[all-commits] [llvm/llvm-project] 8a52bd: [AMDGPU] Only select VOP3 forms of VOP2 instructions

[Jay Foad via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 8a52bd82e36855b3ad842f2535d0c78a97db55dc
      https://github.com/llvm/llvm-project/commit/8a52bd82e36855b3ad842f2535d0c78a97db55dc
  Author: Jay Foad <jay.foad at amd.com>
  Date:   2021-11-24 (Wed, 24 Nov 2021)

  Changed paths:
    M llvm/lib/Target/AMDGPU/SIInstrInfo.td
    M llvm/test/CodeGen/AMDGPU/ashr.v2i16.ll
    M llvm/test/CodeGen/AMDGPU/bfe-patterns.ll
    M llvm/test/CodeGen/AMDGPU/commute-shifts.ll
    M llvm/test/CodeGen/AMDGPU/ctpop16.ll
    M llvm/test/CodeGen/AMDGPU/extract-lowbits.ll
    M llvm/test/CodeGen/AMDGPU/flat-scratch.ll
    M llvm/test/CodeGen/AMDGPU/idot8s.ll
    M llvm/test/CodeGen/AMDGPU/idot8u.ll
    M llvm/test/CodeGen/AMDGPU/inline-asm.ll
    M llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll
    M llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wqm.demote.ll
    M llvm/test/CodeGen/AMDGPU/lshr.v2i16.ll
    M llvm/test/CodeGen/AMDGPU/sdwa-peephole.ll
    M llvm/test/CodeGen/AMDGPU/select-constant-xor.ll
    M llvm/test/CodeGen/AMDGPU/sext-in-reg.ll
    M llvm/test/CodeGen/AMDGPU/shl.ll
    M llvm/test/CodeGen/AMDGPU/shl.v2i16.ll
    M llvm/test/CodeGen/AMDGPU/ssubsat.ll
    M llvm/test/CodeGen/AMDGPU/stack-realign.ll
    M llvm/test/CodeGen/AMDGPU/wwm-reserved-spill.ll

  Log Message:
  -----------
  [AMDGPU] Only select VOP3 forms of VOP2 instructions

Change VOP_PAT_GEN to default to not generating an instruction selection
pattern for the VOP2 (e32) form of an instruction, only for the VOP3
(e64) form. This allows SIFoldOperands maximum freedom to fold copies
into the operands of an instruction, before SIShrinkInstructions tries
to shrink it back to the smaller encoding.

This affects the following VOP2 instructions:
v_min_i32
v_max_i32
v_min_u32
v_max_u32
v_and_b32
v_or_b32
v_xor_b32
v_lshr_b32
v_ashr_i32
v_lshl_b32

A further cleanup could simplify or remove VOP_PAT_GEN, since its
optional second argument is never used.

Differential Revision: https://reviews.llvm.org/D114252