diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -7599,8 +7599,19 @@
 }
 
 bool SIInstrInfo::isBasicBlockPrologue(const MachineInstr &MI) const {
-  return !MI.isTerminator() && MI.getOpcode() != AMDGPU::COPY &&
-         MI.modifiesRegister(AMDGPU::EXEC, &RI);
+  if (MI.isTerminator())
+    return false;
+  switch (MI.getOpcode()) {
+  case AMDGPU::COPY:
+  case AMDGPU::SI_IF:
+  case AMDGPU::SI_IF_BREAK:
+  case AMDGPU::SI_LOOP:
+  case AMDGPU::SI_ELSE:
+  case AMDGPU::SI_END_CF:
+    return false;
+  default:
+    return MI.modifiesRegister(AMDGPU::EXEC, &RI);
+  }
 }
 
 MachineInstrBuilder
diff --git a/llvm/test/CodeGen/AMDGPU/sink-after-control-flow.mir b/llvm/test/CodeGen/AMDGPU/sink-after-control-flow.mir
--- a/llvm/test/CodeGen/AMDGPU/sink-after-control-flow.mir
+++ b/llvm/test/CodeGen/AMDGPU/sink-after-control-flow.mir
@@ -3,6 +3,7 @@
 
 # Test that MachineSink pass respects block prologues when sinking instructions.
 # Specifically an instruction must not be sunk before exec mask manipulation.
+# With the exception that control flow pseudo instructions are not part of the prologue.
 
 ---
 name:            _amdgpu_hs_main
@@ -120,3 +121,133 @@
     S_ENDPGM 0
 
 ...
+---
+name:            test_sink_after_end_cf
+tracksRegLiveness: true
+machineFunctionInfo:
+  isEntryFunction: true
+body:             |
+  ; GFX10-LABEL: name: test_sink_after_end_cf
+  ; GFX10: bb.0.entry:
+  ; GFX10-NEXT:   successors: %bb.1(0x80000000)
+  ; GFX10-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT: bb.1:
+  ; GFX10-NEXT:   successors: %bb.2(0x40000000), %bb.3(0x40000000)
+  ; GFX10-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0
+  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:vreg_64 = COPY [[S_MOV_B32_]]
+  ; GFX10-NEXT:   [[GLOBAL_LOAD_DWORD:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_DWORD [[COPY3]], 0, 0, implicit $exec :: (load (s32), addrspace 1)
+  ; GFX10-NEXT:   [[GLOBAL_LOAD_DWORD1:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_DWORD [[COPY2]], 0, 0, implicit $exec :: (load (s32), addrspace 1)
+  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
+  ; GFX10-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 1
+  ; GFX10-NEXT:   [[V_CMP_LT_I32_e64_:%[0-9]+]]:sreg_32 = V_CMP_LT_I32_e64 [[COPY4]](s32), [[S_MOV_B32_1]], implicit $exec
+  ; GFX10-NEXT:   [[SI_IF:%[0-9]+]]:sreg_32 = SI_IF [[V_CMP_LT_I32_e64_]], %bb.3, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
+  ; GFX10-NEXT:   S_BRANCH %bb.2
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT: bb.2:
+  ; GFX10-NEXT:   successors: %bb.3(0x80000000)
+  ; GFX10-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT:   S_NOP 0
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT: bb.3:
+  ; GFX10-NEXT:   successors: %bb.4(0x40000000), %bb.6(0x40000000)
+  ; GFX10-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT:   %6:vgpr_32 = contract nofpexcept V_FMAC_F32_e64 0, [[GLOBAL_LOAD_DWORD]], 0, [[COPY]], 0, [[COPY1]], 0, 0, implicit $mode, implicit $exec
+  ; GFX10-NEXT:   %8:vgpr_32 = contract nofpexcept V_FMAC_F32_e64 0, [[GLOBAL_LOAD_DWORD1]], 0, [[COPY]], 0, [[COPY1]], 0, 0, implicit $mode, implicit $exec
+  ; GFX10-NEXT:   SI_END_CF [[SI_IF]], implicit-def dead $exec, implicit-def dead $scc, implicit $exec
+  ; GFX10-NEXT:   S_NOP 0, implicit %6, implicit %8
+  ; GFX10-NEXT:   S_CBRANCH_EXECZ %bb.6, implicit $exec
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT: bb.4:
+  ; GFX10-NEXT:   successors: %bb.5(0x04000000), %bb.4(0x7c000000)
+  ; GFX10-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT:   S_NOP 0
+  ; GFX10-NEXT:   S_CBRANCH_EXECZ %bb.4, implicit $exec
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT: bb.5:
+  ; GFX10-NEXT:   successors: %bb.6(0x80000000)
+  ; GFX10-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT:   S_NOP 0
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT: bb.6:
+  ; GFX10-NEXT:   successors: %bb.7(0x04000000), %bb.1(0x7c000000)
+  ; GFX10-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT:   S_CBRANCH_VCCZ %bb.1, implicit $vcc
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT: bb.7:
+  ; GFX10-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+  ; GFX10-NEXT: {{  $}}
+  ; GFX10-NEXT:   S_ENDPGM 0
+  bb.0.entry:
+    successors: %bb.1(0x80000000)
+
+    liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+    %101:vgpr_32 = COPY $vgpr0
+    %102:vgpr_32 = COPY $vgpr1
+    %15:vreg_64 = COPY $vgpr2_vgpr3
+
+  bb.1:
+    successors: %bb.2(0x40000000), %bb.3(0x40000000)
+    liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+
+    %20:sreg_32 = S_MOV_B32 0
+    %30:vreg_64 = COPY %20
+    %29:vgpr_32 = GLOBAL_LOAD_DWORD %30, 0, 0, implicit $exec :: (load (s32), addrspace 1)
+    %6:vgpr_32 = contract nofpexcept V_FMAC_F32_e64 0, %29, 0, %101, 0, %102, 0, 0, implicit $mode, implicit $exec
+    %31:vgpr_32 = GLOBAL_LOAD_DWORD  %15, 0, 0, implicit $exec :: (load (s32), addrspace 1)
+    %7:vgpr_32 = contract nofpexcept V_FMAC_F32_e64 0, %31, 0, %101, 0, %102, 0, 0, implicit $mode, implicit $exec
+    %16:vgpr_32(s32) = COPY $vgpr0
+    %23:sreg_32 = S_MOV_B32 1
+    %24:sreg_32 = V_CMP_LT_I32_e64 %16(s32), %23, implicit $exec
+    %0:sreg_32 = COPY %24
+    %5:sreg_32 = SI_IF %0, %bb.3, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
+    S_BRANCH %bb.2
+
+  bb.2:
+    successors: %bb.3(0x80000000)
+    liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+
+    S_NOP 0
+
+  bb.3:
+    successors: %bb.4(0x40000000), %bb.6(0x40000000)
+    liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+
+    SI_END_CF %5, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
+    S_NOP 0, implicit %6, implicit %7
+    S_CBRANCH_EXECZ %bb.6, implicit $exec
+
+  bb.4:
+    successors: %bb.5(0x04000000), %bb.4(0x7c000000)
+    liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+
+    S_NOP 0
+    S_CBRANCH_EXECZ %bb.4, implicit $exec
+
+  bb.5:
+    successors: %bb.6(0x80000000)
+    liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+
+    S_NOP 0
+
+  bb.6:
+    successors: %bb.7(0x04000000), %bb.1(0x7c000000)
+    liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+
+    S_CBRANCH_VCCZ %bb.1, implicit $vcc
+
+  bb.7:
+    liveins: $vgpr0, $vgpr1, $vgpr2_vgpr3, $vcc
+    S_ENDPGM 0
+...