[AMDGPU] Derive GCNSubtarget from MF to get overridden target features

[dstuttard]

Summary

AMDGPUAsmPrinter has a getSTI function that derives a GCNSubtarget from the
TM. However, this means that overridden target features are not detected and can
result in incorrect behaviour.

Switch to using STM which is a GCNSubtarget derived from the MF (used elsewhere
in the same function).

Change-Id: Ib6328ad667b7fcdc87e9c06344e59859207db9b0

[dstuttard]

Scott - you made some changes here most recently so adding you as the reviewer.

[scott.linder]

I don't quite understand what distinguishes the two versions of MCSubtargetInfo (getSTI() vs STM), but it appears a patch from Konstantin earlier this year changed this specific instance from STM.getFeatureBits() to getSTI(). Changing this back LGTM but I don't know what the rationale was in the first place, if it wasn't a typo, so I don't want to sign off without asking.

[dstuttard]

In D54301#1296006, @scott.linder wrote:

I don't quite understand what distinguishes the two versions of MCSubtargetInfo (getSTI() vs STM), but it appears a patch from Konstantin earlier this year changed this specific instance from STM.getFeatureBits() to getSTI(). Changing this back LGTM but I don't know what the rationale was in the first place, if it wasn't a typo, so I don't want to sign off without asking.

My assumption is that it was neater to use the getSTI() member function provided - but that get the GCNSubtarget from the TargetMachine rather than MachineFunction - in most cases I guess this doesn't matter - but in our case the MF has got subtarget features enabled that the TM version doesn't. I wonder if the getSTI should be changed to use the MF version and all references to STM moved over to use getSTI().

Like you I'm not 100% sure there isn't a reason to use the TargetMachine derived version - so I guess we need Konstantin to comment.

[arsenm]

Testcase?

[kzhuravl]

LGTM

llvm/trunk/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp

void AMDGPUAsmPrinter::getSIProgramInfo(SIProgramInfo &ProgInfo,
   if (MFI->getLDSSize() > static_cast<unsigned>(STM.getLocalMemorySize())) {
     LLVMContext &Ctx = MF.getFunction().getContext();
     DiagnosticInfoResourceLimit Diag(MF.getFunction(), "local memory",
                                      MFI->getLDSSize(), DS_Error);
     Ctx.diagnose(Diag);
   }
 
   ProgInfo.SGPRBlocks = IsaInfo::getNumSGPRBlocks(
-      getSTI(), ProgInfo.NumSGPRsForWavesPerEU);
+      &STM, ProgInfo.NumSGPRsForWavesPerEU);
   ProgInfo.VGPRBlocks = IsaInfo::getNumVGPRBlocks(
-      getSTI(), ProgInfo.NumVGPRsForWavesPerEU);
+      &STM, ProgInfo.NumVGPRsForWavesPerEU);
 
   // Update DebuggerWavefrontPrivateSegmentOffsetSGPR and
   // DebuggerPrivateSegmentBufferSGPR fields if "amdgpu-debugger-emit-prologue"
   // attribute was requested.
   if (STM.debuggerEmitPrologue()) {
     ProgInfo.DebuggerWavefrontPrivateSegmentOffsetSGPR =
       RI->getHWRegIndex(MFI->getScratchWaveOffsetReg());
     ProgInfo.DebuggerPrivateSegmentBufferSGPR =