[PATCH] D25277: [test-suite] Run FP tests twice with -ffp-contract=off/on

[Sebastian Pop via llvm-commits]

sebpop added a comment.

In https://reviews.llvm.org/D25277#562503, @MatzeB wrote:

> Ugh, I don't like the added complexity and disk space bloat.
>
> Is there no way to summarize/hash/compress a list of floating point numbers in a way that we can compare them against a reference output and at the same time allow some fluctuations in floatingpoint accuracy?


If those tests were written with testing with FP_TOLERANCE in mind, we won't have all those problems.
In the case of Polybench in general and specifically for bicg and mvt, the tests are dumping 4000 by 4000 double matrices, which is ridiculous, all Polybench should be restricted to only compare against a an exact reference hash.  These tests (and maybe others) should only run under -ffp-contract=off.


https://reviews.llvm.org/D25277