From e4a6e97b91da8109840b59697a3b77f93ade956c Mon Sep 17 00:00:00 2001
From: Snappy Team <no-reply@google.com>
Date: Mon, 16 Nov 2020 15:10:23 +0000
Subject: [PATCH] Extend validate benchmarks over all types and also add a
 medley for validation.

I also made the compression happen only once per benchmark. This way we get a cleaner measurement of #branch-misses using "perf stat". Compression suffers naturally from a large number of branch misses which was polluting the measurements.

This showed that with the new decompression the branch misses is actually much lower then initially reported, only .2% and very stable, ie. doesn't really fluctuate with how you execute the benchmarks.

PiperOrigin-RevId: 342628576
---
 snappy_unittest.cc | 51 ++++++++++++++++++++++++++++++++--------------
 1 file changed, 36 insertions(+), 15 deletions(-)

diff --git a/snappy_unittest.cc b/snappy_unittest.cc
index d433545..0e207db 100644
--- a/snappy_unittest.cc
+++ b/snappy_unittest.cc
@@ -1284,27 +1284,33 @@ static void BM_UFlat(int iters, int arg) {
 }
 BENCHMARK(BM_UFlat)->DenseRange(0, ARRAYSIZE(files) - 1);
 
-static void BM_UFlatMedley(testing::benchmark::State& state) {
-  constexpr int kFiles = ARRAYSIZE(files);
+struct SourceFiles {
+  SourceFiles() {
+    for (int i = 0; i < kFiles; i++) {
+      std::string contents =
+          ReadTestDataFile(files[i].filename, files[i].size_limit);
+      max_size = std::max(max_size, contents.size());
+      sizes[i] = contents.size();
+      snappy::Compress(contents.data(), contents.size(), &zcontents[i]);
+    }
+  }
+  static constexpr int kFiles = ARRAYSIZE(files);
   std::string zcontents[kFiles];
   size_t sizes[kFiles];
   size_t max_size = 0;
-  for (int i = 0; i < kFiles; i++) {
-    std::string contents =
-        ReadTestDataFile(files[i].filename, files[i].size_limit);
-    max_size = std::max(max_size, contents.size());
-    sizes[i] = contents.size();
-    snappy::Compress(contents.data(), contents.size(), &zcontents[i]);
-  }
+};
 
-  std::vector<char> dst(max_size);
+static void BM_UFlatMedley(testing::benchmark::State& state) {
+  static const SourceFiles* const source = new SourceFiles();
+
+  std::vector<char> dst(source->max_size);
 
   size_t processed = 0;
   for (auto s : state) {
-    for (int i = 0; i < kFiles; i++) {
-      CHECK(snappy::RawUncompress(zcontents[i].data(), zcontents[i].size(),
-                                  dst.data()));
-      processed += sizes[i];
+    for (int i = 0; i < SourceFiles::kFiles; i++) {
+      CHECK(snappy::RawUncompress(source->zcontents[i].data(),
+                                  source->zcontents[i].size(), dst.data()));
+      processed += source->sizes[i];
     }
   }
   SetBenchmarkBytesProcessed(processed);
@@ -1332,7 +1338,22 @@ static void BM_UValidate(int iters, int arg) {
   }
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_UValidate)->DenseRange(0, 4);
+BENCHMARK(BM_UValidate)->DenseRange(0, ARRAYSIZE(files) - 1);
+
+static void BM_UValidateMedley(testing::benchmark::State& state) {
+  static const SourceFiles* const source = new SourceFiles();
+
+  size_t processed = 0;
+  for (auto s : state) {
+    for (int i = 0; i < SourceFiles::kFiles; i++) {
+      CHECK(snappy::IsValidCompressedBuffer(source->zcontents[i].data(),
+                                            source->zcontents[i].size()));
+      processed += source->sizes[i];
+    }
+  }
+  SetBenchmarkBytesProcessed(processed);
+}
+BENCHMARK(BM_UValidateMedley);
 
 static void BM_UIOVec(int iters, int arg) {
   StopBenchmarkTiming();