[PATCH] D112488: AMDGPU: Assume all amdhsa kernarg passed implicit arguments by default

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp:659
+
+  // Assume all implicit inputs are used by default
+  return AMDGPU::getIntegerAttribute(F, "amdgpu-implicitarg-num-bytes", 56);
----------------
rampitec wrote:
> rampitec wrote:
> > arsenm wrote:
> > > rampitec wrote:
> > > > It is not immediately obvious we have a HSA kernel at this point.
> > > This is only meaningfully called with a kernel
> > With MESA ruled out does that mean only a HSA kernel remains? Maybe add at least an assert?
> Is this also true for SPIR_KERNEL?
We never distinguished them (although SPIR_KERNEL should never really be seen, but nobody ever implemented SPIR correctly)


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D112488/new/

https://reviews.llvm.org/D112488