[PATCH] D64207: [AMDGPU] DPP combiner: recognize identities for more opcodes

[Jay Foad via Phabricator via llvm-commits]

foad added a comment.

In D64207#1570526 <https://reviews.llvm.org/D64207#1570526>, @vpykhtin wrote:

> Right, this is hard to follow even for me :). 3rd operand is src1_modifiers, you can use a junk value for this to check whether the DPP combiner don't crash and don't combine it.


I still can't understand the instruction tables but according to my debugger, the third operand is clamp:

  (gdb) call OrigMI.dump()
    %10:vgpr_32 = V_ADD_U32_e64 %9:vgpr_32, %1:vgpr_32, 99, implicit $exec
  (gdb) p TII->getNamedOperand(OrigMI, AMDGPU::OpName::src0_modifiers)
  $5 = (llvm::MachineOperand *) 0x0
  (gdb) p TII->getNamedOperand(OrigMI, AMDGPU::OpName::src1_modifiers)
  $6 = (llvm::MachineOperand *) 0x0
  (gdb) p TII->getNamedOperand(OrigMI, AMDGPU::OpName::clamp)
  $7 = (llvm::MachineOperand *) 0x6d1eef0
  (gdb) p TII->getNamedOperand(OrigMI, AMDGPU::OpName::omod)
  $8 = (llvm::MachineOperand *) 0x0
  (gdb) call TII->getNamedOperand(OrigMI, AMDGPU::OpName::clamp)->dump()
  99


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D64207/new/

https://reviews.llvm.org/D64207