Merge #18011: Replace current benchmarking framework with nanobench

78c312c Replace current benchmarking framework with nanobench (Martin Ankerl)

Pull request description:

  Replace current benchmarking framework with nanobench

  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

ACKs for top commit:
  laanwj:
    ACK 78c312c

Tree-SHA512: 9e18770b18b6f95a7d0105a4a5497d31cf4eb5efe6574f4482f6f1b4c88d7e0946b9a4a1e9e8e6ecbf41a3f2d7571240677dcb45af29a6f0584e89b25f32e49e

Author: laanwj

.appveyor.yml

@@ -55,7 +55,7 @@ after_build:
 #- 7z a bitcoin-%APPVEYOR_BUILD_VERSION%.zip %APPVEYOR_BUILD_FOLDER%\build_msvc\%platform%\%configuration%\*.exe
 test_script:
 - cmd: src\test_bitcoin.exe -l test_suite
-- cmd: src\bench_bitcoin.exe -evals=1 -scaling=0 > NUL
+- cmd: src\bench_bitcoin.exe > NUL
 - ps:  python test\util\bitcoin-util-test.py
 - cmd: python test\util\rpcauth-test.py
 # Fee estimation test failing on appveyor with: WinError 10048] Only one usage of each socket address (protocol/network address/port) is normally permitted.

contrib/devtools/copyright_header.py

@@ -22,6 +22,7 @@
     'src/reverse_iterator.h',
     'src/test/fuzz/FuzzedDataProvider.h',
     'src/tinyformat.h',
+    'src/bench/nanobench.h',
     'test/functional/test_framework/bignum.py',
     # python init:
     '*__init__.py',

doc/benchmarking.md

@@ -19,8 +19,10 @@ After compiling bitcoin-core, the benchmarks can be run with:
 
 The output will look similar to:
 ```
-# Benchmark, evals, iterations, total, min, max, median
-AssembleBlock, 5, 700, 1.79954, 0.000510913, 0.000517018, 0.000514497
+|             ns/byte |              byte/s | error % | benchmark
+|--------------------:|--------------------:|--------:|:----------------------------------------------
+|               64.13 |       15,592,356.01 |    0.1% | `Base58CheckEncode`
+|               24.56 |       40,722,672.68 |    0.2% | `Base58Decode`
 ...
 ```
 

src/Makefile.bench.include

@@ -33,6 +33,8 @@ bench_bench_bitcoin_SOURCES = \
   bench/merkle_root.cpp \
   bench/mempool_eviction.cpp \
   bench/mempool_stress.cpp \
+  bench/nanobench.h \
+  bench/nanobench.cpp \
   bench/rpc_blockchain.cpp \
   bench/rpc_mempool.cpp \
   bench/util_time.cpp \

src/Makefile.test.include

@@ -1237,8 +1237,8 @@ endif
 if TARGET_WINDOWS
 else
 if ENABLE_BENCH
-	@echo "Running bench/bench_bitcoin -evals=1 -scaling=0..."
-	$(BENCH_BINARY) -evals=1 -scaling=0 > /dev/null
+	@echo "Running bench/bench_bitcoin ..."
+	$(BENCH_BINARY) > /dev/null
 endif
 endif
 	$(AM_V_at)$(MAKE) $(AM_MAKEFLAGS) -C secp256k1 check

src/bench/addrman.cpp

@@ -67,52 +67,52 @@ static void FillAddrMan(CAddrMan& addrman)
 
 /* Benchmarks */
 
-static void AddrManAdd(benchmark::State& state)
+static void AddrManAdd(benchmark::Bench& bench)
 {
     CreateAddresses();
 
     CAddrMan addrman;
 
-    while (state.KeepRunning()) {
+    bench.run([&] {
         AddAddressesToAddrMan(addrman);
         addrman.Clear();
-    }
+    });
 }
 
-static void AddrManSelect(benchmark::State& state)
+static void AddrManSelect(benchmark::Bench& bench)
 {
     CAddrMan addrman;
 
     FillAddrMan(addrman);
 
-    while (state.KeepRunning()) {
+    bench.run([&] {
         const auto& address = addrman.Select();
         assert(address.GetPort() > 0);
-    }
+    });
 }
 
-static void AddrManGetAddr(benchmark::State& state)
+static void AddrManGetAddr(benchmark::Bench& bench)
 {
     CAddrMan addrman;
 
     FillAddrMan(addrman);
 
-    while (state.KeepRunning()) {
+    bench.run([&] {
         const auto& addresses = addrman.GetAddr();
         assert(addresses.size() > 0);
-    }
+    });
 }
 
-static void AddrManGood(benchmark::State& state)
+static void AddrManGood(benchmark::Bench& bench)
 {
     /* Create many CAddrMan objects - one to be modified at each loop iteration.
      * This is necessary because the CAddrMan::Good() method modifies the
      * object, affecting the timing of subsequent calls to the same method and
      * we want to do the same amount of work in every loop iteration. */
 
-    const uint64_t numLoops = state.m_num_iters * state.m_num_evals;
+    bench.epochs(5).epochIterations(1);
 
-    std::vector<CAddrMan> addrmans(numLoops);
+    std::vector<CAddrMan> addrmans(bench.epochs() * bench.epochIterations());
     for (auto& addrman : addrmans) {
         FillAddrMan(addrman);
     }
@@ -128,13 +128,13 @@ static void AddrManGood(benchmark::State& state)
     };
 
     uint64_t i = 0;
-    while (state.KeepRunning()) {
+    bench.run([&] {
         markSomeAsGood(addrmans.at(i));
         ++i;
-    }
+    });
 }
 
-BENCHMARK(AddrManAdd, 5);
-BENCHMARK(AddrManSelect, 1000000);
-BENCHMARK(AddrManGetAddr, 500);
-BENCHMARK(AddrManGood, 2);
+BENCHMARK(AddrManAdd);
+BENCHMARK(AddrManSelect);
+BENCHMARK(AddrManGetAddr);
+BENCHMARK(AddrManGood);

src/bench/base58.cpp

@@ -10,7 +10,7 @@
 #include <vector>
 
 
-static void Base58Encode(benchmark::State& state)
+static void Base58Encode(benchmark::Bench& bench)
 {
     static const std::array<unsigned char, 32> buff = {
         {
@@ -19,13 +19,13 @@ static void Base58Encode(benchmark::State& state)
             200, 24
         }
     };
-    while (state.KeepRunning()) {
+    bench.batch(buff.size()).unit("byte").run([&] {
         EncodeBase58(buff.data(), buff.data() + buff.size());
-    }
+    });
 }
 
 
-static void Base58CheckEncode(benchmark::State& state)
+static void Base58CheckEncode(benchmark::Bench& bench)
 {
     static const std::array<unsigned char, 32> buff = {
         {
@@ -36,22 +36,22 @@ static void Base58CheckEncode(benchmark::State& state)
     };
     std::vector<unsigned char> vch;
     vch.assign(buff.begin(), buff.end());
-    while (state.KeepRunning()) {
+    bench.batch(buff.size()).unit("byte").run([&] {
         EncodeBase58Check(vch);
-    }
+    });
 }
 
 
-static void Base58Decode(benchmark::State& state)
+static void Base58Decode(benchmark::Bench& bench)
 {
     const char* addr = "17VZNX1SN5NtKa8UQFxwQbFeFc3iqRYhem";
     std::vector<unsigned char> vch;
-    while (state.KeepRunning()) {
+    bench.batch(strlen(addr)).unit("byte").run([&] {
         (void) DecodeBase58(addr, vch, 64);
-    }
+    });
 }
 
 
-BENCHMARK(Base58Encode, 470 * 1000);
-BENCHMARK(Base58CheckEncode, 320 * 1000);
-BENCHMARK(Base58Decode, 800 * 1000);
+BENCHMARK(Base58Encode);
+BENCHMARK(Base58CheckEncode);
+BENCHMARK(Base58Decode);

src/bench/bech32.cpp

@@ -3,6 +3,7 @@
 // file COPYING or http://www.opensource.org/licenses/mit-license.php.
 
 #include <bench/bench.h>
+#include <bench/nanobench.h>
 
 #include <bech32.h>
 #include <util/strencodings.h>
@@ -11,26 +12,26 @@
 #include <vector>
 
 
-static void Bech32Encode(benchmark::State& state)
+static void Bech32Encode(benchmark::Bench& bench)
 {
     std::vector<uint8_t> v = ParseHex("c97f5a67ec381b760aeaf67573bc164845ff39a3bb26a1cee401ac67243b48db");
     std::vector<unsigned char> tmp = {0};
     tmp.reserve(1 + 32 * 8 / 5);
     ConvertBits<8, 5, true>([&](unsigned char c) { tmp.push_back(c); }, v.begin(), v.end());
-    while (state.KeepRunning()) {
+    bench.batch(v.size()).unit("byte").run([&] {
         bech32::Encode("bc", tmp);
-    }
+    });
 }
 
 
-static void Bech32Decode(benchmark::State& state)
+static void Bech32Decode(benchmark::Bench& bench)
 {
     std::string addr = "bc1qkallence7tjawwvy0dwt4twc62qjgaw8f4vlhyd006d99f09";
-    while (state.KeepRunning()) {
+    bench.batch(addr.size()).unit("byte").run([&] {
         bech32::Decode(addr);
-    }
+    });
 }
 
 
-BENCHMARK(Bech32Encode, 800 * 1000);
-BENCHMARK(Bech32Decode, 800 * 1000);
+BENCHMARK(Bech32Encode);
+BENCHMARK(Bech32Decode);