[PATCH] D46365: AMDGPU: Separate R600 and GCN TableGen files

[Jan Vesely via Phabricator via llvm-commits]

jvesely added a comment.

I added the below snippet to check whether the caymanISA feature gets initialized correctly:

  @@ -415,7 +417,10 @@ R600Subtarget::R600Subtarget(const Triple &TT, StringRef GPU, StringRef FS,
     TLInfo(TM, initializeSubtargetDependencies(TT, GPU, FS)),
     DX10Clamp(false),
     InstrItins(getInstrItineraryForCPU(GPU)),
  -  AS (AMDGPU::getAMDGPUAS(TT)) { }
  +  AS (AMDGPU::getAMDGPUAS(TT)) {
  +  fprintf(stderr, "R600 FEATURE STRING: %s\n", FS.data());
  +  fprintf(stderr, "R600 Has Cayman ISA: %s\n", CaymanISA ? "YES" : "NO");
  +}

As expected it randomly on occasion printed:

  '-fp32-denormals' is not a recognized feature for this target (ignoring feature)
  '-fp32-denormals' is not a recognized feature for this target (ignoring feature)
  '-fp32-denormals' is not a recognized feature for this target (ignoring feature)
  R600 FEATURE STRING: -fp32-denormals
  R600 Has Cayman ISA: YES

running llc through valgrind produced flood of 'Conditional jump or move depends on uninitialised value(s)'
269 errors from 24 contexts. Initialzieng just CaymanISA in R600SUbtarget gets rid of most of them.


Repository:
  rL LLVM

https://reviews.llvm.org/D46365