llvm/test/CodeGen/AMDGPU/collapse-endcf2.mir

# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mtriple=amdgcn -verify-machineinstrs -run-pass=si-optimize-exec-masking-pre-ra %s -o - | FileCheck -check-prefix=GCN %s

--- |
  define amdgpu_kernel void @call_no_explicit_exec_dependency () {
    unreachable
  }

  declare void @func()

...

# Call should be assumed to read exec
---
name: call_no_explicit_exec_dependency
tracksRegLiveness: true
liveins:
  - { reg: '$vgpr0', virtual-reg: '%0' }
  - { reg: '$sgpr0_sgpr1', virtual-reg: '%1' }
machineFunctionInfo:
  isEntryFunction: true
body:             |
  ; GCN-LABEL: name: call_no_explicit_exec_dependency
  ; GCN: bb.0:
  ; GCN-NEXT:   successors: %bb.1(0x40000000), %bb.4(0x40000000)
  ; GCN-NEXT:   liveins: $vgpr0, $sgpr0_sgpr1
  ; GCN-NEXT: {{  $}}
  ; GCN-NEXT:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr0_sgpr1
  ; GCN-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GCN-NEXT:   [[V_CMP_LT_U32_e64_:%[0-9]+]]:sreg_64 = V_CMP_LT_U32_e64 1, [[COPY1]], implicit $exec
  ; GCN-NEXT:   [[COPY2:%[0-9]+]]:sreg_64 = COPY $exec, implicit-def $exec
  ; GCN-NEXT:   [[S_AND_B64_:%[0-9]+]]:sreg_64 = S_AND_B64 [[COPY2]], [[V_CMP_LT_U32_e64_]], implicit-def dead $scc
  ; GCN-NEXT:   $exec = S_MOV_B64_term [[S_AND_B64_]]
  ; GCN-NEXT:   S_CBRANCH_EXECZ %bb.4, implicit $exec
  ; GCN-NEXT:   S_BRANCH %bb.1
  ; GCN-NEXT: {{  $}}
  ; GCN-NEXT: bb.1:
  ; GCN-NEXT:   successors: %bb.2(0x40000000), %bb.3(0x40000000)
  ; GCN-NEXT: {{  $}}
  ; GCN-NEXT:   undef %5.sub0_sub1:sgpr_128 = S_LOAD_DWORDX2_IMM [[COPY]], 9, 0 :: (dereferenceable invariant load (s64), align 4, addrspace 4)
  ; GCN-NEXT:   undef %6.sub0:vreg_64 = V_LSHLREV_B32_e32 2, [[COPY1]], implicit $exec
  ; GCN-NEXT:   %6.sub1:vreg_64 = V_MOV_B32_e32 0, implicit $exec
  ; GCN-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY %5.sub1
  ; GCN-NEXT:   undef %8.sub0:vreg_64, %9:sreg_64_xexec = V_ADD_CO_U32_e64 %5.sub0, %6.sub0, 0, implicit $exec
  ; GCN-NEXT:   %8.sub1:vreg_64, dead %10:sreg_64_xexec = V_ADDC_U32_e64 0, [[COPY3]], %9, 0, implicit $exec
  ; GCN-NEXT:   %5.sub3:sgpr_128 = S_MOV_B32 61440
  ; GCN-NEXT:   %5.sub2:sgpr_128 = S_MOV_B32 0
  ; GCN-NEXT:   BUFFER_STORE_DWORD_ADDR64 %6.sub1, %6, %5, 0, 0, 0, 0, implicit $exec :: (store (s32), addrspace 1)
  ; GCN-NEXT:   [[V_CMP_NE_U32_e64_:%[0-9]+]]:sreg_64 = V_CMP_NE_U32_e64 2, [[COPY1]], implicit $exec
  ; GCN-NEXT:   [[COPY4:%[0-9]+]]:sreg_64 = COPY $exec, implicit-def $exec
  ; GCN-NEXT:   [[S_AND_B64_1:%[0-9]+]]:sreg_64 = S_AND_B64 [[COPY4]], [[V_CMP_NE_U32_e64_]], implicit-def dead $scc
  ; GCN-NEXT:   $exec = S_MOV_B64_term [[S_AND_B64_1]]
  ; GCN-NEXT:   S_CBRANCH_EXECZ %bb.3, implicit $exec
  ; GCN-NEXT:   S_BRANCH %bb.2
  ; GCN-NEXT: {{  $}}
  ; GCN-NEXT: bb.2:
  ; GCN-NEXT:   successors: %bb.3(0x80000000)
  ; GCN-NEXT: {{  $}}
  ; GCN-NEXT:   %5.sub0:sgpr_128 = COPY %5.sub2
  ; GCN-NEXT:   %5.sub1:sgpr_128 = COPY %5.sub2
  ; GCN-NEXT:   [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 1, implicit $exec
  ; GCN-NEXT:   BUFFER_STORE_DWORD_ADDR64 [[V_MOV_B32_e32_]], %8, %5, 0, 4, 0, 0, implicit $exec :: (store (s32), addrspace 1)
  ; GCN-NEXT: {{  $}}
  ; GCN-NEXT: bb.3:
  ; GCN-NEXT:   successors: %bb.4(0x80000000)
  ; GCN-NEXT: {{  $}}
  ; GCN-NEXT:   $exec = S_OR_B64 $exec, [[COPY4]], implicit-def $scc
  ; GCN-NEXT:   [[DEF:%[0-9]+]]:sreg_64 = IMPLICIT_DEF
  ; GCN-NEXT:   dead %16:sreg_64 = SI_CALL [[DEF]], @func, csr_amdgpu
  ; GCN-NEXT: {{  $}}
  ; GCN-NEXT: bb.4:
  ; GCN-NEXT:   $exec = S_OR_B64 $exec, [[COPY2]], implicit-def $scc
  ; GCN-NEXT:   [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 3, implicit $exec
  ; GCN-NEXT:   [[V_MOV_B32_e32_2:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
  ; GCN-NEXT:   $m0 = S_MOV_B32 -1
  ; GCN-NEXT:   DS_WRITE_B32 [[V_MOV_B32_e32_2]], [[V_MOV_B32_e32_1]], 0, 0, implicit $m0, implicit $exec :: (store (s32), addrspace 3)
  ; GCN-NEXT:   S_ENDPGM 0
  bb.0:
    successors: %bb.1, %bb.4
    liveins: $vgpr0, $sgpr0_sgpr1

    %1:sgpr_64 = COPY $sgpr0_sgpr1
    %0:vgpr_32 = COPY $vgpr0
    %2:sreg_64 = V_CMP_LT_U32_e64 1, %0, implicit $exec
    %3:sreg_64 = COPY $exec, implicit-def $exec
    %4:sreg_64 = S_AND_B64 %3, %2, implicit-def dead $scc
    $exec = S_MOV_B64_term %4
    S_CBRANCH_EXECZ %bb.4, implicit $exec
    S_BRANCH %bb.1

  bb.1:
    successors: %bb.2, %bb.3

    undef %5.sub0_sub1:sgpr_128 = S_LOAD_DWORDX2_IMM %1, 9, 0 :: (dereferenceable invariant load (s64), align 4, addrspace 4)
    undef %6.sub0:vreg_64 = V_LSHLREV_B32_e32 2, %0, implicit $exec
    %6.sub1:vreg_64 = V_MOV_B32_e32 0, implicit $exec
    %7:vgpr_32 = COPY %5.sub1
    undef %8.sub0:vreg_64, %9:sreg_64_xexec = V_ADD_CO_U32_e64 %5.sub0, %6.sub0, 0, implicit $exec
    %8.sub1:vreg_64, dead %10:sreg_64_xexec = V_ADDC_U32_e64 0, %7, %9, 0, implicit $exec
    %5.sub3:sgpr_128 = S_MOV_B32 61440
    %5.sub2:sgpr_128 = S_MOV_B32 0
    BUFFER_STORE_DWORD_ADDR64 %6.sub1, %6, %5, 0, 0, 0, 0, implicit $exec :: (store (s32), addrspace 1)
    %11:sreg_64 = V_CMP_NE_U32_e64 2, %0, implicit $exec
    %12:sreg_64 = COPY $exec, implicit-def $exec
    %13:sreg_64 = S_AND_B64 %12, %11, implicit-def dead $scc
    $exec = S_MOV_B64_term %13
    S_CBRANCH_EXECZ %bb.3, implicit $exec
    S_BRANCH %bb.2

  bb.2:
    %5.sub0:sgpr_128 = COPY %5.sub2
    %5.sub1:sgpr_128 = COPY %5.sub2
    %14:vgpr_32 = V_MOV_B32_e32 1, implicit $exec
    BUFFER_STORE_DWORD_ADDR64 %14, %8, %5, 0, 4, 0, 0, implicit $exec :: (store (s32), addrspace 1)

  bb.3:
    $exec = S_OR_B64 $exec, %12,  implicit-def $scc
    %20:sreg_64 = IMPLICIT_DEF
    %21:sreg_64 = SI_CALL %20, @func, csr_amdgpu

  bb.4:
    $exec = S_OR_B64 $exec, %3, implicit-def $scc
    %17:vgpr_32 = V_MOV_B32_e32 3, implicit $exec
    %18:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
    $m0 = S_MOV_B32 -1
    DS_WRITE_B32 %18, %17, 0, 0, implicit $m0, implicit $exec :: (store (s32), addrspace 3)
    S_ENDPGM 0

...