[PATCH] D65423: [AMDGPU] Fix high occupancy calculation and print it

Stanislav Mekhanoshin via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Jul 29 17:24:11 PDT 2019


rampitec marked 2 inline comments as done.
rampitec added inline comments.


================
Comment at: lib/Target/AMDGPU/AMDGPUSubtarget.h:1322
+
+  unsigned getMaxWavesPerEU() const override { return 10; }
 };
----------------
rampitec wrote:
> arsenm wrote:
> > Should just set a field in the constructor rather than introducing a virtual 
> That is too early, constructor does not know it yet.
> Also look at the other functions of the same family, they are all virtual: getMaxWorkGroupsPerCU, getMinFlatWorkGroupSize, getMaxFlatWorkGroupSize, getMinWavesPerEU, getMaxWavesPerEU(unsigned FlatWorkGroupSize).
Well, OK, I still can do it. Just overwrite value in GCNSubtarget(). It looks a little weird because I still have to duplicate this function in the GCNSubtarget(), otherwise C++ has problems with name resolve. Either this or we will always have to specify AMDGPUSubtarget:: prefix.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D65423/new/

https://reviews.llvm.org/D65423





More information about the llvm-commits mailing list