[PATCH] D46992: [AMDGPU] Add perf hints to functions

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec added inline comments.


================
Comment at: lib/Target/AMDGPU/AMDGPUPerfHint.cpp:289
+  }
+  FIM[&F] = FI;
+}
----------------
rampitec wrote:
> arsenm wrote:
> > rampitec wrote:
> > > t-tye wrote:
> > > > Should this be done at the beginning of the visit to ensure will terminate for mutual recursive functions?
> > > Thank you!
> > Can you add a test for this case
> Only as opt test. BE will fail on recursion.
Actually since it is now an on-demand analysis I cannot do it even with opt. We will need to generally fix recursion handling in the BE, it is not a problem specific to this patch.


================
Comment at: lib/Target/AMDGPU/SIDefines.h:538-539
+
+#define ATTR_MEMBOUND_HINT       "amdgpu-hint-memory-bound"
+#define ATTR_WAVE_LIMITER_HINT   "amdgpu-hint-wave-limiter"
 } // End namespace llvm
----------------
arsenm wrote:
> Should be able to also remove this
It was removed when pass was converted to analysis. Please check the current patch.


https://reviews.llvm.org/D46992