Merge bitcoin/bitcoin#27598: bench: Add SHA256 implementation specific benchmarks

ce6df7df9bab2405cfe7d6e382f5682cf30de476 bench: Add SHA256 implementation specific benchmarks (Hennadii Stepanov)
5f72417176cfffece9a5aa11e97d5a6599c51e7a Add ability to specify SHA256 implementation for benchmark purposes (Hennadii Stepanov)

Pull request description:

  On the master branch, only the best available `SHA256` implementation is being benchmarked. This PR makes `bench_bitcoin` benchmark all `SHA256` implementations that are available on the system.

  For  example:
  - on Linux:
  ```
  $ ./src/bench/bench_bitcoin -filter=SHA.*
  Using the 'x86_shani(1way,2way)' SHA256 implementation

  |             ns/byte |              byte/s |    err% |     total | benchmark
  |--------------------:|--------------------:|--------:|----------:|:----------
  |                1.00 |    1,002,545,462.93 |    0.4% |      0.01 | `SHA1`
  |                2.91 |      344,117,991.18 |    0.1% |      0.03 | `SHA256 using the 'standard' SHA256 implementation`
  |                2.21 |      453,081,794.40 |    0.1% |      0.02 | `SHA256 using the 'sse4(1way),sse41(4way)' SHA256 implementation`
  |                2.21 |      453,396,506.58 |    0.1% |      0.02 | `SHA256 using the 'sse4(1way),sse41(4way),avx2(8way)' SHA256 implementation`
  |                0.53 |    1,870,520,687.49 |    0.1% |      0.01 | `SHA256 using the 'x86_shani(1way,2way)' SHA256 implementation`
  |                7.90 |      126,627,134.33 |    0.0% |      0.01 | `SHA256D64_1024 using the 'standard' SHA256 implementation`
  |                3.94 |      253,850,206.07 |    0.0% |      0.01 | `SHA256D64_1024 using the 'sse4(1way),sse41(4way)' SHA256 implementation`
  |                1.40 |      716,247,553.38 |    0.4% |      0.01 | `SHA256D64_1024 using the 'sse4(1way),sse41(4way),avx2(8way)' SHA256 implementation`
  |                1.26 |      792,706,270.13 |    0.9% |      0.01 | `SHA256D64_1024 using the 'x86_shani(1way,2way)' SHA256 implementation`
  |                6.75 |      148,172,097.64 |    0.2% |      0.01 | `SHA256_32b using the 'standard' SHA256 implementation`
  |                4.90 |      204,156,289.96 |    0.1% |      0.01 | `SHA256_32b using the 'sse4(1way),sse41(4way)' SHA256 implementation`
  |                4.90 |      204,101,274.22 |    0.1% |      0.01 | `SHA256_32b using the 'sse4(1way),sse41(4way),avx2(8way)' SHA256 implementation`
  |                1.70 |      589,052,595.35 |    0.4% |      0.01 | `SHA256_32b using the 'x86_shani(1way,2way)' SHA256 implementation`
  |                2.21 |      453,441,736.14 |    1.0% |      0.02 | `SHA3_256_1M`
  |                1.92 |      521,807,101.48 |    1.0% |      0.02 | `SHA512`
  ```

  - on macOS (M1):
  ```
  % ./src/bench/bench_bitcoin -filter=SHA.\*
  Using the 'arm_shani(1way,2way)' SHA256 implementation

  |             ns/byte |              byte/s |    err% |     total | benchmark
  |--------------------:|--------------------:|--------:|----------:|:----------
  |                1.36 |      737,644,274.00 |    0.6% |      0.02 | `SHA1`
  |                3.08 |      324,556,777.15 |    0.2% |      0.03 | `SHA256 using the 'standard' SHA256 implementation`
  |                0.45 |    2,198,104,135.18 |    0.3% |      0.01 | `SHA256 using the 'arm_shani(1way,2way)' SHA256 implementation`
  |                8.84 |      113,131,299.18 |    0.0% |      0.01 | `SHA256D64_1024 using the 'standard' SHA256 implementation`
  |                0.94 |    1,059,406,239.36 |    0.0% |      0.01 | `SHA256D64_1024 using the 'arm_shani(1way,2way)' SHA256 implementation`
  |                6.17 |      162,050,659.51 |    0.2% |      0.01 | `SHA256_32b using the 'standard' SHA256 implementation`
  |                1.15 |      866,637,155.98 |    0.0% |      0.01 | `SHA256_32b using the 'arm_shani(1way,2way)' SHA256 implementation`
  |                1.69 |      592,636,491.59 |    0.2% |      0.02 | `SHA3_256_1M`
  |                1.89 |      528,785,775.66 |    0.0% |      0.02 | `SHA512`
  ```

  Found it useful, while working on https://github.com/bitcoin/bitcoin/pull/24773.

ACKs for top commit:
  martinus:
    ACK ce6df7df9bab2405cfe7d6e382f5682cf30de476. I would have created a helper function in the test to avoid the code duplication for each test, but that's just me nitpicking. Here are results from my Ryzen 7950X, with `./src/bench/bench_bitcoin -filter="SHA256.*" -min-time=1000`:
  MarcoFalke:
    review ACK ce6df7df9bab2405cfe7d6e382f5682cf30de476 🏵
  sipa:
    ACK ce6df7df9bab2405cfe7d6e382f5682cf30de476

Tree-SHA512: e3de50e11b9a3a0d1e05583786041d4dc9afa2022e2115d75d6d1f63b11f62f6336f093001e53a631431d558c4dae29c596755c9e2d6aa78c382270116cc1f7f

3 files changed, 163 insertions, 23 deletions

diff --git a/src/bench/crypto_hash.cpp b/src/bench/crypto_hash.cpp
index cf8d807d7b..1685a120b4 100644
--- a/src/bench/crypto_hash.cpp
+++ b/src/bench/crypto_hash.cpp
@@ -13,6 +13,7 @@
 #include <crypto/siphash.h>
 #include <hash.h>
 #include <random.h>
+#include <tinyformat.h>
 #include <uint256.h>
 
 /* Number of bytes to hash per iteration */
@@ -36,13 +37,48 @@ static void SHA1(benchmark::Bench& bench)
     });
 }
 
-static void SHA256(benchmark::Bench& bench)
+static void SHA256_STANDARD(benchmark::Bench& bench)
 {
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::STANDARD)));
     uint8_t hash[CSHA256::OUTPUT_SIZE];
     std::vector<uint8_t> in(BUFFER_SIZE,0);
     bench.batch(in.size()).unit("byte").run([&] {
         CSHA256().Write(in.data(), in.size()).Finalize(hash);
     });
+    SHA256AutoDetect();
+}
+
+static void SHA256_SSE4(benchmark::Bench& bench)
+{
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::USE_SSE4)));
+    uint8_t hash[CSHA256::OUTPUT_SIZE];
+    std::vector<uint8_t> in(BUFFER_SIZE,0);
+    bench.batch(in.size()).unit("byte").run([&] {
+        CSHA256().Write(in.data(), in.size()).Finalize(hash);
+    });
+    SHA256AutoDetect();
+}
+
+static void SHA256_AVX2(benchmark::Bench& bench)
+{
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::USE_SSE4_AND_AVX2)));
+    uint8_t hash[CSHA256::OUTPUT_SIZE];
+    std::vector<uint8_t> in(BUFFER_SIZE,0);
+    bench.batch(in.size()).unit("byte").run([&] {
+        CSHA256().Write(in.data(), in.size()).Finalize(hash);
+    });
+    SHA256AutoDetect();
+}
+
+static void SHA256_SHANI(benchmark::Bench& bench)
+{
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::USE_SSE4_AND_SHANI)));
+    uint8_t hash[CSHA256::OUTPUT_SIZE];
+    std::vector<uint8_t> in(BUFFER_SIZE,0);
+    bench.batch(in.size()).unit("byte").run([&] {
+        CSHA256().Write(in.data(), in.size()).Finalize(hash);
+    });
+    SHA256AutoDetect();
 }
 
 static void SHA3_256_1M(benchmark::Bench& bench)
@@ -54,22 +90,92 @@ static void SHA3_256_1M(benchmark::Bench& bench)
     });
 }
 
-static void SHA256_32b(benchmark::Bench& bench)
+static void SHA256_32b_STANDARD(benchmark::Bench& bench)
+{
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::STANDARD)));
+    std::vector<uint8_t> in(32,0);
+    bench.batch(in.size()).unit("byte").run([&] {
+        CSHA256()
+            .Write(in.data(), in.size())
+            .Finalize(in.data());
+    });
+    SHA256AutoDetect();
+}
+
+static void SHA256_32b_SSE4(benchmark::Bench& bench)
+{
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::USE_SSE4)));
+    std::vector<uint8_t> in(32,0);
+    bench.batch(in.size()).unit("byte").run([&] {
+        CSHA256()
+            .Write(in.data(), in.size())
+            .Finalize(in.data());
+    });
+    SHA256AutoDetect();
+}
+
+static void SHA256_32b_AVX2(benchmark::Bench& bench)
+{
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::USE_SSE4_AND_AVX2)));
+    std::vector<uint8_t> in(32,0);
+    bench.batch(in.size()).unit("byte").run([&] {
+        CSHA256()
+            .Write(in.data(), in.size())
+            .Finalize(in.data());
+    });
+    SHA256AutoDetect();
+}
+
+static void SHA256_32b_SHANI(benchmark::Bench& bench)
 {
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::USE_SSE4_AND_SHANI)));
     std::vector<uint8_t> in(32,0);
     bench.batch(in.size()).unit("byte").run([&] {
         CSHA256()
             .Write(in.data(), in.size())
             .Finalize(in.data());
     });
+    SHA256AutoDetect();
+}
+
+static void SHA256D64_1024_STANDARD(benchmark::Bench& bench)
+{
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::STANDARD)));
+    std::vector<uint8_t> in(64 * 1024, 0);
+    bench.batch(in.size()).unit("byte").run([&] {
+        SHA256D64(in.data(), in.data(), 1024);
+    });
+    SHA256AutoDetect();
+}
+
+static void SHA256D64_1024_SSE4(benchmark::Bench& bench)
+{
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::USE_SSE4)));
+    std::vector<uint8_t> in(64 * 1024, 0);
+    bench.batch(in.size()).unit("byte").run([&] {
+        SHA256D64(in.data(), in.data(), 1024);
+    });
+    SHA256AutoDetect();
+}
+
+static void SHA256D64_1024_AVX2(benchmark::Bench& bench)
+{
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::USE_SSE4_AND_AVX2)));
+    std::vector<uint8_t> in(64 * 1024, 0);
+    bench.batch(in.size()).unit("byte").run([&] {
+        SHA256D64(in.data(), in.data(), 1024);
+    });
+    SHA256AutoDetect();
 }
 
-static void SHA256D64_1024(benchmark::Bench& bench)
+static void SHA256D64_1024_SHANI(benchmark::Bench& bench)
 {
+    bench.name(strprintf("%s using the '%s' SHA256 implementation", __func__, SHA256AutoDetect(sha256_implementation::USE_SSE4_AND_SHANI)));
     std::vector<uint8_t> in(64 * 1024, 0);
     bench.batch(in.size()).unit("byte").run([&] {
         SHA256D64(in.data(), in.data(), 1024);
     });
+    SHA256AutoDetect();
 }
 
 static void SHA512(benchmark::Bench& bench)
@@ -152,13 +258,22 @@ static void MuHashPrecompute(benchmark::Bench& bench)
 
 BENCHMARK(BenchRIPEMD160, benchmark::PriorityLevel::HIGH);
 BENCHMARK(SHA1, benchmark::PriorityLevel::HIGH);
-BENCHMARK(SHA256, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256_STANDARD, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256_SSE4, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256_AVX2, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256_SHANI, benchmark::PriorityLevel::HIGH);
 BENCHMARK(SHA512, benchmark::PriorityLevel::HIGH);
 BENCHMARK(SHA3_256_1M, benchmark::PriorityLevel::HIGH);
 
-BENCHMARK(SHA256_32b, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256_32b_STANDARD, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256_32b_SSE4, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256_32b_AVX2, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256_32b_SHANI, benchmark::PriorityLevel::HIGH);
 BENCHMARK(SipHash_32b, benchmark::PriorityLevel::HIGH);
-BENCHMARK(SHA256D64_1024, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256D64_1024_STANDARD, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256D64_1024_SSE4, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256D64_1024_AVX2, benchmark::PriorityLevel::HIGH);
+BENCHMARK(SHA256D64_1024_SHANI, benchmark::PriorityLevel::HIGH);
 BENCHMARK(FastRandom_32bit, benchmark::PriorityLevel::HIGH);
 BENCHMARK(FastRandom_1bit, benchmark::PriorityLevel::HIGH);
 
diff --git a/src/crypto/sha256.cpp b/src/crypto/sha256.cpp
index a4eef36480..5eacaa44e1 100644
--- a/src/crypto/sha256.cpp
+++ b/src/crypto/sha256.cpp
@@ -579,9 +579,15 @@ bool AVXEnabled()
 } // namespace
 
 
-std::string SHA256AutoDetect()
+std::string SHA256AutoDetect(sha256_implementation::UseImplementation use_implementation)
 {
     std::string ret = "standard";
+    Transform = sha256::Transform;
+    TransformD64 = sha256::TransformD64;
+    TransformD64_2way = nullptr;
+    TransformD64_4way = nullptr;
+    TransformD64_8way = nullptr;
+
 #if defined(USE_ASM) && defined(HAVE_GETCPUID)
     bool have_sse4 = false;
     bool have_xsave = false;
@@ -592,7 +598,9 @@ std::string SHA256AutoDetect()
 
     uint32_t eax, ebx, ecx, edx;
     GetCPUID(1, 0, eax, ebx, ecx, edx);
-    have_sse4 = (ecx >> 19) & 1;
+    if (use_implementation & sha256_implementation::USE_SSE4) {
+        have_sse4 = (ecx >> 19) & 1;
+    }
     have_xsave = (ecx >> 27) & 1;
     have_avx = (ecx >> 28) & 1;
     if (have_xsave && have_avx) {
@@ -600,8 +608,12 @@ std::string SHA256AutoDetect()
     }
     if (have_sse4) {
         GetCPUID(7, 0, eax, ebx, ecx, edx);
-        have_avx2 = (ebx >> 5) & 1;
-        have_x86_shani = (ebx >> 29) & 1;
+        if (use_implementation & sha256_implementation::USE_AVX2) {
+            have_avx2 = (ebx >> 5) & 1;
+        }
+        if (use_implementation & sha256_implementation::USE_SHANI) {
+            have_x86_shani = (ebx >> 29) & 1;
+        }
     }
 
 #if defined(ENABLE_X86_SHANI) && !defined(BUILD_BITCOIN_INTERNAL)
@@ -637,27 +649,28 @@ std::string SHA256AutoDetect()
 
 #if defined(ENABLE_ARM_SHANI) && !defined(BUILD_BITCOIN_INTERNAL)
     bool have_arm_shani = false;
-
+    if (use_implementation & sha256_implementation::USE_SHANI) {
 #if defined(__linux__)
 #if defined(__arm__) // 32-bit
-    if (getauxval(AT_HWCAP2) & HWCAP2_SHA2) {
-        have_arm_shani = true;
-    }
+        if (getauxval(AT_HWCAP2) & HWCAP2_SHA2) {
+            have_arm_shani = true;
+        }
 #endif
 #if defined(__aarch64__) // 64-bit
-    if (getauxval(AT_HWCAP) & HWCAP_SHA2) {
-        have_arm_shani = true;
-    }
+        if (getauxval(AT_HWCAP) & HWCAP_SHA2) {
+            have_arm_shani = true;
+        }
 #endif
 #endif
 
 #if defined(MAC_OSX)
-    int val = 0;
-    size_t len = sizeof(val);
-    if (sysctlbyname("hw.optional.arm.FEAT_SHA256", &val, &len, nullptr, 0) == 0) {
-        have_arm_shani = val != 0;
-    }
+        int val = 0;
+        size_t len = sizeof(val);
+        if (sysctlbyname("hw.optional.arm.FEAT_SHA256", &val, &len, nullptr, 0) == 0) {
+            have_arm_shani = val != 0;
+        }
 #endif
+    }
 
     if (have_arm_shani) {
         Transform = sha256_arm_shani::Transform;
diff --git a/src/crypto/sha256.h b/src/crypto/sha256.h
index 7625508665..b1348631d3 100644
--- a/src/crypto/sha256.h
+++ b/src/crypto/sha256.h
@@ -26,10 +26,22 @@ public:
     CSHA256& Reset();
 };
 
+namespace sha256_implementation {
+enum UseImplementation : uint8_t {
+    STANDARD = 0,
+    USE_SSE4 = 1 << 0,
+    USE_AVX2 = 1 << 1,
+    USE_SHANI = 1 << 2,
+    USE_SSE4_AND_AVX2 = USE_SSE4 | USE_AVX2,
+    USE_SSE4_AND_SHANI = USE_SSE4 | USE_SHANI,
+    USE_ALL = USE_SSE4 | USE_AVX2 | USE_SHANI,
+};
+}
+
 /** Autodetect the best available SHA256 implementation.
  *  Returns the name of the implementation.
  */
-std::string SHA256AutoDetect();
+std::string SHA256AutoDetect(sha256_implementation::UseImplementation use_implementation = sha256_implementation::USE_ALL);
 
 /** Compute multiple double-SHA256's of 64-byte blobs.
  *  output:  pointer to a blocks*32 byte output buffer