[PATCH] [x86] generalize reassociation optimization in machine combiner to 2 instructions
mehdi.amini at apple.com
Mon Jun 22 14:26:54 PDT 2015
For the record: the test didn’t come from an X86 test, it is a simplified version of a real-world GPU shader.
> On Jun 22, 2015, at 11:48 AM, Gerolf Hoflehner <ghoflehner at apple.com> wrote:
> LGTM, but for compile time please add a FIXME before commit. What more evidence does it need? "On Mehdi's (hopefully degenerate for x86) test case from the r236031 post-commit thread, the compile-time increases from ~0.2 sec to 5.0 sec". However, in the current form the patch should have negligible ct impact in general.
> EMAIL PREFERENCES
More information about the llvm-commits