[PATCH] D138141: [amdgpu] Reimplement LDS lowering

Wed Nov 23 15:53:49 PST 2022

JonChesterfield added a comment.

Something of a speed bump. https://reviews.llvm.org/D131246 breaks the table and hybrid implementation here. Fails check-openmp reliably. Partial stack trace.

   #9 0x0000000000c0a341 (anonymous namespace)::SIMCCodeEmitter::encodeInstruction(llvm::MCInst const&, llvm::raw_ostream&, llvm::SmallVectorImpl<llvm::MCFixup>&, llvm::MCSubtargetInfo const&)
  #10 0x0000000000e27a57 llvm::MCELFStreamer::emitInstToData(llvm::MCInst const&, llvm::MCSubtargetInfo const&)
  #11 0x0000000000e31964 llvm::MCObjectStreamer::emitInstruction(llvm::MCInst const&, llvm::MCSubtargetInfo const&)
  #12 0x000000000082d8f0 llvm::AMDGPUAsmPrinter::emitInstruction(llvm::MachineInstr const*)
  #13 0x00000000017a3df7 llvm::AsmPrinter::emitFunctionBody()
  #14 0x000000000081cb38 llvm::AMDGPUAsmPrinter::runOnMachineFunction(llvm::MachineFunction&)

I'm unclear whether this is a latent problem exposed by D131246 <https://reviews.llvm.org/D131246> or a novel failure mode. Since this patch works as written - a benefit from defaulting to module - I'm hoping we can land this, then fix that lowering bug, then change the default to hybrid. Sound reasonable?

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D138141/new/

https://reviews.llvm.org/D138141