llvm/test/CodeGen/AMDGPU/subvector-test.mir

# RUN: llc -mtriple=amdgpu10.10 -start-before=greedy -verify-machineinstrs -o - %s | FileCheck -check-prefix=GCN %s
# RUN: llc -mtriple=amdgpu11.00 -start-before=greedy -verify-machineinstrs -o - %s | FileCheck -check-prefix=GCN %s
...
# GCN-LABEL: {{^}}"subvector-basic-bb"
# GCN: s_subvector_loop_begin [[RS:s[0-9]]], .LBB0_2
# GCN: s_subvector_loop_end [[RS]], .LBB0_1
name:            subvector-basic-bb
tracksRegLiveness: true
machineFunctionInfo:
  scratchRSrcReg: $sgpr0_sgpr1_sgpr2_sgpr3
  frameOffsetReg: $sgpr5
  stackPtrOffsetReg: $sgpr32
body:             |
  bb.0:
    liveins: $sgpr0_sgpr1
    successors: %bb.1, %bb.2

    %1:sgpr_64 = COPY $sgpr0_sgpr1
    %4:sgpr_128 = S_LOAD_DWORDX4_IMM %1, 36, 0
    %11:sreg_64_xexec = S_LOAD_DWORDX2_IMM %4.sub2_sub3, 0, 0
    undef %15.sub0:vreg_64 = COPY %4.sub0
    %15.sub1:vreg_64 = COPY %4.sub1
    %16:vgpr_32 = COPY %1.sub0
    S_SUBVECTOR_LOOP_BEGIN %bb.2, undef %19:sreg_32, implicit-def $exec, implicit $exec, implicit-def %19

  bb.1:
    successors: %bb.1, %bb.2

    %14:sreg_32_xm0 = S_ADD_I32 %11.sub0, %11.sub1, implicit-def dead $scc
    %16:vgpr_32 = COPY %14
    S_SUBVECTOR_LOOP_END %bb.1, %19:sreg_32, implicit-def $exec, implicit $exec, implicit-def %19

  bb.2:

    GLOBAL_STORE_DWORD %15, %16, 0, 0, implicit $exec
    S_ENDPGM 0
...