[PATCH] D24593: Standford/Bubble sort code restructure

[Chandler Carruth via llvm-commits]

chandlerc requested changes to this revision.
chandlerc added inline comments.
This revision now requires changes to proceed.


> mehdi_amini wrote in Bubblesort.c:159
> It seems to me that you're creating a "new" benchmark by doing that.
> It does not seem legit to change a benchmark because you find it hard to analyze from within the compiler.
> 
> (That does not fit my definition of flaky either)

What Mehdi said.

This is not flaky. You don't like what a particular transform does to the performance here. You can't fix this by changing the benchmark.

You can add more benchmarks that show that a particular transform is *worth* the regression in this benchmark.

Or you can teach the transform to not regress the code pattern that this benchmark is testing. Which would seem the more likely solution.

No matter what, *changing the benchmark code is not the answer*.

https://reviews.llvm.org/D24593