[PATCH] D49563: [ARM] Add new target feature to fuse literal generation
Evandro Menezes via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Fri Jul 27 10:11:28 PDT 2018
evandro added inline comments.
================
Comment at: llvm/lib/Target/ARM/ARM.td:145
+// Fast execution of bottom and top halves of literal generation
+def FeatureFuseLiterals : SubtargetFeature<"fuse-literals", "HasFuseLiterals", "true",
+ "CPU fuses literal generation operations">;
----------------
fhahn wrote:
> evandro wrote:
> > fhahn wrote:
> > > FeatureFuseLiterals should be added to the cortex-a57 and cortex-a72 features according to the optimization guides. Can you also add a run lines with those CPUs to the test?
> > It's more complicated than that. I had to enable `FeatureUseMISched` as well for this to be effective. This is fine for Exynos, but it'd require more extensive testing on Cortex.
> Ah right! No worries then
Of course, Exynos needs this feature because there's no pipeline model for it in the ARM target.
Repository:
rL LLVM
https://reviews.llvm.org/D49563
More information about the llvm-commits
mailing list