# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX6 %s # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX7 %s # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX10 %s # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX10 %s --- name: test_fmad_s16_flush machineFunctionInfo: mode: fp64-fp16-input-denormals: false fp64-fp16-output-denormals: false body: | bb.0: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-LABEL: name: test_fmad_s16_flush ; GFX6: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-NEXT: {{ $}} ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX6-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX6-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX6-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32) ; GFX6-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX6-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX6-NEXT: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[FPEXT1]] ; GFX6-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX6-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC]](s16) ; GFX6-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX6-NEXT: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[FPEXT2]], [[FPEXT3]] ; GFX6-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32) ; GFX6-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC1]](s16) ; GFX6-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX7-LABEL: name: test_fmad_s16_flush ; GFX7: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX7-NEXT: {{ $}} ; GFX7-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX7-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX7-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX7-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX7-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX7-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32) ; GFX7-NEXT: [[FMAD:%[0-9]+]]:_(s16) = G_FMAD [[TRUNC]], [[TRUNC1]], [[TRUNC2]] ; GFX7-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FMAD]](s16) ; GFX7-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX10-LABEL: name: test_fmad_s16_flush ; GFX10: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX10-NEXT: {{ $}} ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32) ; GFX10-NEXT: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC1]] ; GFX10-NEXT: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[FMUL]], [[TRUNC2]] ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) ; GFX10-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $vgpr1 %2:_(s32) = COPY $vgpr1 %3:_(s16) = G_TRUNC %0 %4:_(s16) = G_TRUNC %1 %5:_(s16) = G_TRUNC %2 %6:_(s16) = G_FMAD %3, %4, %5 %7:_(s32) = G_ANYEXT %6 $vgpr0 = COPY %7 ... --- name: test_fmad_v2s16_flush machineFunctionInfo: mode: fp64-fp16-input-denormals: false fp64-fp16-output-denormals: false body: | bb.0: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-LABEL: name: test_fmad_v2s16_flush ; GFX6: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-NEXT: {{ $}} ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2 ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX6-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX6-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX6-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX6-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>) ; GFX6-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX6-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX6-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX6-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX6-NEXT: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[FPEXT1]] ; GFX6-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX6-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC]](s16) ; GFX6-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; GFX6-NEXT: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[FPEXT2]], [[FPEXT3]] ; GFX6-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32) ; GFX6-NEXT: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX6-NEXT: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX6-NEXT: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[FPEXT5]] ; GFX6-NEXT: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX6-NEXT: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC2]](s16) ; GFX6-NEXT: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; GFX6-NEXT: [[FADD1:%[0-9]+]]:_(s32) = G_FADD [[FPEXT6]], [[FPEXT7]] ; GFX6-NEXT: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32) ; GFX6-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC1]](s16) ; GFX6-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC3]](s16) ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX6-NEXT: $vgpr0 = COPY [[BITCAST3]](<2 x s16>) ; GFX7-LABEL: name: test_fmad_v2s16_flush ; GFX7: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX7-NEXT: {{ $}} ; GFX7-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX7-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX7-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2 ; GFX7-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX7-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX7-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX7-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX7-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX7-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX7-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX7-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX7-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX7-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>) ; GFX7-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX7-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX7-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX7-NEXT: [[FMAD:%[0-9]+]]:_(s16) = G_FMAD [[TRUNC]], [[TRUNC2]], [[TRUNC4]] ; GFX7-NEXT: [[FMAD1:%[0-9]+]]:_(s16) = G_FMAD [[TRUNC1]], [[TRUNC3]], [[TRUNC5]] ; GFX7-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FMAD]](s16) ; GFX7-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FMAD1]](s16) ; GFX7-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX7-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX7-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX7-NEXT: $vgpr0 = COPY [[BITCAST3]](<2 x s16>) ; GFX10-LABEL: name: test_fmad_v2s16_flush ; GFX10: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX10-NEXT: {{ $}} ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2 ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>) ; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX10-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10-NEXT: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC2]] ; GFX10-NEXT: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[FMUL]], [[TRUNC4]] ; GFX10-NEXT: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[TRUNC3]] ; GFX10-NEXT: [[FADD1:%[0-9]+]]:_(s16) = G_FADD [[FMUL1]], [[TRUNC5]] ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>) %0:_(<2 x s16>) = COPY $vgpr0 %1:_(<2 x s16>) = COPY $vgpr1 %2:_(<2 x s16>) = COPY $vgpr2 %3:_(<2 x s16>) = G_FMAD %0, %1, %2 $vgpr0 = COPY %3 ... --- name: test_fmad_v4s16_flush machineFunctionInfo: mode: fp64-fp16-input-denormals: false fp64-fp16-output-denormals: false body: | bb.0: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX6-LABEL: name: test_fmad_v4s16_flush ; GFX6: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX6-NEXT: {{ $}} ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr4_vgpr5 ; GFX6-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX6-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX6-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX6-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX6-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX6-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX6-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX6-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX6-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX6-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX6-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY2]](<4 x s16>) ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX6-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; GFX6-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; GFX6-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX6-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; GFX6-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; GFX6-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32) ; GFX6-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX6-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; GFX6-NEXT: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[FPEXT1]] ; GFX6-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX6-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC]](s16) ; GFX6-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC8]](s16) ; GFX6-NEXT: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[FPEXT2]], [[FPEXT3]] ; GFX6-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32) ; GFX6-NEXT: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX6-NEXT: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; GFX6-NEXT: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[FPEXT5]] ; GFX6-NEXT: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX6-NEXT: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC2]](s16) ; GFX6-NEXT: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC9]](s16) ; GFX6-NEXT: [[FADD1:%[0-9]+]]:_(s32) = G_FADD [[FPEXT6]], [[FPEXT7]] ; GFX6-NEXT: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32) ; GFX6-NEXT: [[FPEXT8:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX6-NEXT: [[FPEXT9:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC6]](s16) ; GFX6-NEXT: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT8]], [[FPEXT9]] ; GFX6-NEXT: [[FPTRUNC4:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32) ; GFX6-NEXT: [[FPEXT10:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC4]](s16) ; GFX6-NEXT: [[FPEXT11:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC10]](s16) ; GFX6-NEXT: [[FADD2:%[0-9]+]]:_(s32) = G_FADD [[FPEXT10]], [[FPEXT11]] ; GFX6-NEXT: [[FPTRUNC5:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD2]](s32) ; GFX6-NEXT: [[FPEXT12:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX6-NEXT: [[FPEXT13:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC7]](s16) ; GFX6-NEXT: [[FMUL3:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT12]], [[FPEXT13]] ; GFX6-NEXT: [[FPTRUNC6:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL3]](s32) ; GFX6-NEXT: [[FPEXT14:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC6]](s16) ; GFX6-NEXT: [[FPEXT15:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC11]](s16) ; GFX6-NEXT: [[FADD3:%[0-9]+]]:_(s32) = G_FADD [[FPEXT14]], [[FPEXT15]] ; GFX6-NEXT: [[FPTRUNC7:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD3]](s32) ; GFX6-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC1]](s16) ; GFX6-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC3]](s16) ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX6-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC5]](s16) ; GFX6-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC7]](s16) ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32) ; GFX6-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) ; GFX6-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) ; GFX7-LABEL: name: test_fmad_v4s16_flush ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX7-NEXT: {{ $}} ; GFX7-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX7-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX7-NEXT: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr4_vgpr5 ; GFX7-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX7-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX7-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX7-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX7-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX7-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX7-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX7-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX7-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX7-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX7-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX7-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX7-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX7-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX7-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX7-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX7-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX7-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX7-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX7-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY2]](<4 x s16>) ; GFX7-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX7-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; GFX7-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; GFX7-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) ; GFX7-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX7-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; GFX7-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; GFX7-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32) ; GFX7-NEXT: [[FMAD:%[0-9]+]]:_(s16) = G_FMAD [[TRUNC]], [[TRUNC4]], [[TRUNC8]] ; GFX7-NEXT: [[FMAD1:%[0-9]+]]:_(s16) = G_FMAD [[TRUNC1]], [[TRUNC5]], [[TRUNC9]] ; GFX7-NEXT: [[FMAD2:%[0-9]+]]:_(s16) = G_FMAD [[TRUNC2]], [[TRUNC6]], [[TRUNC10]] ; GFX7-NEXT: [[FMAD3:%[0-9]+]]:_(s16) = G_FMAD [[TRUNC3]], [[TRUNC7]], [[TRUNC11]] ; GFX7-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FMAD]](s16) ; GFX7-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FMAD1]](s16) ; GFX7-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX7-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX7-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX7-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FMAD2]](s16) ; GFX7-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[FMAD3]](s16) ; GFX7-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32) ; GFX7-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; GFX7-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX7-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) ; GFX7-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) ; GFX10-LABEL: name: test_fmad_v4s16_flush ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX10-NEXT: {{ $}} ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr4_vgpr5 ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX10-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX10-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX10-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX10-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX10-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY2]](<4 x s16>) ; GFX10-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX10-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; GFX10-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; GFX10-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) ; GFX10-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX10-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; GFX10-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; GFX10-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32) ; GFX10-NEXT: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC4]] ; GFX10-NEXT: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[FMUL]], [[TRUNC8]] ; GFX10-NEXT: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[TRUNC5]] ; GFX10-NEXT: [[FADD1:%[0-9]+]]:_(s16) = G_FADD [[FMUL1]], [[TRUNC9]] ; GFX10-NEXT: [[FMUL2:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC2]], [[TRUNC6]] ; GFX10-NEXT: [[FADD2:%[0-9]+]]:_(s16) = G_FADD [[FMUL2]], [[TRUNC10]] ; GFX10-NEXT: [[FMUL3:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC3]], [[TRUNC7]] ; GFX10-NEXT: [[FADD3:%[0-9]+]]:_(s16) = G_FADD [[FMUL3]], [[TRUNC11]] ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX10-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD2]](s16) ; GFX10-NEXT: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD3]](s16) ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[ANYEXT3]](s32) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 %2:_(<4 x s16>) = COPY $vgpr4_vgpr5 %3:_(<4 x s16>) = G_FMAD %0, %1, %2 $vgpr0_vgpr1 = COPY %3 ... --- name: test_fmad_s16_denorm machineFunctionInfo: mode: fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-LABEL: name: test_fmad_s16_denorm ; GFX6: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-NEXT: {{ $}} ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX6-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX6-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX6-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32) ; GFX6-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX6-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX6-NEXT: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[FPEXT1]] ; GFX6-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX6-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC]](s16) ; GFX6-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX6-NEXT: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[FPEXT2]], [[FPEXT3]] ; GFX6-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32) ; GFX6-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC1]](s16) ; GFX6-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX7-LABEL: name: test_fmad_s16_denorm ; GFX7: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX7-NEXT: {{ $}} ; GFX7-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX7-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX7-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX7-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX7-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX7-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32) ; GFX7-NEXT: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC1]] ; GFX7-NEXT: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[FMUL]], [[TRUNC2]] ; GFX7-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) ; GFX7-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX10-LABEL: name: test_fmad_s16_denorm ; GFX10: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX10-NEXT: {{ $}} ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32) ; GFX10-NEXT: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC1]] ; GFX10-NEXT: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[FMUL]], [[TRUNC2]] ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) ; GFX10-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $vgpr1 %2:_(s32) = COPY $vgpr1 %3:_(s16) = G_TRUNC %0 %4:_(s16) = G_TRUNC %1 %5:_(s16) = G_TRUNC %2 %6:_(s16) = G_FMAD %3, %4, %5 %7:_(s32) = G_ANYEXT %6 $vgpr0 = COPY %7 ... --- name: test_fmad_s16_denorm_flags machineFunctionInfo: mode: fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-LABEL: name: test_fmad_s16_denorm_flags ; GFX6: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-NEXT: {{ $}} ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX6-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX6-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX6-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32) ; GFX6-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX6-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX6-NEXT: [[FMUL:%[0-9]+]]:_(s32) = nnan G_FMUL [[FPEXT]], [[FPEXT1]] ; GFX6-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX6-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC]](s16) ; GFX6-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX6-NEXT: [[FADD:%[0-9]+]]:_(s32) = nnan G_FADD [[FPEXT2]], [[FPEXT3]] ; GFX6-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32) ; GFX6-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC1]](s16) ; GFX6-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX7-LABEL: name: test_fmad_s16_denorm_flags ; GFX7: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX7-NEXT: {{ $}} ; GFX7-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX7-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX7-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX7-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX7-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX7-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32) ; GFX7-NEXT: [[FMUL:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC]], [[TRUNC1]] ; GFX7-NEXT: [[FADD:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL]], [[TRUNC2]] ; GFX7-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) ; GFX7-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX10-LABEL: name: test_fmad_s16_denorm_flags ; GFX10: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX10-NEXT: {{ $}} ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32) ; GFX10-NEXT: [[FMUL:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC]], [[TRUNC1]] ; GFX10-NEXT: [[FADD:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL]], [[TRUNC2]] ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) ; GFX10-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $vgpr1 %2:_(s32) = COPY $vgpr1 %3:_(s16) = G_TRUNC %0 %4:_(s16) = G_TRUNC %1 %5:_(s16) = G_TRUNC %2 %6:_(s16) = nnan G_FMAD %3, %4, %5 %7:_(s32) = G_ANYEXT %6 $vgpr0 = COPY %7 ... --- name: test_fmad_v2s16_denorm machineFunctionInfo: mode: fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-LABEL: name: test_fmad_v2s16_denorm ; GFX6: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-NEXT: {{ $}} ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2 ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX6-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX6-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX6-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX6-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>) ; GFX6-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX6-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX6-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX6-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX6-NEXT: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[FPEXT1]] ; GFX6-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX6-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC]](s16) ; GFX6-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; GFX6-NEXT: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[FPEXT2]], [[FPEXT3]] ; GFX6-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32) ; GFX6-NEXT: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX6-NEXT: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX6-NEXT: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[FPEXT5]] ; GFX6-NEXT: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX6-NEXT: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC2]](s16) ; GFX6-NEXT: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; GFX6-NEXT: [[FADD1:%[0-9]+]]:_(s32) = G_FADD [[FPEXT6]], [[FPEXT7]] ; GFX6-NEXT: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32) ; GFX6-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC1]](s16) ; GFX6-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC3]](s16) ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX6-NEXT: $vgpr0 = COPY [[BITCAST3]](<2 x s16>) ; GFX7-LABEL: name: test_fmad_v2s16_denorm ; GFX7: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX7-NEXT: {{ $}} ; GFX7-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX7-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX7-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2 ; GFX7-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX7-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX7-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX7-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX7-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX7-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX7-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX7-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX7-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX7-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>) ; GFX7-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX7-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX7-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX7-NEXT: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC2]] ; GFX7-NEXT: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[FMUL]], [[TRUNC4]] ; GFX7-NEXT: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[TRUNC3]] ; GFX7-NEXT: [[FADD1:%[0-9]+]]:_(s16) = G_FADD [[FMUL1]], [[TRUNC5]] ; GFX7-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FADD]](s16) ; GFX7-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FADD1]](s16) ; GFX7-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX7-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX7-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX7-NEXT: $vgpr0 = COPY [[BITCAST3]](<2 x s16>) ; GFX10-LABEL: name: test_fmad_v2s16_denorm ; GFX10: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX10-NEXT: {{ $}} ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2 ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>) ; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX10-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10-NEXT: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC2]] ; GFX10-NEXT: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[FMUL]], [[TRUNC4]] ; GFX10-NEXT: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[TRUNC3]] ; GFX10-NEXT: [[FADD1:%[0-9]+]]:_(s16) = G_FADD [[FMUL1]], [[TRUNC5]] ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>) %0:_(<2 x s16>) = COPY $vgpr0 %1:_(<2 x s16>) = COPY $vgpr1 %2:_(<2 x s16>) = COPY $vgpr2 %3:_(<2 x s16>) = G_FMAD %0, %1, %2 $vgpr0 = COPY %3 ... --- name: test_fmad_v2s16_denorm_flags machineFunctionInfo: mode: fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-LABEL: name: test_fmad_v2s16_denorm_flags ; GFX6: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX6-NEXT: {{ $}} ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2 ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX6-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX6-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX6-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX6-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>) ; GFX6-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX6-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX6-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX6-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX6-NEXT: [[FMUL:%[0-9]+]]:_(s32) = nnan G_FMUL [[FPEXT]], [[FPEXT1]] ; GFX6-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX6-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC]](s16) ; GFX6-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; GFX6-NEXT: [[FADD:%[0-9]+]]:_(s32) = nnan G_FADD [[FPEXT2]], [[FPEXT3]] ; GFX6-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32) ; GFX6-NEXT: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX6-NEXT: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX6-NEXT: [[FMUL1:%[0-9]+]]:_(s32) = nnan G_FMUL [[FPEXT4]], [[FPEXT5]] ; GFX6-NEXT: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX6-NEXT: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC2]](s16) ; GFX6-NEXT: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; GFX6-NEXT: [[FADD1:%[0-9]+]]:_(s32) = nnan G_FADD [[FPEXT6]], [[FPEXT7]] ; GFX6-NEXT: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32) ; GFX6-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC1]](s16) ; GFX6-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC3]](s16) ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX6-NEXT: $vgpr0 = COPY [[BITCAST3]](<2 x s16>) ; GFX7-LABEL: name: test_fmad_v2s16_denorm_flags ; GFX7: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX7-NEXT: {{ $}} ; GFX7-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX7-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX7-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2 ; GFX7-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX7-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX7-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX7-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX7-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX7-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX7-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX7-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX7-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX7-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>) ; GFX7-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX7-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX7-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX7-NEXT: [[FMUL:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC]], [[TRUNC2]] ; GFX7-NEXT: [[FADD:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL]], [[TRUNC4]] ; GFX7-NEXT: [[FMUL1:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC1]], [[TRUNC3]] ; GFX7-NEXT: [[FADD1:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL1]], [[TRUNC5]] ; GFX7-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FADD]](s16) ; GFX7-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FADD1]](s16) ; GFX7-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX7-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX7-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX7-NEXT: $vgpr0 = COPY [[BITCAST3]](<2 x s16>) ; GFX10-LABEL: name: test_fmad_v2s16_denorm_flags ; GFX10: liveins: $vgpr0, $vgpr1, $vgpr2 ; GFX10-NEXT: {{ $}} ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2 ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>) ; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX10-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10-NEXT: [[FMUL:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC]], [[TRUNC2]] ; GFX10-NEXT: [[FADD:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL]], [[TRUNC4]] ; GFX10-NEXT: [[FMUL1:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC1]], [[TRUNC3]] ; GFX10-NEXT: [[FADD1:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL1]], [[TRUNC5]] ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>) %0:_(<2 x s16>) = COPY $vgpr0 %1:_(<2 x s16>) = COPY $vgpr1 %2:_(<2 x s16>) = COPY $vgpr2 %3:_(<2 x s16>) = nnan G_FMAD %0, %1, %2 $vgpr0 = COPY %3 ... --- name: test_fmad_v4s16_denorm machineFunctionInfo: mode: fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX6-LABEL: name: test_fmad_v4s16_denorm ; GFX6: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX6-NEXT: {{ $}} ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr4_vgpr5 ; GFX6-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX6-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX6-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX6-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX6-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX6-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX6-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX6-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX6-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX6-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX6-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY2]](<4 x s16>) ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX6-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; GFX6-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; GFX6-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX6-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; GFX6-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; GFX6-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32) ; GFX6-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX6-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; GFX6-NEXT: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[FPEXT1]] ; GFX6-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX6-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC]](s16) ; GFX6-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC8]](s16) ; GFX6-NEXT: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[FPEXT2]], [[FPEXT3]] ; GFX6-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32) ; GFX6-NEXT: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX6-NEXT: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; GFX6-NEXT: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[FPEXT5]] ; GFX6-NEXT: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX6-NEXT: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC2]](s16) ; GFX6-NEXT: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC9]](s16) ; GFX6-NEXT: [[FADD1:%[0-9]+]]:_(s32) = G_FADD [[FPEXT6]], [[FPEXT7]] ; GFX6-NEXT: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32) ; GFX6-NEXT: [[FPEXT8:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX6-NEXT: [[FPEXT9:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC6]](s16) ; GFX6-NEXT: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT8]], [[FPEXT9]] ; GFX6-NEXT: [[FPTRUNC4:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32) ; GFX6-NEXT: [[FPEXT10:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC4]](s16) ; GFX6-NEXT: [[FPEXT11:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC10]](s16) ; GFX6-NEXT: [[FADD2:%[0-9]+]]:_(s32) = G_FADD [[FPEXT10]], [[FPEXT11]] ; GFX6-NEXT: [[FPTRUNC5:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD2]](s32) ; GFX6-NEXT: [[FPEXT12:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX6-NEXT: [[FPEXT13:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC7]](s16) ; GFX6-NEXT: [[FMUL3:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT12]], [[FPEXT13]] ; GFX6-NEXT: [[FPTRUNC6:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL3]](s32) ; GFX6-NEXT: [[FPEXT14:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC6]](s16) ; GFX6-NEXT: [[FPEXT15:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC11]](s16) ; GFX6-NEXT: [[FADD3:%[0-9]+]]:_(s32) = G_FADD [[FPEXT14]], [[FPEXT15]] ; GFX6-NEXT: [[FPTRUNC7:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD3]](s32) ; GFX6-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC1]](s16) ; GFX6-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC3]](s16) ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX6-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC5]](s16) ; GFX6-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC7]](s16) ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32) ; GFX6-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) ; GFX6-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) ; GFX7-LABEL: name: test_fmad_v4s16_denorm ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX7-NEXT: {{ $}} ; GFX7-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX7-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX7-NEXT: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr4_vgpr5 ; GFX7-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX7-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX7-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX7-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX7-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX7-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX7-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX7-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX7-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX7-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX7-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX7-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX7-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX7-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX7-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX7-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX7-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX7-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX7-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX7-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY2]](<4 x s16>) ; GFX7-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX7-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; GFX7-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; GFX7-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) ; GFX7-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX7-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; GFX7-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; GFX7-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32) ; GFX7-NEXT: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC4]] ; GFX7-NEXT: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[FMUL]], [[TRUNC8]] ; GFX7-NEXT: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[TRUNC5]] ; GFX7-NEXT: [[FADD1:%[0-9]+]]:_(s16) = G_FADD [[FMUL1]], [[TRUNC9]] ; GFX7-NEXT: [[FMUL2:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC2]], [[TRUNC6]] ; GFX7-NEXT: [[FADD2:%[0-9]+]]:_(s16) = G_FADD [[FMUL2]], [[TRUNC10]] ; GFX7-NEXT: [[FMUL3:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC3]], [[TRUNC7]] ; GFX7-NEXT: [[FADD3:%[0-9]+]]:_(s16) = G_FADD [[FMUL3]], [[TRUNC11]] ; GFX7-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FADD]](s16) ; GFX7-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FADD1]](s16) ; GFX7-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX7-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX7-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX7-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FADD2]](s16) ; GFX7-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[FADD3]](s16) ; GFX7-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32) ; GFX7-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; GFX7-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX7-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) ; GFX7-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) ; GFX10-LABEL: name: test_fmad_v4s16_denorm ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX10-NEXT: {{ $}} ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr4_vgpr5 ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX10-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX10-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX10-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX10-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX10-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY2]](<4 x s16>) ; GFX10-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX10-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; GFX10-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; GFX10-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) ; GFX10-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX10-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; GFX10-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; GFX10-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32) ; GFX10-NEXT: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC4]] ; GFX10-NEXT: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[FMUL]], [[TRUNC8]] ; GFX10-NEXT: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[TRUNC5]] ; GFX10-NEXT: [[FADD1:%[0-9]+]]:_(s16) = G_FADD [[FMUL1]], [[TRUNC9]] ; GFX10-NEXT: [[FMUL2:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC2]], [[TRUNC6]] ; GFX10-NEXT: [[FADD2:%[0-9]+]]:_(s16) = G_FADD [[FMUL2]], [[TRUNC10]] ; GFX10-NEXT: [[FMUL3:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC3]], [[TRUNC7]] ; GFX10-NEXT: [[FADD3:%[0-9]+]]:_(s16) = G_FADD [[FMUL3]], [[TRUNC11]] ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX10-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD2]](s16) ; GFX10-NEXT: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD3]](s16) ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[ANYEXT3]](s32) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 %2:_(<4 x s16>) = COPY $vgpr4_vgpr5 %3:_(<4 x s16>) = G_FMAD %0, %1, %2 $vgpr0_vgpr1 = COPY %3 ... --- name: test_fmad_v4s16_denorm_flags machineFunctionInfo: mode: fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX6-LABEL: name: test_fmad_v4s16_denorm_flags ; GFX6: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX6-NEXT: {{ $}} ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr4_vgpr5 ; GFX6-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX6-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX6-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX6-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX6-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX6-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX6-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX6-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX6-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX6-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX6-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY2]](<4 x s16>) ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX6-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; GFX6-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; GFX6-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX6-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; GFX6-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; GFX6-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32) ; GFX6-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX6-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; GFX6-NEXT: [[FMUL:%[0-9]+]]:_(s32) = nnan G_FMUL [[FPEXT]], [[FPEXT1]] ; GFX6-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX6-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC]](s16) ; GFX6-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC8]](s16) ; GFX6-NEXT: [[FADD:%[0-9]+]]:_(s32) = nnan G_FADD [[FPEXT2]], [[FPEXT3]] ; GFX6-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32) ; GFX6-NEXT: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX6-NEXT: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; GFX6-NEXT: [[FMUL1:%[0-9]+]]:_(s32) = nnan G_FMUL [[FPEXT4]], [[FPEXT5]] ; GFX6-NEXT: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX6-NEXT: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC2]](s16) ; GFX6-NEXT: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC9]](s16) ; GFX6-NEXT: [[FADD1:%[0-9]+]]:_(s32) = nnan G_FADD [[FPEXT6]], [[FPEXT7]] ; GFX6-NEXT: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32) ; GFX6-NEXT: [[FPEXT8:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX6-NEXT: [[FPEXT9:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC6]](s16) ; GFX6-NEXT: [[FMUL2:%[0-9]+]]:_(s32) = nnan G_FMUL [[FPEXT8]], [[FPEXT9]] ; GFX6-NEXT: [[FPTRUNC4:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32) ; GFX6-NEXT: [[FPEXT10:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC4]](s16) ; GFX6-NEXT: [[FPEXT11:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC10]](s16) ; GFX6-NEXT: [[FADD2:%[0-9]+]]:_(s32) = nnan G_FADD [[FPEXT10]], [[FPEXT11]] ; GFX6-NEXT: [[FPTRUNC5:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD2]](s32) ; GFX6-NEXT: [[FPEXT12:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX6-NEXT: [[FPEXT13:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC7]](s16) ; GFX6-NEXT: [[FMUL3:%[0-9]+]]:_(s32) = nnan G_FMUL [[FPEXT12]], [[FPEXT13]] ; GFX6-NEXT: [[FPTRUNC6:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL3]](s32) ; GFX6-NEXT: [[FPEXT14:%[0-9]+]]:_(s32) = G_FPEXT [[FPTRUNC6]](s16) ; GFX6-NEXT: [[FPEXT15:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC11]](s16) ; GFX6-NEXT: [[FADD3:%[0-9]+]]:_(s32) = nnan G_FADD [[FPEXT14]], [[FPEXT15]] ; GFX6-NEXT: [[FPTRUNC7:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD3]](s32) ; GFX6-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC1]](s16) ; GFX6-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC3]](s16) ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX6-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC5]](s16) ; GFX6-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC7]](s16) ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32) ; GFX6-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) ; GFX6-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) ; GFX7-LABEL: name: test_fmad_v4s16_denorm_flags ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX7-NEXT: {{ $}} ; GFX7-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX7-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX7-NEXT: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr4_vgpr5 ; GFX7-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX7-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX7-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX7-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX7-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX7-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX7-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX7-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX7-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX7-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX7-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX7-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX7-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX7-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX7-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX7-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX7-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX7-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX7-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX7-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY2]](<4 x s16>) ; GFX7-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX7-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; GFX7-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; GFX7-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) ; GFX7-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX7-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; GFX7-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; GFX7-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32) ; GFX7-NEXT: [[FMUL:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC]], [[TRUNC4]] ; GFX7-NEXT: [[FADD:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL]], [[TRUNC8]] ; GFX7-NEXT: [[FMUL1:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC1]], [[TRUNC5]] ; GFX7-NEXT: [[FADD1:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL1]], [[TRUNC9]] ; GFX7-NEXT: [[FMUL2:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC2]], [[TRUNC6]] ; GFX7-NEXT: [[FADD2:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL2]], [[TRUNC10]] ; GFX7-NEXT: [[FMUL3:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC3]], [[TRUNC7]] ; GFX7-NEXT: [[FADD3:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL3]], [[TRUNC11]] ; GFX7-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FADD]](s16) ; GFX7-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FADD1]](s16) ; GFX7-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; GFX7-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; GFX7-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; GFX7-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FADD2]](s16) ; GFX7-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[FADD3]](s16) ; GFX7-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32) ; GFX7-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; GFX7-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX7-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) ; GFX7-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) ; GFX10-LABEL: name: test_fmad_v4s16_denorm_flags ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr4_vgpr5 ; GFX10-NEXT: {{ $}} ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr4_vgpr5 ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX10-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX10-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX10-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX10-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX10-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY2]](<4 x s16>) ; GFX10-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX10-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; GFX10-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; GFX10-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) ; GFX10-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX10-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; GFX10-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; GFX10-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32) ; GFX10-NEXT: [[FMUL:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC]], [[TRUNC4]] ; GFX10-NEXT: [[FADD:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL]], [[TRUNC8]] ; GFX10-NEXT: [[FMUL1:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC1]], [[TRUNC5]] ; GFX10-NEXT: [[FADD1:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL1]], [[TRUNC9]] ; GFX10-NEXT: [[FMUL2:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC2]], [[TRUNC6]] ; GFX10-NEXT: [[FADD2:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL2]], [[TRUNC10]] ; GFX10-NEXT: [[FMUL3:%[0-9]+]]:_(s16) = nnan G_FMUL [[TRUNC3]], [[TRUNC7]] ; GFX10-NEXT: [[FADD3:%[0-9]+]]:_(s16) = nnan G_FADD [[FMUL3]], [[TRUNC11]] ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX10-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD2]](s16) ; GFX10-NEXT: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD3]](s16) ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[ANYEXT3]](s32) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 %2:_(<4 x s16>) = COPY $vgpr4_vgpr5 %3:_(<4 x s16>) = nnan G_FMAD %0, %1, %2 $vgpr0_vgpr1 = COPY %3 ...