[libcxx-commits] [libcxx] [libc++] Fix the batch sized used in the std::gcd benchmark (PR #120618)

[via libcxx-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-libcxx

Author: Louis Dionne (ldionne)

<details>
<summary>Changes</summary>

Since that benchmark is testing n*n inputs, the batch size reported to GoogleBenchmark should be that amount. Otherwise, GoogleBenchmark reports the timing for calling std::gcd on the whole sequence, which is misleading.

---
Full diff: https://github.com/llvm/llvm-project/pull/120618.diff


1 Files Affected:

- (modified) libcxx/test/benchmarks/numeric/gcd.bench.cpp (+1-1) 


``````````diff
diff --git a/libcxx/test/benchmarks/numeric/gcd.bench.cpp b/libcxx/test/benchmarks/numeric/gcd.bench.cpp
index abbc7e9dd04f96..ca5fed59463a21 100644
--- a/libcxx/test/benchmarks/numeric/gcd.bench.cpp
+++ b/libcxx/test/benchmarks/numeric/gcd.bench.cpp
@@ -25,7 +25,7 @@ static std::array<T, 1000> generate(std::uniform_int_distribution<T> distributio
 
 static void bm_gcd_random(benchmark::State& state) {
   std::array data = generate<int>();
-  while (state.KeepRunningBatch(data.size()))
+  while (state.KeepRunningBatch(data.size() * data.size()))
     for (auto v0 : data)
       for (auto v1 : data)
         benchmark::DoNotOptimize(std::gcd(v0, v1));

``````````

</details>


https://github.com/llvm/llvm-project/pull/120618