, no need to specify number of evaluations. * measure instructions, cycles, branches, instructions per cycle, branch misses (only Linux, when performance counters are available) * output in markdown table format. * Warn about unstable environment (frequency scaling, turbo, ...) * For better profiling, it is possible to set the environment variable NANOBENCH_ENDLESS to force endless running of a particular benchmark without the need to recompile. This makes it to e.g. run "perf top" and look at hotspots. Here is an example copy & pasted from the terminal output: | ns/byte | byte/s | err% | ins/byte | cyc/byte | IPC | bra/byte | miss% | total | benchmark |--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:---------- | 2.52 | 396,529,415.94 | 0.6% | 25.42 | 8.02 | 3.169 | 0.06 | 0.0% | 0.03 | `bench/crypto_hash.cpp RIPEMD160` | 1.87 | 535,161,444.83 | 0.3% | 21.36 | 5.95 | 3.589 | 0.06 | 0.0% | 0.02 | `bench/crypto_hash.cpp SHA1` | 3.22 | 310,344,174.79 | 1.1% | 36.80 | 10.22 | 3.601 | 0.09 | 0.0% | 0.04 | `bench/crypto_hash.cpp SHA256` | 2.01 | 496,375,796.23 | 0.0% | 18.72 | 6.43 | 2.911 | 0.01 | 1.0% | 0.00 | `bench/crypto_hash.cpp SHA256D64_1024` | 7.23 | 138,263,519.35 | 0.1% | 82.66 | 23.11 | 3.577 | 1.63 | 0.1% | 0.00 | `bench/crypto_hash.cpp SHA256_32b` | 3.04 | 328,780,166.40 | 0.3% | 35.82 | 9.69 | 3.696 | 0.03 | 0.0% | 0.03 | `bench/crypto_hash.cpp SHA512` [1] https://github.com/martinus/nanobench * Adds support for asymptotes This adds support to calculate asymptotic complexity of a benchmark. This is similar to bitcoin#17375, but currently only one asymptote is supported, and I have added support in the benchmark `ComplexMemPool` as an example. Usage is e.g. like this: ``` ./bench_bitcoin -filter=ComplexMemPool -asymptote=25,50,100,200,400,600,800 ``` This runs the benchmark `ComplexMemPool` several times but with different complexityN settings. The benchmark can extract that number and use it accordingly. Here, it's used for `childTxs`. The output is this: | complexityN | 8000 ns/op | op/s | err% | ins/op | cyc/op | IPC | total | benchmark |------------:|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|----------:|:---------- | 25 | 1,064,241.00 | 939.64 | 1.4% | 3,960,279.00 | 2,829,708.00 | 1.400 | 0.01 | `ComplexMemPool` | 50 | 1,579,530.00 | 633.10 | 1.0% | 6,231,810.00 | 4,412,674.00 | 1.412 | 0.02 | `ComplexMemPool` | 100 | 4,022,774.00 | 248.58 | 0.6% | 16,544,406.00 | 11,889,535.00 | 1.392 | 0.04 | `ComplexMemPool` | 200 | 15,390,986.00 | 64.97 | 0.2% | 63,904,254.00 | 47,731,705.00 | 1.339 | 0.17 | `ComplexMemPool` | 400 | 69,394,711.00 | 14.41 | 0.1% | 272,602,461.00 | 219,014,691.00 | 1.245 | 0.76 | `ComplexMemPool` | 600 | 168,977,165.00 | 5.92 | 0.1% | 639,108,082.00 | 535,316,887.00 | 1.194 | 1.86 | `ComplexMemPool` | 800 | 310,109,077.00 | 3.22 | 0.1% |1,149,134,246.00 | 984,620,812.00 | 1.167 | 3.41 | `ComplexMemPool` | coefficient | err% | complexity |--------------:|-------:|------------ | 4.78486e-07 | 4.5% | O(n^2) | 6.38557e-10 | 21.7% | O(n^3) | 3.42338e-05 | 38.0% | O(n log n) | 0.000313914 | 46.9% | O(n) | 0.0129823 | 114.4% | O(log n) | 0.0815055 | 133.8% | O(1) The best fitting curve is O(n^2), so the algorithm seems to scale quadratic with `childTxs` in the range 25 to 800.

This replaces the current benchmarking framework with nanobench [1], an MIT licensed single-header benchmarking library, of which I am the autor. This has in my opinion several advantages, especially on Linux: * fast: Running all benchmarks takes ~6 seconds instead of 4m13s on an Intel i7-8700 CPU @ 3.20GHz. * accurate: I ran e.g. the benchmark for SipHash_32b 10 times and calculate standard deviation / mean = coefficient of variation: * 0.57% CV for old benchmarking framework * 0.20% CV for nanobench So the benchmark results with nanobench seem to vary less than with the old framework. * It automatically determines runtime based on clock precision, no need to specify number of evaluations. * measure instructions, cycles, branches, instructions per cycle, branch misses (only Linux, when performance counters are available) * output in markdown table format. * Warn about unstable environment (frequency scaling, turbo, ...) * For better profiling, it is possible to set the environment variable NANOBENCH_ENDLESS to force endless running of a particular benchmark without the need to recompile. This makes it to e.g. run "perf top" and look at hotspots. Here is an example copy & pasted from the terminal output: | ns/byte | byte/s | err% | ins/byte | cyc/byte | IPC | bra/byte | miss% | total | benchmark |--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:---------- | 2.52 | 396,529,415.94 | 0.6% | 25.42 | 8.02 | 3.169 | 0.06 | 0.0% | 0.03 | `bench/crypto_hash.cpp RIPEMD160` | 1.87 | 535,161,444.83 | 0.3% | 21.36 | 5.95 | 3.589 | 0.06 | 0.0% | 0.02 | `bench/crypto_hash.cpp SHA1` | 3.22 | 310,344,174.79 | 1.1% | 36.80 | 10.22 | 3.601 | 0.09 | 0.0% | 0.04 | `bench/crypto_hash.cpp SHA256` | 2.01 | 496,375,796.23 | 0.0% | 18.72 | 6.43 | 2.911 | 0.01 | 1.0% | 0.00 | `bench/crypto_hash.cpp SHA256D64_1024` | 7.23 | 138,263,519.35 | 0.1% | 82.66 | 23.11 | 3.577 | 1.63 | 0.1% | 0.00 | `bench/crypto_hash.cpp SHA256_32b` | 3.04 | 328,780,166.40 | 0.3% | 35.82 | 9.69 | 3.696 | 0.03 | 0.0% | 0.03 | `bench/crypto_hash.cpp SHA512` [1] https://github.com/martinus/nanobench * Adds support for asymptotes This adds support to calculate asymptotic complexity of a benchmark. This is similar to bitcoin#17375, but currently only one asymptote is supported, and I have added support in the benchmark `ComplexMemPool` as an example. Usage is e.g. like this: ``` ./bench_bitcoin -filter=ComplexMemPool -asymptote=25,50,100,200,400,600,800 ``` This runs the benchmark `ComplexMemPool` several times but with different complexityN settings. The benchmark can extract that number and use it accordingly. Here, it's used for `childTxs`. The output is this: | complexityN | ns/op | op/s | err% | ins/op | cyc/op | IPC | total | benchmark |------------:|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|----------:|:---------- | 25 | 1,064,241.00 | 939.64 | 1.4% | 3,960,279.00 | 2,829,708.00 | 1.400 | 0.01 | `ComplexMemPool` | 50 | 1,579,530.00 | 633.10 | 1.0% | 6,231,810.00 | 4,412,674.00 | 1.412 | 0.02 | `ComplexMemPool` | 100 | 4,022,774.00 | 248.58 | 0.6% | 16,544,406.00 | 11,889,535.00 | 1.392 | 0.04 | `ComplexMemPool` | 200 | 15,390,986.00 | 64.97 | 0.2% | 63,904,254.00 | 47,731,705.00 | 1.339 | 0.17 | `ComplexMemPool` | 400 | 69,394,711.00 | 14.41 | 0.1% | 272,602,461.00 | 219,014,691.00 | 1.245 | 0.76 | `ComplexMemPool` | 600 | 168,977,165.00 | 5.92 | 0.1% | 639,108,082.00 | 535,316,887.00 | 1.194 | 1.86 | `ComplexMemPool` | 800 | 310,109,077.00 | 3.22 | 0.1% |1,149,134,246.00 | 984,620,812.00 | 1.167 | 3.41 | `ComplexMemPool` | coefficient | err% | complexity |--------------:|-------:|------------ | 4.78486e-07 | 4.5% | O(n^2) | 6.38557e-10 | 21.7% | O(n^3) | 3.42338e-05 | 38.0% | O(n log n) | 0.000313914 | 46.9% | O(n) | 0.0129823 | 114.4% | O(log n) | 0.0815055 | 133.8% | O(1) The best fitting curve is O(n^2), so the algorithm seems to scale quadratic with `childTxs` in the range 25 to 800.

This replaces the current benchmarking framework with nanobench [1], an MIT licensed single-header benchmarking library, of which I am the autor. This has in my opinion several advantages, especially on Linux: * fast: Running all benchmarks takes ~6 seconds instead of 4m13s on an Intel i7-8700 CPU @ 3.20GHz. * accurate: I ran e.g. the benchmark for SipHash_32b 10 times and calculate standard deviation / mean = coefficient of variation: * 0.57% CV for old benchmarking framework * 0.20% CV for nanobench So the benchmark results with nanobench seem to vary less than with the old framework. * It automatically determines runtime based on clock precision, no need to specify number of evaluations. * measure instructions, cycles, branches, instructions per cycle, branch misses (only Linux, when performance counters are available) * output in markdown table format. * Warn about unstable environment (frequency scaling, turbo, ...) * For better profiling, it is possible to set the environment variable NANOBENCH_ENDLESS to force endless running of a particular benchmark without the need to recompile. This makes it to e.g. run "perf top" and look at hotspots. Here is an example copy & pasted from the terminal output: | ns/byte | byte/s | err% | ins/byte | cyc/byte | IPC | bra/byte | miss% | total | benchmark |--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:---------- | 2.52 | 396,529,415.94 | 0.6% | 25.42 | 8.02 | 3.169 | 0.06 | 0.0% | 0.03 | `bench/crypto_hash.cpp RIPEMD160` | 1.87 | 535,161,444.83 | 0.3% | 21.36 | 5.95 | 3.589 | 0.06 | 0.0% | 0.02 | `bench/crypto_hash.cpp SHA1` | 3.22 | 310,344,174.79 | 1.1% | 36.80 | 10.22 | 3.601 | 0.09 | 0.0% | 0.04 | `bench/crypto_hash.cpp SHA256` | 2.01 | 496,375,796.23 | 0.0% | 18.72 | 6.43 | 2.911 | 0.01 | 1.0% | 0.00 | `bench/crypto_hash.cpp SHA256D64_1024` | 7.23 | 138,263,519.35 | 0.1% | 82.66 | 23.11 | 3.577 | 1.63 | 0.1% | 0.00 | `bench/crypto_hash.cpp SHA256_32b` | 3.04 | 328,780,166.40 | 0.3% | 35.82 | 9.69 | 3.696 | 0.03 | 0.0% | 0.03 | `bench/crypto_hash.cpp SHA512` [1] https://github.com/martinus/nanobench * Adds support for asymptotes This adds support to calculate asymptotic complexity of a benchmark. This is similar to bitcoin#17375, but currently only one asymptote is supported, and I have added support in the benchmark `ComplexMemPool` as an example. Usage is e.g. like this: ``` ./bench_bitcoin -filter=ComplexMemPool -asymptote=25,50,100,200,400,600,800 ``` This runs the benchmark `ComplexMemPool` several times but with different complexityN settings. The benchmark can extract that number and use it accordingly. Here, it's used for `childTxs`. The output is this: | complexityN | ns/op | op/s | err% | ins/op | cyc/op | IPC | total | benchmark |------------:|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|----------:|:---------- | 25 | 1,064,241.00 | 939.64 | 1.4% | 3,960,279.00 | 2,829,708.00 | 1.400 | 0.01 | `ComplexMemPool` | 50 | 1,579,530.00 | 633.10 | 1.0% | 6,231,810.00 | 4,412,674.00 | 1.412 | 0.02 | `ComplexMemPool` | 100 | 4,022,774.00 | 248.58 | 0.6% | 16,544,406.00 | 11,889,535.00 | 1.392 | 0.04 | `ComplexMemPool` | 200 | 15,390,986.00 | 64.97 | 0.2% | 63,904,254.00 | 47,731,705.00 | 1.339 | 0.17 | `ComplexMemPool` | 400 | 69,394,711.00 | 14.41 | 0.1% | 272,602,461.00 | 219,014,691.00 | 1.245 | 0.76 | `ComplexMemPool` | 600 | 168,977,165.00 | 5.92 | 0.1% | 639,108,082.00 | 535,316,887.00 | 1.194 | 1.86 | `ComplexMemPool` | 800 | 310,109,077.00 | 3.22 | 0.1% |1,149,134,246.00 | 984,620,812.00 | 1.167 | 3.41 | `ComplexMemPool` | coefficient | err% | complexity |--------------:|-------:|------------ | 4.78486e-07 | 4.5% | O(n^2) | 6.38557e-10 | 21.7% | O(n^3) | 3.42338e-05 | 38.0% | O(n log n) | 0.000313914 | 46.9% | O(n) | 0.0129823 | 114.4% | O(log n) | 0.0815055 | 133.8% | O(1) The best fitting curve is O(n^2), so the algorithm seems to scale quadratic with `childTxs` in the range 25 to 800. # Conflicts: # .appveyor.yml # contrib/devtools/copyright_header.py # doc/benc 7656 hmarking.md # src/Makefile.bench.include # src/Makefile.test.include # src/bench/addrman.cpp # src/bench/base58.cpp # src/bench/bech32.cpp # src/bench/bench.cpp # src/bench/bench.h # src/bench/bench_bitcoin.cpp # src/bench/block_assemble.cpp # src/bench/ccoins_caching.cpp # src/bench/checkblock.cpp # src/bench/coin_selection.cpp # src/bench/crypto_hash.cpp # src/bench/duplicate_inputs.cpp # src/bench/examples.cpp # src/bench/gcs_filter.cpp # src/bench/hashpadding.cpp # src/bench/mempool_eviction.cpp # src/bench/mempool_stress.cpp # src/bench/prevector.cpp # src/bench/rollingbloom.cpp # src/bench/rpc_blockchain.cpp # src/bench/rpc_mempool.cpp # src/bench/verify_script.cpp # src/bench/wallet_balance.cpp # test/lint/lint-include-guards.sh

fanquake added the Tests label Nov 5, 2019

laanwj reviewed Nov 6, 2019

View reviewed changes

Add asymptotes for benchmarking framework

32862e5

JeremyRubin force-pushed the asymptotic-benchmarks branch from f160100 to 32862e5 Compare November 11, 2019 19:45

DrahtBot added the Needs rebase label Jan 2, 2020

JeremyRubin mentioned this pull request Jan 27, 2020

Replace current benchmarking framework with nanobench #18011

Merged

JeremyRubin closed this Jan 28, 2020

bitcoin locked as resolved and limited conversation to collaborators Feb 15, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add asymptotes for benchmarking framework #17375

Add asymptotes for benchmarking framework #17375

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Add asymptotes for benchmarking framework #17375

Add asymptotes for benchmarking framework #17375

Uh oh!

Conversation

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Conflicts

Uh oh!

Uh oh!

Uh oh!

Uh oh!