[libcxx-commits] [libcxx] [libc++] Optimize ranges::{for_each, for_each_n} for segmented iterators (PR #132896)

[Louis Dionne via libcxx-commits]

================
@@ -33,15 +35,19 @@ int main(int argc, char** argv) {
 
             for ([[maybe_unused]] auto _ : st) {
               benchmark::DoNotOptimize(c);
-              auto result = for_each(first, last, [](int& x) { x = std::clamp(x, 10, 100); });
+              auto result = for_each(first, last, [](ElemType& x) { x = std::clamp<ElemType>(x, 10, 100); });
               benchmark::DoNotOptimize(result);
             }
           })
           ->Arg(8)
           ->Arg(32)
           ->Arg(50) // non power-of-two
+          ->Arg(1024)
+          ->Arg(4096)
           ->Arg(8192)
-          ->Arg(1 << 20);
+          ->Arg(1 << 14)
+          ->Arg(1 << 16)
+          ->Arg(1 << 18);
----------------
ldionne wrote:

I believe it would be better to leave the old benchmark values in place. They are less comprehensive but we need to achieve a tradeoff between comprehensiveness and the time it takes to run these benchmarks.

https://github.com/llvm/llvm-project/pull/132896