# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=SI %s # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX9 %s # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -global-isel-abort=0 -enable-unsafe-fp-math -o - %s | FileCheck -check-prefix=GFX9-UNSAFE %s # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX10 %s --- name: test_fdiv_s16 machineFunctionInfo: mode: fp32-input-denormals: true fp32-output-denormals: true fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0, $vgpr1 ; SI-LABEL: name: test_fdiv_s16 ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; SI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[FPEXT1]](s32), [[FPEXT]](s32) ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT6]](s32) ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC]](s16) ; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI-LABEL: name: test_fdiv_s16 ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; VI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; VI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; VI: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; VI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; VI: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC1]](s16), [[TRUNC]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-LABEL: name: test_fdiv_s16 ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX9: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX9: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX9: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX9: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX9: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC1]](s16), [[TRUNC]](s16) ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX9: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-UNSAFE-LABEL: name: test_fdiv_s16 ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9-UNSAFE: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX9-UNSAFE: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX9-UNSAFE: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC1]](s16) ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[INT]] ; GFX9-UNSAFE: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL]](s16) ; GFX9-UNSAFE: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX10-LABEL: name: test_fdiv_s16 ; GFX10: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX10: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) ; GFX10: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX10: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX10: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX10: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX10: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC1]](s16), [[TRUNC]](s16) ; GFX10: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX10: $vgpr0 = COPY [[ANYEXT]](s32) %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $vgpr1 %2:_(s16) = G_TRUNC %0 %3:_(s16) = G_TRUNC %1 %4:_(s16) = G_FDIV %2, %3 %5:_(s32) = G_ANYEXT %4 $vgpr0 = COPY %5 ... --- name: test_fdiv_s32_denorms_on machineFunctionInfo: mode: fp32-input-denormals: true fp32-output-denormals: true fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0, $vgpr1 ; SI-LABEL: name: test_fdiv_s32_denorms_on ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; SI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; SI: $vgpr0 = COPY [[INT6]](s32) ; VI-LABEL: name: test_fdiv_s32_denorms_on ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; VI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; VI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; VI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; VI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; VI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; VI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; VI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; VI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; VI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; VI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; VI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; VI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; VI: $vgpr0 = COPY [[INT6]](s32) ; GFX9-LABEL: name: test_fdiv_s32_denorms_on ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX9: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; GFX9: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; GFX9: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX9: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX9: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX9: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX9: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX9: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX9: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX9: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; GFX9: $vgpr0 = COPY [[INT6]](s32) ; GFX9-UNSAFE-LABEL: name: test_fdiv_s32_denorms_on ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9-UNSAFE: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[COPY1]](s32) ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[COPY]], [[INT]] ; GFX9-UNSAFE: $vgpr0 = COPY [[FMUL]](s32) ; GFX10-LABEL: name: test_fdiv_s32_denorms_on ; GFX10: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX10: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; GFX10: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; GFX10: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX10: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX10: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX10: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX10: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX10: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX10: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX10: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; GFX10: $vgpr0 = COPY [[INT6]](s32) %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $vgpr1 %2:_(s32) = G_FDIV %0, %1 $vgpr0 = COPY %2 ... --- name: test_fdiv_s32_denorms_off machineFunctionInfo: mode: fp32-input-denormals: false fp32-output-denormals: false fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0, $vgpr1 ; SI-LABEL: name: test_fdiv_s32_denorms_off ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; SI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; SI: $vgpr0 = COPY [[INT6]](s32) ; VI-LABEL: name: test_fdiv_s32_denorms_off ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; VI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; VI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; VI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; VI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; VI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; VI: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; VI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; VI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; VI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; VI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; VI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; VI: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; VI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; VI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; VI: $vgpr0 = COPY [[INT6]](s32) ; GFX9-LABEL: name: test_fdiv_s32_denorms_off ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX9: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; GFX9: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; GFX9: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX9: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX9: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; GFX9: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX9: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX9: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX9: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX9: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX9: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; GFX9: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; GFX9: $vgpr0 = COPY [[INT6]](s32) ; GFX9-UNSAFE-LABEL: name: test_fdiv_s32_denorms_off ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9-UNSAFE: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[COPY1]](s32) ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[COPY]], [[INT]] ; GFX9-UNSAFE: $vgpr0 = COPY [[FMUL]](s32) ; GFX10-LABEL: name: test_fdiv_s32_denorms_off ; GFX10: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX10: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; GFX10: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; GFX10: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX10: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX10: S_DENORM_MODE 15, implicit-def $mode, implicit $mode ; GFX10: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX10: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX10: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX10: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX10: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX10: S_DENORM_MODE 12, implicit-def $mode, implicit $mode ; GFX10: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; GFX10: $vgpr0 = COPY [[INT6]](s32) %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $vgpr1 %2:_(s32) = G_FDIV %0, %1 $vgpr0 = COPY %2 ... --- name: test_fdiv_s32_denorms_off_arcp machineFunctionInfo: mode: fp32-input-denormals: false fp32-output-denormals: false fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0, $vgpr1 ; SI-LABEL: name: test_fdiv_s32_denorms_off_arcp ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; SI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = arcp G_FNEG [[INT]] ; SI: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; SI: [[FMA:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[INT4]], [[C]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = arcp G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = arcp G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = arcp G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; SI: [[INT5:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; SI: $vgpr0 = COPY [[INT6]](s32) ; VI-LABEL: name: test_fdiv_s32_denorms_off_arcp ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; VI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; VI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; VI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; VI: [[INT4:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; VI: [[FNEG:%[0-9]+]]:_(s32) = arcp G_FNEG [[INT]] ; VI: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; VI: [[FMA:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[INT4]], [[C]] ; VI: [[FMA1:%[0-9]+]]:_(s32) = arcp G_FMA [[FMA]], [[INT4]], [[INT4]] ; VI: [[FMUL:%[0-9]+]]:_(s32) = arcp G_FMUL [[INT2]], [[FMA1]] ; VI: [[FMA2:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; VI: [[FMA3:%[0-9]+]]:_(s32) = arcp G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; VI: [[FMA4:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; VI: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; VI: [[INT5:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; VI: [[INT6:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; VI: $vgpr0 = COPY [[INT6]](s32) ; GFX9-LABEL: name: test_fdiv_s32_denorms_off_arcp ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX9: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; GFX9: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; GFX9: [[INT4:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX9: [[FNEG:%[0-9]+]]:_(s32) = arcp G_FNEG [[INT]] ; GFX9: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; GFX9: [[FMA:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX9: [[FMA1:%[0-9]+]]:_(s32) = arcp G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = arcp G_FMUL [[INT2]], [[FMA1]] ; GFX9: [[FMA2:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX9: [[FMA3:%[0-9]+]]:_(s32) = arcp G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX9: [[FMA4:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX9: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; GFX9: [[INT5:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; GFX9: $vgpr0 = COPY [[INT6]](s32) ; GFX9-UNSAFE-LABEL: name: test_fdiv_s32_denorms_off_arcp ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9-UNSAFE: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[COPY1]](s32) ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s32) = arcp G_FMUL [[COPY]], [[INT]] ; GFX9-UNSAFE: $vgpr0 = COPY [[FMUL]](s32) ; GFX10-LABEL: name: test_fdiv_s32_denorms_off_arcp ; GFX10: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX10: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 0 ; GFX10: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s32), [[COPY1]](s32), 1 ; GFX10: [[INT4:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX10: [[FNEG:%[0-9]+]]:_(s32) = arcp G_FNEG [[INT]] ; GFX10: S_DENORM_MODE 15, implicit-def $mode, implicit $mode ; GFX10: [[FMA:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX10: [[FMA1:%[0-9]+]]:_(s32) = arcp G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = arcp G_FMUL [[INT2]], [[FMA1]] ; GFX10: [[FMA2:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX10: [[FMA3:%[0-9]+]]:_(s32) = arcp G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX10: [[FMA4:%[0-9]+]]:_(s32) = arcp G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX10: S_DENORM_MODE 12, implicit-def $mode, implicit $mode ; GFX10: [[INT5:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s32) = arcp G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY1]](s32), [[COPY]](s32) ; GFX10: $vgpr0 = COPY [[INT6]](s32) %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $vgpr1 %2:_(s32) = arcp G_FDIV %0, %1 $vgpr0 = COPY %2 ... --- name: test_fdiv_s64 machineFunctionInfo: mode: fp32-input-denormals: false fp32-output-denormals: false fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 ; SI-LABEL: name: test_fdiv_s64 ; SI: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3 ; SI: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s64), [[COPY1]](s64), 0 ; SI: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; SI: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; SI: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; SI: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; SI: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; SI: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s64), [[COPY1]](s64), 1 ; SI: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; SI: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; SI: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](s64) ; SI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64) ; SI: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[INT]](s64) ; SI: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[INT3]](s64) ; SI: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[UV1]](s32), [[UV7]] ; SI: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[UV3]](s32), [[UV5]] ; SI: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP]], [[ICMP1]] ; SI: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[XOR]](s1) ; SI: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY1]](s64), [[COPY]](s64) ; SI: $vgpr0_vgpr1 = COPY [[INT6]](s64) ; VI-LABEL: name: test_fdiv_s64 ; VI: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3 ; VI: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; VI: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s64), [[COPY1]](s64), 0 ; VI: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; VI: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; VI: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; VI: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; VI: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; VI: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s64), [[COPY1]](s64), 1 ; VI: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; VI: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; VI: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; VI: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; VI: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY1]](s64), [[COPY]](s64) ; VI: $vgpr0_vgpr1 = COPY [[INT6]](s64) ; GFX9-LABEL: name: test_fdiv_s64 ; GFX9: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3 ; GFX9: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX9: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s64), [[COPY1]](s64), 0 ; GFX9: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; GFX9: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; GFX9: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; GFX9: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; GFX9: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; GFX9: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s64), [[COPY1]](s64), 1 ; GFX9: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; GFX9: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; GFX9: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; GFX9: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY1]](s64), [[COPY]](s64) ; GFX9: $vgpr0_vgpr1 = COPY [[INT6]](s64) ; GFX9-UNSAFE-LABEL: name: test_fdiv_s64 ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; GFX9-UNSAFE: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3 ; GFX9-UNSAFE: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[COPY1]] ; GFX9-UNSAFE: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[COPY1]](s64) ; GFX9-UNSAFE: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT]], [[C]] ; GFX9-UNSAFE: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[FMA]], [[INT]], [[INT]] ; GFX9-UNSAFE: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; GFX9-UNSAFE: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA2]], [[FMA1]], [[FMA1]] ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[COPY]], [[FMA3]] ; GFX9-UNSAFE: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[COPY]] ; GFX9-UNSAFE: [[FMA5:%[0-9]+]]:_(s64) = G_FMA [[FMA4]], [[FMA3]], [[FMUL]] ; GFX9-UNSAFE: $vgpr0_vgpr1 = COPY [[FMA5]](s64) ; GFX10-LABEL: name: test_fdiv_s64 ; GFX10: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; GFX10: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3 ; GFX10: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX10: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s64), [[COPY1]](s64), 0 ; GFX10: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; GFX10: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; GFX10: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; GFX10: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; GFX10: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; GFX10: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[COPY]](s64), [[COPY1]](s64), 1 ; GFX10: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; GFX10: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; GFX10: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; GFX10: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY1]](s64), [[COPY]](s64) ; GFX10: $vgpr0_vgpr1 = COPY [[INT6]](s64) %0:_(s64) = COPY $vgpr0_vgpr1 %1:_(s64) = COPY $vgpr2_vgpr3 %2:_(s64) = G_FDIV %0, %1 $vgpr0_vgpr1 = COPY %2 ... --- name: test_fdiv_v2s32 machineFunctionInfo: mode: fp32-input-denormals: false fp32-output-denormals: false fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 ; SI-LABEL: name: test_fdiv_v2s32 ; SI: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) ; SI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) ; SI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV2]](s32), [[UV]](s32) ; SI: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 0 ; SI: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 1 ; SI: [[INT11:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; SI: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[INT7]] ; SI: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; SI: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[INT11]], [[C]] ; SI: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FMA5]], [[INT11]], [[INT11]] ; SI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[INT9]], [[FMA6]] ; SI: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; SI: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; SI: [[FMA9:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; SI: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; SI: [[INT12:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; SI: [[INT13:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV3]](s32), [[UV1]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32) ; SI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) ; VI-LABEL: name: test_fdiv_v2s32 ; VI: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) ; VI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; VI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 0 ; VI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 1 ; VI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; VI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; VI: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; VI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; VI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; VI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; VI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; VI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; VI: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; VI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; VI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV2]](s32), [[UV]](s32) ; VI: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 0 ; VI: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 1 ; VI: [[INT11:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; VI: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[INT7]] ; VI: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; VI: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[INT11]], [[C]] ; VI: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FMA5]], [[INT11]], [[INT11]] ; VI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[INT9]], [[FMA6]] ; VI: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; VI: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; VI: [[FMA9:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; VI: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; VI: [[INT12:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; VI: [[INT13:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV3]](s32), [[UV1]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32) ; VI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) ; GFX9-LABEL: name: test_fdiv_v2s32 ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) ; GFX9: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX9: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 0 ; GFX9: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 1 ; GFX9: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX9: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX9: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; GFX9: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX9: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX9: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX9: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX9: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX9: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; GFX9: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV2]](s32), [[UV]](s32) ; GFX9: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 0 ; GFX9: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 1 ; GFX9: [[INT11:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; GFX9: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[INT7]] ; GFX9: S_SETREG_IMM32_B32 3, 2305, implicit-def $mode, implicit $mode ; GFX9: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[INT11]], [[C]] ; GFX9: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FMA5]], [[INT11]], [[INT11]] ; GFX9: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[INT9]], [[FMA6]] ; GFX9: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; GFX9: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; GFX9: [[FMA9:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; GFX9: S_SETREG_IMM32_B32 0, 2305, implicit-def $mode, implicit $mode ; GFX9: [[INT12:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; GFX9: [[INT13:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV3]](s32), [[UV1]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32) ; GFX9: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) ; GFX9-UNSAFE-LABEL: name: test_fdiv_v2s32 ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 ; GFX9-UNSAFE: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 ; GFX9-UNSAFE: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) ; GFX9-UNSAFE: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[UV2]](s32) ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[UV]], [[INT]] ; GFX9-UNSAFE: [[INT1:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[UV3]](s32) ; GFX9-UNSAFE: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[UV1]], [[INT1]] ; GFX9-UNSAFE: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32) ; GFX9-UNSAFE: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) ; GFX10-LABEL: name: test_fdiv_v2s32 ; GFX10: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 ; GFX10: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 ; GFX10: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) ; GFX10: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) ; GFX10: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX10: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 0 ; GFX10: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 1 ; GFX10: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX10: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX10: S_DENORM_MODE 15, implicit-def $mode, implicit $mode ; GFX10: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX10: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX10: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX10: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX10: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX10: S_DENORM_MODE 12, implicit-def $mode, implicit $mode ; GFX10: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV2]](s32), [[UV]](s32) ; GFX10: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 0 ; GFX10: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 1 ; GFX10: [[INT11:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; GFX10: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[INT7]] ; GFX10: S_DENORM_MODE 15, implicit-def $mode, implicit $mode ; GFX10: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[INT11]], [[C]] ; GFX10: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FMA5]], [[INT11]], [[INT11]] ; GFX10: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[INT9]], [[FMA6]] ; GFX10: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; GFX10: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; GFX10: [[FMA9:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; GFX10: S_DENORM_MODE 12, implicit-def $mode, implicit $mode ; GFX10: [[INT12:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; GFX10: [[INT13:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV3]](s32), [[UV1]](s32) ; GFX10: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32) ; GFX10: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) %0:_(<2 x s32>) = COPY $vgpr0_vgpr1 %1:_(<2 x s32>) = COPY $vgpr2_vgpr3 %2:_(<2 x s32>) = G_FDIV %0, %1 $vgpr0_vgpr1 = COPY %2 ... --- name: test_fdiv_v2s32_flags body: | bb.0: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 ; SI-LABEL: name: test_fdiv_v2s32_flags ; SI: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) ; SI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) ; SI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = nnan G_FNEG [[INT]] ; SI: [[FMA:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[INT4]], [[C]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = nnan G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: [[INT5:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV2]](s32), [[UV]](s32) ; SI: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 0 ; SI: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 1 ; SI: [[INT11:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; SI: [[FNEG1:%[0-9]+]]:_(s32) = nnan G_FNEG [[INT7]] ; SI: [[FMA5:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[INT11]], [[C]] ; SI: [[FMA6:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA5]], [[INT11]], [[INT11]] ; SI: [[FMUL1:%[0-9]+]]:_(s32) = nnan G_FMUL [[INT9]], [[FMA6]] ; SI: [[FMA7:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; SI: [[FMA8:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; SI: [[FMA9:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; SI: [[INT12:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; SI: [[INT13:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV3]](s32), [[UV1]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32) ; SI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) ; VI-LABEL: name: test_fdiv_v2s32_flags ; VI: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) ; VI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; VI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 0 ; VI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 1 ; VI: [[INT4:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; VI: [[FNEG:%[0-9]+]]:_(s32) = nnan G_FNEG [[INT]] ; VI: [[FMA:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[INT4]], [[C]] ; VI: [[FMA1:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA]], [[INT4]], [[INT4]] ; VI: [[FMUL:%[0-9]+]]:_(s32) = nnan G_FMUL [[INT2]], [[FMA1]] ; VI: [[FMA2:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; VI: [[FMA3:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; VI: [[FMA4:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; VI: [[INT5:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; VI: [[INT6:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV2]](s32), [[UV]](s32) ; VI: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 0 ; VI: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 1 ; VI: [[INT11:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; VI: [[FNEG1:%[0-9]+]]:_(s32) = nnan G_FNEG [[INT7]] ; VI: [[FMA5:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[INT11]], [[C]] ; VI: [[FMA6:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA5]], [[INT11]], [[INT11]] ; VI: [[FMUL1:%[0-9]+]]:_(s32) = nnan G_FMUL [[INT9]], [[FMA6]] ; VI: [[FMA7:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; VI: [[FMA8:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; VI: [[FMA9:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; VI: [[INT12:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; VI: [[INT13:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV3]](s32), [[UV1]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32) ; VI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) ; GFX9-LABEL: name: test_fdiv_v2s32_flags ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) ; GFX9: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX9: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 0 ; GFX9: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 1 ; GFX9: [[INT4:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX9: [[FNEG:%[0-9]+]]:_(s32) = nnan G_FNEG [[INT]] ; GFX9: [[FMA:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX9: [[FMA1:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = nnan G_FMUL [[INT2]], [[FMA1]] ; GFX9: [[FMA2:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX9: [[FMA3:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX9: [[FMA4:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX9: [[INT5:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV2]](s32), [[UV]](s32) ; GFX9: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 0 ; GFX9: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 1 ; GFX9: [[INT11:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; GFX9: [[FNEG1:%[0-9]+]]:_(s32) = nnan G_FNEG [[INT7]] ; GFX9: [[FMA5:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[INT11]], [[C]] ; GFX9: [[FMA6:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA5]], [[INT11]], [[INT11]] ; GFX9: [[FMUL1:%[0-9]+]]:_(s32) = nnan G_FMUL [[INT9]], [[FMA6]] ; GFX9: [[FMA7:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; GFX9: [[FMA8:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; GFX9: [[FMA9:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; GFX9: [[INT12:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; GFX9: [[INT13:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV3]](s32), [[UV1]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32) ; GFX9: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) ; GFX9-UNSAFE-LABEL: name: test_fdiv_v2s32_flags ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 ; GFX9-UNSAFE: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 ; GFX9-UNSAFE: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) ; GFX9-UNSAFE: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[UV2]](s32) ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s32) = nnan G_FMUL [[UV]], [[INT]] ; GFX9-UNSAFE: [[INT1:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[UV3]](s32) ; GFX9-UNSAFE: [[FMUL1:%[0-9]+]]:_(s32) = nnan G_FMUL [[UV1]], [[INT1]] ; GFX9-UNSAFE: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32) ; GFX9-UNSAFE: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) ; GFX10-LABEL: name: test_fdiv_v2s32_flags ; GFX10: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 ; GFX10: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 ; GFX10: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) ; GFX10: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) ; GFX10: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX10: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 0 ; GFX10: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV2]](s32), 1 ; GFX10: [[INT4:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX10: [[FNEG:%[0-9]+]]:_(s32) = nnan G_FNEG [[INT]] ; GFX10: [[FMA:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX10: [[FMA1:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = nnan G_FMUL [[INT2]], [[FMA1]] ; GFX10: [[FMA2:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX10: [[FMA3:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX10: [[FMA4:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX10: [[INT5:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV2]](s32), [[UV]](s32) ; GFX10: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 0 ; GFX10: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV3]](s32), 1 ; GFX10: [[INT11:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; GFX10: [[FNEG1:%[0-9]+]]:_(s32) = nnan G_FNEG [[INT7]] ; GFX10: [[FMA5:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[INT11]], [[C]] ; GFX10: [[FMA6:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA5]], [[INT11]], [[INT11]] ; GFX10: [[FMUL1:%[0-9]+]]:_(s32) = nnan G_FMUL [[INT9]], [[FMA6]] ; GFX10: [[FMA7:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; GFX10: [[FMA8:%[0-9]+]]:_(s32) = nnan G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; GFX10: [[FMA9:%[0-9]+]]:_(s32) = nnan G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; GFX10: [[INT12:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; GFX10: [[INT13:%[0-9]+]]:_(s32) = nnan G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV3]](s32), [[UV1]](s32) ; GFX10: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32) ; GFX10: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) %0:_(<2 x s32>) = COPY $vgpr0_vgpr1 %1:_(<2 x s32>) = COPY $vgpr2_vgpr3 %2:_(<2 x s32>) = nnan G_FDIV %0, %1 $vgpr0_vgpr1 = COPY %2 ... --- name: test_fdiv_v3s32 body: | bb.0: liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; SI-LABEL: name: test_fdiv_v3s32 ; SI: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2 ; SI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5 ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>) ; SI: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) ; SI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV3]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV3]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV3]](s32), [[UV]](s32) ; SI: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV4]](s32), 0 ; SI: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV4]](s32), 1 ; SI: [[INT11:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; SI: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[INT7]] ; SI: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[INT11]], [[C]] ; SI: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FMA5]], [[INT11]], [[INT11]] ; SI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[INT9]], [[FMA6]] ; SI: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; SI: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; SI: [[FMA9:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; SI: [[INT12:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; SI: [[INT13:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV4]](s32), [[UV1]](s32) ; SI: [[INT14:%[0-9]+]]:_(s32), [[INT15:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV2]](s32), [[UV5]](s32), 0 ; SI: [[INT16:%[0-9]+]]:_(s32), [[INT17:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV2]](s32), [[UV5]](s32), 1 ; SI: [[INT18:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT14]](s32) ; SI: [[FNEG2:%[0-9]+]]:_(s32) = G_FNEG [[INT14]] ; SI: [[FMA10:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[INT18]], [[C]] ; SI: [[FMA11:%[0-9]+]]:_(s32) = G_FMA [[FMA10]], [[INT18]], [[INT18]] ; SI: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[INT16]], [[FMA11]] ; SI: [[FMA12:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMUL2]], [[INT16]] ; SI: [[FMA13:%[0-9]+]]:_(s32) = G_FMA [[FMA12]], [[FMA11]], [[FMUL2]] ; SI: [[FMA14:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMA13]], [[INT16]] ; SI: [[INT19:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA14]](s32), [[FMA11]](s32), [[FMA13]](s32), [[INT17]](s1) ; SI: [[INT20:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT19]](s32), [[UV5]](s32), [[UV2]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32), [[INT20]](s32) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) ; VI-LABEL: name: test_fdiv_v3s32 ; VI: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2 ; VI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5 ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>) ; VI: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; VI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV3]](s32), 0 ; VI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV3]](s32), 1 ; VI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; VI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; VI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; VI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; VI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; VI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; VI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; VI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; VI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV3]](s32), [[UV]](s32) ; VI: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV4]](s32), 0 ; VI: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV4]](s32), 1 ; VI: [[INT11:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; VI: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[INT7]] ; VI: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[INT11]], [[C]] ; VI: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FMA5]], [[INT11]], [[INT11]] ; VI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[INT9]], [[FMA6]] ; VI: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; VI: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; VI: [[FMA9:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; VI: [[INT12:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; VI: [[INT13:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV4]](s32), [[UV1]](s32) ; VI: [[INT14:%[0-9]+]]:_(s32), [[INT15:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV2]](s32), [[UV5]](s32), 0 ; VI: [[INT16:%[0-9]+]]:_(s32), [[INT17:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV2]](s32), [[UV5]](s32), 1 ; VI: [[INT18:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT14]](s32) ; VI: [[FNEG2:%[0-9]+]]:_(s32) = G_FNEG [[INT14]] ; VI: [[FMA10:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[INT18]], [[C]] ; VI: [[FMA11:%[0-9]+]]:_(s32) = G_FMA [[FMA10]], [[INT18]], [[INT18]] ; VI: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[INT16]], [[FMA11]] ; VI: [[FMA12:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMUL2]], [[INT16]] ; VI: [[FMA13:%[0-9]+]]:_(s32) = G_FMA [[FMA12]], [[FMA11]], [[FMUL2]] ; VI: [[FMA14:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMA13]], [[INT16]] ; VI: [[INT19:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA14]](s32), [[FMA11]](s32), [[FMA13]](s32), [[INT17]](s1) ; VI: [[INT20:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT19]](s32), [[UV5]](s32), [[UV2]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32), [[INT20]](s32) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) ; GFX9-LABEL: name: test_fdiv_v3s32 ; GFX9: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2 ; GFX9: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5 ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>) ; GFX9: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX9: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV3]](s32), 0 ; GFX9: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV3]](s32), 1 ; GFX9: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX9: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX9: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX9: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX9: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX9: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX9: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX9: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV3]](s32), [[UV]](s32) ; GFX9: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV4]](s32), 0 ; GFX9: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV4]](s32), 1 ; GFX9: [[INT11:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; GFX9: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[INT7]] ; GFX9: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[INT11]], [[C]] ; GFX9: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FMA5]], [[INT11]], [[INT11]] ; GFX9: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[INT9]], [[FMA6]] ; GFX9: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; GFX9: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; GFX9: [[FMA9:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; GFX9: [[INT12:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; GFX9: [[INT13:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV4]](s32), [[UV1]](s32) ; GFX9: [[INT14:%[0-9]+]]:_(s32), [[INT15:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV2]](s32), [[UV5]](s32), 0 ; GFX9: [[INT16:%[0-9]+]]:_(s32), [[INT17:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV2]](s32), [[UV5]](s32), 1 ; GFX9: [[INT18:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT14]](s32) ; GFX9: [[FNEG2:%[0-9]+]]:_(s32) = G_FNEG [[INT14]] ; GFX9: [[FMA10:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[INT18]], [[C]] ; GFX9: [[FMA11:%[0-9]+]]:_(s32) = G_FMA [[FMA10]], [[INT18]], [[INT18]] ; GFX9: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[INT16]], [[FMA11]] ; GFX9: [[FMA12:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMUL2]], [[INT16]] ; GFX9: [[FMA13:%[0-9]+]]:_(s32) = G_FMA [[FMA12]], [[FMA11]], [[FMUL2]] ; GFX9: [[FMA14:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMA13]], [[INT16]] ; GFX9: [[INT19:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA14]](s32), [[FMA11]](s32), [[FMA13]](s32), [[INT17]](s1) ; GFX9: [[INT20:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT19]](s32), [[UV5]](s32), [[UV2]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32), [[INT20]](s32) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) ; GFX9-UNSAFE-LABEL: name: test_fdiv_v3s32 ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2 ; GFX9-UNSAFE: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5 ; GFX9-UNSAFE: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>) ; GFX9-UNSAFE: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[UV3]](s32) ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[UV]], [[INT]] ; GFX9-UNSAFE: [[INT1:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[UV4]](s32) ; GFX9-UNSAFE: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[UV1]], [[INT1]] ; GFX9-UNSAFE: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[UV5]](s32) ; GFX9-UNSAFE: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[UV2]], [[INT2]] ; GFX9-UNSAFE: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32), [[FMUL2]](s32) ; GFX9-UNSAFE: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) ; GFX10-LABEL: name: test_fdiv_v3s32 ; GFX10: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2 ; GFX10: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5 ; GFX10: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>) ; GFX10: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) ; GFX10: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX10: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV3]](s32), 0 ; GFX10: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s32), [[UV3]](s32), 1 ; GFX10: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX10: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX10: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX10: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX10: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX10: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX10: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX10: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[UV3]](s32), [[UV]](s32) ; GFX10: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV4]](s32), 0 ; GFX10: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s32), [[UV4]](s32), 1 ; GFX10: [[INT11:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; GFX10: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[INT7]] ; GFX10: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[INT11]], [[C]] ; GFX10: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FMA5]], [[INT11]], [[INT11]] ; GFX10: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[INT9]], [[FMA6]] ; GFX10: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; GFX10: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; GFX10: [[FMA9:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; GFX10: [[INT12:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; GFX10: [[INT13:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[UV4]](s32), [[UV1]](s32) ; GFX10: [[INT14:%[0-9]+]]:_(s32), [[INT15:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV2]](s32), [[UV5]](s32), 0 ; GFX10: [[INT16:%[0-9]+]]:_(s32), [[INT17:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV2]](s32), [[UV5]](s32), 1 ; GFX10: [[INT18:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT14]](s32) ; GFX10: [[FNEG2:%[0-9]+]]:_(s32) = G_FNEG [[INT14]] ; GFX10: [[FMA10:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[INT18]], [[C]] ; GFX10: [[FMA11:%[0-9]+]]:_(s32) = G_FMA [[FMA10]], [[INT18]], [[INT18]] ; GFX10: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[INT16]], [[FMA11]] ; GFX10: [[FMA12:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMUL2]], [[INT16]] ; GFX10: [[FMA13:%[0-9]+]]:_(s32) = G_FMA [[FMA12]], [[FMA11]], [[FMUL2]] ; GFX10: [[FMA14:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMA13]], [[INT16]] ; GFX10: [[INT19:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA14]](s32), [[FMA11]](s32), [[FMA13]](s32), [[INT17]](s1) ; GFX10: [[INT20:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT19]](s32), [[UV5]](s32), [[UV2]](s32) ; GFX10: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[INT6]](s32), [[INT13]](s32), [[INT20]](s32) ; GFX10: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2 %1:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5 %2:_(<3 x s32>) = G_FDIV %0, %1 $vgpr0_vgpr1_vgpr2 = COPY %2 ... --- name: test_fdiv_v2s64 body: | bb.0: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5_vgpr6_vgpr7 ; SI-LABEL: name: test_fdiv_v2s64 ; SI: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 ; SI: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 ; SI: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>) ; SI: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>) ; SI: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s64), [[UV2]](s64), 0 ; SI: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; SI: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; SI: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; SI: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; SI: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; SI: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s64), [[UV2]](s64), 1 ; SI: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; SI: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; SI: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; SI: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](s64) ; SI: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV2]](s64) ; SI: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[INT]](s64) ; SI: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[INT3]](s64) ; SI: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[UV5]](s32), [[UV11]] ; SI: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[UV7]](s32), [[UV9]] ; SI: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP]], [[ICMP1]] ; SI: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[XOR]](s1) ; SI: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[UV2]](s64), [[UV]](s64) ; SI: [[INT7:%[0-9]+]]:_(s64), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s64), [[UV3]](s64), 0 ; SI: [[FNEG1:%[0-9]+]]:_(s64) = G_FNEG [[INT7]] ; SI: [[INT9:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s64) ; SI: [[FMA5:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[INT9]], [[C]] ; SI: [[FMA6:%[0-9]+]]:_(s64) = G_FMA [[INT9]], [[FMA5]], [[INT9]] ; SI: [[FMA7:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[FMA6]], [[C]] ; SI: [[INT10:%[0-9]+]]:_(s64), [[INT11:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s64), [[UV3]](s64), 1 ; SI: [[FMA8:%[0-9]+]]:_(s64) = G_FMA [[FMA6]], [[FMA7]], [[FMA6]] ; SI: [[FMUL1:%[0-9]+]]:_(s64) = G_FMUL [[INT10]], [[FMA8]] ; SI: [[FMA9:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[FMUL1]], [[INT10]] ; SI: [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64) ; SI: [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64) ; SI: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[INT7]](s64) ; SI: [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[INT10]](s64) ; SI: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[UV13]](s32), [[UV19]] ; SI: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[UV15]](s32), [[UV17]] ; SI: [[XOR1:%[0-9]+]]:_(s1) = G_XOR [[ICMP2]], [[ICMP3]] ; SI: [[INT12:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s64), [[FMA8]](s64), [[FMUL1]](s64), [[XOR1]](s1) ; SI: [[INT13:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s64), [[UV3]](s64), [[UV1]](s64) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[INT6]](s64), [[INT13]](s64) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>) ; VI-LABEL: name: test_fdiv_v2s64 ; VI: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 ; VI: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 ; VI: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>) ; VI: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>) ; VI: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; VI: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s64), [[UV2]](s64), 0 ; VI: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; VI: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; VI: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; VI: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; VI: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; VI: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s64), [[UV2]](s64), 1 ; VI: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; VI: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; VI: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; VI: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; VI: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[UV2]](s64), [[UV]](s64) ; VI: [[INT7:%[0-9]+]]:_(s64), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s64), [[UV3]](s64), 0 ; VI: [[FNEG1:%[0-9]+]]:_(s64) = G_FNEG [[INT7]] ; VI: [[INT9:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s64) ; VI: [[FMA5:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[INT9]], [[C]] ; VI: [[FMA6:%[0-9]+]]:_(s64) = G_FMA [[INT9]], [[FMA5]], [[INT9]] ; VI: [[FMA7:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[FMA6]], [[C]] ; VI: [[INT10:%[0-9]+]]:_(s64), [[INT11:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s64), [[UV3]](s64), 1 ; VI: [[FMA8:%[0-9]+]]:_(s64) = G_FMA [[FMA6]], [[FMA7]], [[FMA6]] ; VI: [[FMUL1:%[0-9]+]]:_(s64) = G_FMUL [[INT10]], [[FMA8]] ; VI: [[FMA9:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[FMUL1]], [[INT10]] ; VI: [[INT12:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s64), [[FMA8]](s64), [[FMUL1]](s64), [[INT11]](s1) ; VI: [[INT13:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s64), [[UV3]](s64), [[UV1]](s64) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[INT6]](s64), [[INT13]](s64) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>) ; GFX9-LABEL: name: test_fdiv_v2s64 ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 ; GFX9: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>) ; GFX9: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>) ; GFX9: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX9: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s64), [[UV2]](s64), 0 ; GFX9: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; GFX9: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; GFX9: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; GFX9: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; GFX9: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; GFX9: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s64), [[UV2]](s64), 1 ; GFX9: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; GFX9: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; GFX9: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; GFX9: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[UV2]](s64), [[UV]](s64) ; GFX9: [[INT7:%[0-9]+]]:_(s64), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s64), [[UV3]](s64), 0 ; GFX9: [[FNEG1:%[0-9]+]]:_(s64) = G_FNEG [[INT7]] ; GFX9: [[INT9:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s64) ; GFX9: [[FMA5:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[INT9]], [[C]] ; GFX9: [[FMA6:%[0-9]+]]:_(s64) = G_FMA [[INT9]], [[FMA5]], [[INT9]] ; GFX9: [[FMA7:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[FMA6]], [[C]] ; GFX9: [[INT10:%[0-9]+]]:_(s64), [[INT11:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s64), [[UV3]](s64), 1 ; GFX9: [[FMA8:%[0-9]+]]:_(s64) = G_FMA [[FMA6]], [[FMA7]], [[FMA6]] ; GFX9: [[FMUL1:%[0-9]+]]:_(s64) = G_FMUL [[INT10]], [[FMA8]] ; GFX9: [[FMA9:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[FMUL1]], [[INT10]] ; GFX9: [[INT12:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s64), [[FMA8]](s64), [[FMUL1]](s64), [[INT11]](s1) ; GFX9: [[INT13:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s64), [[UV3]](s64), [[UV1]](s64) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[INT6]](s64), [[INT13]](s64) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>) ; GFX9-UNSAFE-LABEL: name: test_fdiv_v2s64 ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 ; GFX9-UNSAFE: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 ; GFX9-UNSAFE: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>) ; GFX9-UNSAFE: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>) ; GFX9-UNSAFE: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[UV2]] ; GFX9-UNSAFE: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[UV2]](s64) ; GFX9-UNSAFE: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT]], [[C]] ; GFX9-UNSAFE: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[FMA]], [[INT]], [[INT]] ; GFX9-UNSAFE: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; GFX9-UNSAFE: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA2]], [[FMA1]], [[FMA1]] ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[UV]], [[FMA3]] ; GFX9-UNSAFE: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[UV]] ; GFX9-UNSAFE: [[FMA5:%[0-9]+]]:_(s64) = G_FMA [[FMA4]], [[FMA3]], [[FMUL]] ; GFX9-UNSAFE: [[FNEG1:%[0-9]+]]:_(s64) = G_FNEG [[UV3]] ; GFX9-UNSAFE: [[INT1:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[UV3]](s64) ; GFX9-UNSAFE: [[FMA6:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[INT1]], [[C]] ; GFX9-UNSAFE: [[FMA7:%[0-9]+]]:_(s64) = G_FMA [[FMA6]], [[INT1]], [[INT1]] ; GFX9-UNSAFE: [[FMA8:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[FMA7]], [[C]] ; GFX9-UNSAFE: [[FMA9:%[0-9]+]]:_(s64) = G_FMA [[FMA8]], [[FMA7]], [[FMA7]] ; GFX9-UNSAFE: [[FMUL1:%[0-9]+]]:_(s64) = G_FMUL [[UV1]], [[FMA9]] ; GFX9-UNSAFE: [[FMA10:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[FMUL1]], [[UV1]] ; GFX9-UNSAFE: [[FMA11:%[0-9]+]]:_(s64) = G_FMA [[FMA10]], [[FMA9]], [[FMUL1]] ; GFX9-UNSAFE: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[FMA5]](s64), [[FMA11]](s64) ; GFX9-UNSAFE: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>) ; GFX10-LABEL: name: test_fdiv_v2s64 ; GFX10: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 ; GFX10: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 ; GFX10: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>) ; GFX10: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>) ; GFX10: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX10: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s64), [[UV2]](s64), 0 ; GFX10: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; GFX10: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; GFX10: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; GFX10: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; GFX10: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; GFX10: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV]](s64), [[UV2]](s64), 1 ; GFX10: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; GFX10: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; GFX10: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; GFX10: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[UV2]](s64), [[UV]](s64) ; GFX10: [[INT7:%[0-9]+]]:_(s64), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s64), [[UV3]](s64), 0 ; GFX10: [[FNEG1:%[0-9]+]]:_(s64) = G_FNEG [[INT7]] ; GFX10: [[INT9:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s64) ; GFX10: [[FMA5:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[INT9]], [[C]] ; GFX10: [[FMA6:%[0-9]+]]:_(s64) = G_FMA [[INT9]], [[FMA5]], [[INT9]] ; GFX10: [[FMA7:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[FMA6]], [[C]] ; GFX10: [[INT10:%[0-9]+]]:_(s64), [[INT11:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[UV1]](s64), [[UV3]](s64), 1 ; GFX10: [[FMA8:%[0-9]+]]:_(s64) = G_FMA [[FMA6]], [[FMA7]], [[FMA6]] ; GFX10: [[FMUL1:%[0-9]+]]:_(s64) = G_FMUL [[INT10]], [[FMA8]] ; GFX10: [[FMA9:%[0-9]+]]:_(s64) = G_FMA [[FNEG1]], [[FMUL1]], [[INT10]] ; GFX10: [[INT12:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s64), [[FMA8]](s64), [[FMUL1]](s64), [[INT11]](s1) ; GFX10: [[INT13:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s64), [[UV3]](s64), [[UV1]](s64) ; GFX10: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[INT6]](s64), [[INT13]](s64) ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>) %0:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 %1:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 %2:_(<2 x s64>) = G_FDIV %0, %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2 ... --- name: test_fdiv_v2s16 body: | bb.0: liveins: $vgpr0, $vgpr1 ; SI-LABEL: name: test_fdiv_v2s16 ; SI: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; SI: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; SI: [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C1]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[FPEXT1]](s32), [[FPEXT]](s32) ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT6]](s32) ; SI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; SI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; SI: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT2]](s32), [[FPEXT3]](s32), 0 ; SI: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT2]](s32), [[FPEXT3]](s32), 1 ; SI: [[INT11:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; SI: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[INT7]] ; SI: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[INT11]], [[C1]] ; SI: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FMA5]], [[INT11]], [[INT11]] ; SI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[INT9]], [[FMA6]] ; SI: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; SI: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; SI: [[FMA9:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; SI: [[INT12:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; SI: [[INT13:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[FPEXT3]](s32), [[FPEXT2]](s32) ; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT13]](s32) ; SI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC]](s16) ; SI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC1]](s16) ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; SI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; SI: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; SI: $vgpr0 = COPY [[BITCAST2]](<2 x s16>) ; VI-LABEL: name: test_fdiv_v2s16 ; VI: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; VI: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; VI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; VI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; VI: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; VI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; VI: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC2]](s16), [[TRUNC]](s16) ; VI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; VI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; VI: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32) ; VI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]] ; VI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; VI: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC3]](s16), [[TRUNC1]](s16) ; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[INT1]](s16) ; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[INT3]](s16) ; VI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; VI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; VI: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; VI: $vgpr0 = COPY [[BITCAST2]](<2 x s16>) ; GFX9-LABEL: name: test_fdiv_v2s16 ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX9: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX9: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX9: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX9: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX9: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX9: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC2]](s16), [[TRUNC]](s16) ; GFX9: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX9: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX9: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32) ; GFX9: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]] ; GFX9: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX9: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC3]](s16), [[TRUNC1]](s16) ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX9: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT3]](s16) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX9: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>) ; GFX9-UNSAFE-LABEL: name: test_fdiv_v2s16 ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX9-UNSAFE: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX9-UNSAFE: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9-UNSAFE: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-UNSAFE: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9-UNSAFE: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX9-UNSAFE: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9-UNSAFE: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9-UNSAFE: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC2]](s16) ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[INT]] ; GFX9-UNSAFE: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC3]](s16) ; GFX9-UNSAFE: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[INT1]] ; GFX9-UNSAFE: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL]](s16) ; GFX9-UNSAFE: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL1]](s16) ; GFX9-UNSAFE: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX9-UNSAFE: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>) ; GFX10-LABEL: name: test_fdiv_v2s16 ; GFX10: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX10: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) ; GFX10: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) ; GFX10: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX10: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX10: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX10: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX10: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX10: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX10: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC2]](s16), [[TRUNC]](s16) ; GFX10: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX10: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX10: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32) ; GFX10: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]] ; GFX10: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX10: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC3]](s16), [[TRUNC1]](s16) ; GFX10: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX10: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT3]](s16) ; GFX10: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX10: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>) %0:_(<2 x s16>) = COPY $vgpr0 %1:_(<2 x s16>) = COPY $vgpr1 %2:_(<2 x s16>) = G_FDIV %0, %1 $vgpr0 = COPY %2 ... --- name: test_fdiv_v3s16 body: | bb.0: liveins: $vgpr0, $vgpr1 ; SI-LABEL: name: test_fdiv_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>) ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>) ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; SI: [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C1]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[FPEXT1]](s32), [[FPEXT]](s32) ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT6]](s32) ; SI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; SI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; SI: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT2]](s32), [[FPEXT3]](s32), 0 ; SI: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT2]](s32), [[FPEXT3]](s32), 1 ; SI: [[INT11:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; SI: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[INT7]] ; SI: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[INT11]], [[C1]] ; SI: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FMA5]], [[INT11]], [[INT11]] ; SI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[INT9]], [[FMA6]] ; SI: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; SI: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; SI: [[FMA9:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; SI: [[INT12:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; SI: [[INT13:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[FPEXT3]](s32), [[FPEXT2]](s32) ; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT13]](s32) ; SI: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; SI: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; SI: [[INT14:%[0-9]+]]:_(s32), [[INT15:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT4]](s32), [[FPEXT5]](s32), 0 ; SI: [[INT16:%[0-9]+]]:_(s32), [[INT17:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT4]](s32), [[FPEXT5]](s32), 1 ; SI: [[INT18:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT14]](s32) ; SI: [[FNEG2:%[0-9]+]]:_(s32) = G_FNEG [[INT14]] ; SI: [[FMA10:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[INT18]], [[C1]] ; SI: [[FMA11:%[0-9]+]]:_(s32) = G_FMA [[FMA10]], [[INT18]], [[INT18]] ; SI: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[INT16]], [[FMA11]] ; SI: [[FMA12:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMUL2]], [[INT16]] ; SI: [[FMA13:%[0-9]+]]:_(s32) = G_FMA [[FMA12]], [[FMA11]], [[FMUL2]] ; SI: [[FMA14:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMA13]], [[INT16]] ; SI: [[INT19:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA14]](s32), [[FMA11]](s32), [[FMA13]](s32), [[INT17]](s1) ; SI: [[INT20:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT19]](s32), [[FPEXT5]](s32), [[FPEXT4]](s32) ; SI: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT20]](s32) ; SI: [[DEF3:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; SI: [[DEF4:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC]](s16) ; SI: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC1]](s16) ; SI: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC2]](s16) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32) ; SI: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>) ; VI-LABEL: name: test_fdiv_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>) ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>) ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; VI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; VI: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; VI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; VI: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC3]](s16), [[TRUNC]](s16) ; VI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; VI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; VI: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32) ; VI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]] ; VI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; VI: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC4]](s16), [[TRUNC1]](s16) ; VI: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; VI: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; VI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT5]](s32) ; VI: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[INT4]] ; VI: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32) ; VI: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[TRUNC5]](s16), [[TRUNC2]](s16) ; VI: [[DEF3:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; VI: [[DEF4:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; VI: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT3]](s16) ; VI: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[INT5]](s16) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32) ; VI: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>) ; GFX9-LABEL: name: test_fdiv_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX9: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>) ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX9: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX9: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX9: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX9: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX9: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX9: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX9: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX9: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC3]](s16), [[TRUNC]](s16) ; GFX9: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX9: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; GFX9: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32) ; GFX9: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]] ; GFX9: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX9: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC4]](s16), [[TRUNC1]](s16) ; GFX9: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX9: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; GFX9: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT5]](s32) ; GFX9: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[INT4]] ; GFX9: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32) ; GFX9: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[TRUNC5]](s16), [[TRUNC2]](s16) ; GFX9: [[DEF3:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX9: [[DEF4:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX9: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT3]](s16) ; GFX9: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[INT5]](s16) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32) ; GFX9: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>) ; GFX9-UNSAFE-LABEL: name: test_fdiv_v3s16 ; GFX9-UNSAFE: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; GFX9-UNSAFE: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; GFX9-UNSAFE: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; GFX9-UNSAFE: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>) ; GFX9-UNSAFE: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9-UNSAFE: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-UNSAFE: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9-UNSAFE: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX9-UNSAFE: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9-UNSAFE: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9-UNSAFE: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>) ; GFX9-UNSAFE: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX9-UNSAFE: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX9-UNSAFE: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX9-UNSAFE: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX9-UNSAFE: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC3]](s16) ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[INT]] ; GFX9-UNSAFE: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC4]](s16) ; GFX9-UNSAFE: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[INT1]] ; GFX9-UNSAFE: [[INT2:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC5]](s16) ; GFX9-UNSAFE: [[FMUL2:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC2]], [[INT2]] ; GFX9-UNSAFE: [[DEF3:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX9-UNSAFE: [[DEF4:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; GFX9-UNSAFE: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL]](s16) ; GFX9-UNSAFE: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL1]](s16) ; GFX9-UNSAFE: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL2]](s16) ; GFX9-UNSAFE: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32) ; GFX9-UNSAFE: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>) ; GFX10-LABEL: name: test_fdiv_v3s16 ; GFX10: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; GFX10: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; GFX10: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; GFX10: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>) ; GFX10: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX10: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX10: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX10: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX10: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>) ; GFX10: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX10: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX10: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX10: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX10: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX10: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX10: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX10: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX10: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX10: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC3]](s16), [[TRUNC]](s16) ; GFX10: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX10: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; GFX10: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32) ; GFX10: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]] ; GFX10: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX10: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC4]](s16), [[TRUNC1]](s16) ; GFX10: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX10: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; GFX10: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT5]](s32) ; GFX10: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[INT4]] ; GFX10: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32) ; GFX10: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[TRUNC5]](s16), [[TRUNC2]](s16) ; GFX10: [[DEF3:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10: [[DEF4:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; GFX10: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX10: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT3]](s16) ; GFX10: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[INT5]](s16) ; GFX10: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32) ; GFX10: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF %2:_(<3 x s16>) = G_FDIV %0, %1 %3:_(<3 x s32>) = G_ANYEXT %2 S_NOP 0, implicit %3 ... --- name: test_fdiv_v4s16 body: | bb.0: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 ; SI-LABEL: name: test_fdiv_v4s16 ; SI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; SI: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; SI: [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C1]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[FPEXT1]](s32), [[FPEXT]](s32) ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT6]](s32) ; SI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; SI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; SI: [[INT7:%[0-9]+]]:_(s32), [[INT8:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT2]](s32), [[FPEXT3]](s32), 0 ; SI: [[INT9:%[0-9]+]]:_(s32), [[INT10:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT2]](s32), [[FPEXT3]](s32), 1 ; SI: [[INT11:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT7]](s32) ; SI: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[INT7]] ; SI: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[INT11]], [[C1]] ; SI: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FMA5]], [[INT11]], [[INT11]] ; SI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[INT9]], [[FMA6]] ; SI: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMUL1]], [[INT9]] ; SI: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FMA7]], [[FMA6]], [[FMUL1]] ; SI: [[FMA9:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA8]], [[INT9]] ; SI: [[INT12:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA9]](s32), [[FMA6]](s32), [[FMA8]](s32), [[INT10]](s1) ; SI: [[INT13:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT12]](s32), [[FPEXT3]](s32), [[FPEXT2]](s32) ; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT13]](s32) ; SI: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; SI: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC6]](s16) ; SI: [[INT14:%[0-9]+]]:_(s32), [[INT15:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT4]](s32), [[FPEXT5]](s32), 0 ; SI: [[INT16:%[0-9]+]]:_(s32), [[INT17:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT4]](s32), [[FPEXT5]](s32), 1 ; SI: [[INT18:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT14]](s32) ; SI: [[FNEG2:%[0-9]+]]:_(s32) = G_FNEG [[INT14]] ; SI: [[FMA10:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[INT18]], [[C1]] ; SI: [[FMA11:%[0-9]+]]:_(s32) = G_FMA [[FMA10]], [[INT18]], [[INT18]] ; SI: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[INT16]], [[FMA11]] ; SI: [[FMA12:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMUL2]], [[INT16]] ; SI: [[FMA13:%[0-9]+]]:_(s32) = G_FMA [[FMA12]], [[FMA11]], [[FMUL2]] ; SI: [[FMA14:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMA13]], [[INT16]] ; SI: [[INT19:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA14]](s32), [[FMA11]](s32), [[FMA13]](s32), [[INT17]](s1) ; SI: [[INT20:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT19]](s32), [[FPEXT5]](s32), [[FPEXT4]](s32) ; SI: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT20]](s32) ; SI: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; SI: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC7]](s16) ; SI: [[INT21:%[0-9]+]]:_(s32), [[INT22:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT6]](s32), [[FPEXT7]](s32), 0 ; SI: [[INT23:%[0-9]+]]:_(s32), [[INT24:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT6]](s32), [[FPEXT7]](s32), 1 ; SI: [[INT25:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT21]](s32) ; SI: [[FNEG3:%[0-9]+]]:_(s32) = G_FNEG [[INT21]] ; SI: [[FMA15:%[0-9]+]]:_(s32) = G_FMA [[FNEG3]], [[INT25]], [[C1]] ; SI: [[FMA16:%[0-9]+]]:_(s32) = G_FMA [[FMA15]], [[INT25]], [[INT25]] ; SI: [[FMUL3:%[0-9]+]]:_(s32) = G_FMUL [[INT23]], [[FMA16]] ; SI: [[FMA17:%[0-9]+]]:_(s32) = G_FMA [[FNEG3]], [[FMUL3]], [[INT23]] ; SI: [[FMA18:%[0-9]+]]:_(s32) = G_FMA [[FMA17]], [[FMA16]], [[FMUL3]] ; SI: [[FMA19:%[0-9]+]]:_(s32) = G_FMA [[FNEG3]], [[FMA18]], [[INT23]] ; SI: [[INT26:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA19]](s32), [[FMA16]](s32), [[FMA18]](s32), [[INT24]](s1) ; SI: [[INT27:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT26]](s32), [[FPEXT7]](s32), [[FPEXT6]](s32) ; SI: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT27]](s32) ; SI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC]](s16) ; SI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC1]](s16) ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; SI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; SI: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; SI: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC2]](s16) ; SI: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC3]](s16) ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) ; SI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) ; VI-LABEL: name: test_fdiv_v4s16 ; VI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; VI: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; VI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; VI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; VI: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; VI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; VI: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC4]](s16), [[TRUNC]](s16) ; VI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; VI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; VI: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32) ; VI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]] ; VI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; VI: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC5]](s16), [[TRUNC1]](s16) ; VI: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; VI: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC6]](s16) ; VI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT5]](s32) ; VI: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[INT4]] ; VI: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32) ; VI: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[TRUNC6]](s16), [[TRUNC2]](s16) ; VI: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; VI: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC7]](s16) ; VI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT7]](s32) ; VI: [[FMUL3:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT6]], [[INT6]] ; VI: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL3]](s32) ; VI: [[INT7:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC3]](s16), [[TRUNC7]](s16), [[TRUNC3]](s16) ; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[INT1]](s16) ; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[INT3]](s16) ; VI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; VI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] ; VI: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; VI: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[INT5]](s16) ; VI: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[INT7]](s16) ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) ; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) ; GFX9-LABEL: name: test_fdiv_v4s16 ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX9: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX9: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX9: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX9: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX9: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX9: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX9: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; GFX9: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX9: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX9: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC4]](s16), [[TRUNC]](s16) ; GFX9: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX9: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; GFX9: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32) ; GFX9: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]] ; GFX9: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX9: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC5]](s16), [[TRUNC1]](s16) ; GFX9: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX9: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC6]](s16) ; GFX9: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT5]](s32) ; GFX9: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[INT4]] ; GFX9: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32) ; GFX9: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[TRUNC6]](s16), [[TRUNC2]](s16) ; GFX9: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX9: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC7]](s16) ; GFX9: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT7]](s32) ; GFX9: [[FMUL3:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT6]], [[INT6]] ; GFX9: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL3]](s32) ; GFX9: [[INT7:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC3]](s16), [[TRUNC7]](s16), [[TRUNC3]](s16) ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX9: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT3]](s16) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX9: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[INT5]](s16) ; GFX9: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[INT7]](s16) ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[ANYEXT3]](s32) ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX9: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) ; GFX9-UNSAFE-LABEL: name: test_fdiv_v4s16 ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX9-UNSAFE: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX9-UNSAFE: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX9-UNSAFE: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9-UNSAFE: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-UNSAFE: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9-UNSAFE: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX9-UNSAFE: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9-UNSAFE: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9-UNSAFE: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX9-UNSAFE: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX9-UNSAFE: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX9-UNSAFE: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX9-UNSAFE: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX9-UNSAFE: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX9-UNSAFE: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX9-UNSAFE: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC4]](s16) ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[INT]] ; GFX9-UNSAFE: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC5]](s16) ; GFX9-UNSAFE: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[INT1]] ; GFX9-UNSAFE: [[INT2:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC6]](s16) ; GFX9-UNSAFE: [[FMUL2:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC2]], [[INT2]] ; GFX9-UNSAFE: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC7]](s16) ; GFX9-UNSAFE: [[FMUL3:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC3]], [[INT3]] ; GFX9-UNSAFE: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL]](s16) ; GFX9-UNSAFE: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL1]](s16) ; GFX9-UNSAFE: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX9-UNSAFE: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL2]](s16) ; GFX9-UNSAFE: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL3]](s16) ; GFX9-UNSAFE: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[ANYEXT3]](s32) ; GFX9-UNSAFE: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX9-UNSAFE: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) ; GFX10-LABEL: name: test_fdiv_v4s16 ; GFX10: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 ; GFX10: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 ; GFX10: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) ; GFX10: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX10: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX10: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX10: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX10: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) ; GFX10: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX10: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX10: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX10: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX10: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX10: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX10: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) ; GFX10: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX10: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX10: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC4]](s16), [[TRUNC]](s16) ; GFX10: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) ; GFX10: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) ; GFX10: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32) ; GFX10: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]] ; GFX10: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) ; GFX10: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC5]](s16), [[TRUNC1]](s16) ; GFX10: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; GFX10: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC6]](s16) ; GFX10: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT5]](s32) ; GFX10: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[INT4]] ; GFX10: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32) ; GFX10: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[TRUNC6]](s16), [[TRUNC2]](s16) ; GFX10: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) ; GFX10: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC7]](s16) ; GFX10: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT7]](s32) ; GFX10: [[FMUL3:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT6]], [[INT6]] ; GFX10: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL3]](s32) ; GFX10: [[INT7:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC3]](s16), [[TRUNC7]](s16), [[TRUNC3]](s16) ; GFX10: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX10: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT3]](s16) ; GFX10: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX10: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[INT5]](s16) ; GFX10: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[INT7]](s16) ; GFX10: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[ANYEXT3]](s32) ; GFX10: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX10: $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>) = G_FDIV %0, %1 $vgpr0_vgpr1 = COPY %2 ... --- name: test_fdiv_s16_constant_one_rcp body: | bb.0: liveins: $vgpr0 ; SI-LABEL: name: test_fdiv_s16_constant_one_rcp ; SI: [[C:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xH3C00 ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[C]](s16) ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; SI: [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C1]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[FPEXT1]](s32), [[FPEXT]](s32) ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT6]](s32) ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC]](s16) ; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI-LABEL: name: test_fdiv_s16_constant_one_rcp ; VI: [[C:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xH3C00 ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; VI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[C]](s16) ; VI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; VI: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; VI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; VI: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC]](s16), [[C]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-LABEL: name: test_fdiv_s16_constant_one_rcp ; GFX9: [[C:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xH3C00 ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX9: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[C]](s16) ; GFX9: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX9: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX9: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX9: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC]](s16), [[C]](s16) ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX9: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-UNSAFE-LABEL: name: test_fdiv_s16_constant_one_rcp ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9-UNSAFE: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC]](s16) ; GFX9-UNSAFE: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT]](s16) ; GFX9-UNSAFE: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX10-LABEL: name: test_fdiv_s16_constant_one_rcp ; GFX10: [[C:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xH3C00 ; GFX10: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX10: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[C]](s16) ; GFX10: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX10: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX10: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX10: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC]](s16), [[C]](s16) ; GFX10: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX10: $vgpr0 = COPY [[ANYEXT]](s32) %0:_(s16) = G_FCONSTANT half 1.0 %1:_(s32) = COPY $vgpr0 %2:_(s16) = G_TRUNC %1 %3:_(s16) = G_FDIV %0, %2 %4:_(s32) = G_ANYEXT %3 $vgpr0 = COPY %4 ... --- name: test_fdiv_s16_constant_negative_one_rcp body: | bb.0: liveins: $vgpr0 ; SI-LABEL: name: test_fdiv_s16_constant_negative_one_rcp ; SI: [[C:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xHBC00 ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[C]](s16) ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; SI: [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[FPEXT]](s32), [[FPEXT1]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C1]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[FPEXT1]](s32), [[FPEXT]](s32) ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT6]](s32) ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC]](s16) ; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI-LABEL: name: test_fdiv_s16_constant_negative_one_rcp ; VI: [[C:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xHBC00 ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; VI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[C]](s16) ; VI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; VI: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; VI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; VI: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC]](s16), [[C]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-LABEL: name: test_fdiv_s16_constant_negative_one_rcp ; GFX9: [[C:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xHBC00 ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX9: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[C]](s16) ; GFX9: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX9: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX9: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX9: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC]](s16), [[C]](s16) ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX9: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-UNSAFE-LABEL: name: test_fdiv_s16_constant_negative_one_rcp ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9-UNSAFE: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX9-UNSAFE: [[FNEG:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC]] ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FNEG]](s16) ; GFX9-UNSAFE: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT]](s16) ; GFX9-UNSAFE: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX10-LABEL: name: test_fdiv_s16_constant_negative_one_rcp ; GFX10: [[C:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xHBC00 ; GFX10: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) ; GFX10: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[C]](s16) ; GFX10: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) ; GFX10: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32) ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]] ; GFX10: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) ; GFX10: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC]](s16), [[C]](s16) ; GFX10: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX10: $vgpr0 = COPY [[ANYEXT]](s32) %0:_(s16) = G_FCONSTANT half -1.0 %1:_(s32) = COPY $vgpr0 %2:_(s16) = G_TRUNC %1 %3:_(s16) = G_FDIV %0, %2 %4:_(s32) = G_ANYEXT %3 $vgpr0 = COPY %4 ... --- name: test_fdiv_s32_constant_one_rcp body: | bb.0: liveins: $vgpr0 ; SI-LABEL: name: test_fdiv_s32_constant_one_rcp ; SI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY]](s32), [[C]](s32) ; SI: $vgpr0 = COPY [[INT6]](s32) ; VI-LABEL: name: test_fdiv_s32_constant_one_rcp ; VI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; VI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 0 ; VI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 1 ; VI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; VI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; VI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; VI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; VI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; VI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; VI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; VI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; VI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY]](s32), [[C]](s32) ; VI: $vgpr0 = COPY [[INT6]](s32) ; GFX9-LABEL: name: test_fdiv_s32_constant_one_rcp ; GFX9: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 0 ; GFX9: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 1 ; GFX9: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX9: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX9: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX9: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX9: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX9: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX9: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX9: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY]](s32), [[C]](s32) ; GFX9: $vgpr0 = COPY [[INT6]](s32) ; GFX9-UNSAFE-LABEL: name: test_fdiv_s32_constant_one_rcp ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[COPY]](s32) ; GFX9-UNSAFE: $vgpr0 = COPY [[INT]](s32) ; GFX10-LABEL: name: test_fdiv_s32_constant_one_rcp ; GFX10: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX10: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 0 ; GFX10: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 1 ; GFX10: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX10: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX10: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C]] ; GFX10: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX10: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX10: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX10: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX10: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY]](s32), [[C]](s32) ; GFX10: $vgpr0 = COPY [[INT6]](s32) %0:_(s32) = G_FCONSTANT float 1.0 %1:_(s32) = COPY $vgpr0 %2:_(s32) = G_FDIV %0, %1 $vgpr0 = COPY %2 ... --- name: test_fdiv_s32_constant_negative_one_rcp body: | bb.0: liveins: $vgpr0 ; SI-LABEL: name: test_fdiv_s32_constant_negative_one_rcp ; SI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float -1.000000e+00 ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; SI: [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 0 ; SI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 1 ; SI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; SI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C1]] ; SI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; SI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; SI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; SI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; SI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; SI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY]](s32), [[C]](s32) ; SI: $vgpr0 = COPY [[INT6]](s32) ; VI-LABEL: name: test_fdiv_s32_constant_negative_one_rcp ; VI: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float -1.000000e+00 ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; VI: [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; VI: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 0 ; VI: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 1 ; VI: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; VI: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; VI: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C1]] ; VI: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; VI: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; VI: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; VI: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; VI: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; VI: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY]](s32), [[C]](s32) ; VI: $vgpr0 = COPY [[INT6]](s32) ; GFX9-LABEL: name: test_fdiv_s32_constant_negative_one_rcp ; GFX9: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float -1.000000e+00 ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX9: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 0 ; GFX9: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 1 ; GFX9: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX9: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX9: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C1]] ; GFX9: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX9: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX9: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX9: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX9: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY]](s32), [[C]](s32) ; GFX9: $vgpr0 = COPY [[INT6]](s32) ; GFX9-UNSAFE-LABEL: name: test_fdiv_s32_constant_negative_one_rcp ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX9-UNSAFE: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[COPY]] ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FNEG]](s32) ; GFX9-UNSAFE: $vgpr0 = COPY [[INT]](s32) ; GFX10-LABEL: name: test_fdiv_s32_constant_negative_one_rcp ; GFX10: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float -1.000000e+00 ; GFX10: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10: [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00 ; GFX10: [[INT:%[0-9]+]]:_(s32), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 0 ; GFX10: [[INT2:%[0-9]+]]:_(s32), [[INT3:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s32), [[COPY]](s32), 1 ; GFX10: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s32) ; GFX10: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[INT]] ; GFX10: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[INT4]], [[C1]] ; GFX10: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT4]], [[INT4]] ; GFX10: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[INT2]], [[FMA1]] ; GFX10: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMUL]], [[INT2]] ; GFX10: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FMA2]], [[FMA1]], [[FMUL]] ; GFX10: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA3]], [[INT2]] ; GFX10: [[INT5:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s32), [[FMA1]](s32), [[FMA3]](s32), [[INT3]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s32), [[COPY]](s32), [[C]](s32) ; GFX10: $vgpr0 = COPY [[INT6]](s32) %0:_(s32) = G_FCONSTANT float -1.0 %1:_(s32) = COPY $vgpr0 %2:_(s32) = G_FDIV %0, %1 $vgpr0 = COPY %2 ... --- name: test_fdiv_s64_constant_one_rcp machineFunctionInfo: mode: fp32-input-denormals: false fp32-output-denormals: false fp64-fp16-output-denormals: true fp64-fp16-input-denormals: true body: | bb.0: liveins: $vgpr0_vgpr1 ; SI-LABEL: name: test_fdiv_s64_constant_one_rcp ; SI: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; SI: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; SI: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 0 ; SI: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; SI: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; SI: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; SI: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; SI: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; SI: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 1 ; SI: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; SI: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; SI: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C]](s64) ; SI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](s64) ; SI: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[INT]](s64) ; SI: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[INT3]](s64) ; SI: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[UV1]](s32), [[UV7]] ; SI: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[UV3]](s32), [[UV5]] ; SI: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP]], [[ICMP1]] ; SI: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[XOR]](s1) ; SI: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY]](s64), [[C]](s64) ; SI: $vgpr0_vgpr1 = COPY [[INT6]](s64) ; VI-LABEL: name: test_fdiv_s64_constant_one_rcp ; VI: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; VI: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; VI: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 0 ; VI: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; VI: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; VI: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; VI: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; VI: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; VI: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 1 ; VI: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; VI: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; VI: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; VI: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; VI: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY]](s64), [[C]](s64) ; VI: $vgpr0_vgpr1 = COPY [[INT6]](s64) ; GFX9-LABEL: name: test_fdiv_s64_constant_one_rcp ; GFX9: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX9: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; GFX9: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 0 ; GFX9: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; GFX9: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; GFX9: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; GFX9: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; GFX9: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; GFX9: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 1 ; GFX9: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; GFX9: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; GFX9: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; GFX9: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY]](s64), [[C]](s64) ; GFX9: $vgpr0_vgpr1 = COPY [[INT6]](s64) ; GFX9-UNSAFE-LABEL: name: test_fdiv_s64_constant_one_rcp ; GFX9-UNSAFE: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; GFX9-UNSAFE: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[COPY]] ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[COPY]](s64) ; GFX9-UNSAFE: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT]], [[C]] ; GFX9-UNSAFE: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[FMA]], [[INT]], [[INT]] ; GFX9-UNSAFE: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; GFX9-UNSAFE: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA2]], [[FMA1]], [[FMA1]] ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[C]], [[FMA3]] ; GFX9-UNSAFE: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[C]] ; GFX9-UNSAFE: [[FMA5:%[0-9]+]]:_(s64) = G_FMA [[FMA4]], [[FMA3]], [[FMUL]] ; GFX9-UNSAFE: $vgpr0_vgpr1 = COPY [[FMA5]](s64) ; GFX10-LABEL: name: test_fdiv_s64_constant_one_rcp ; GFX10: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX10: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; GFX10: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 0 ; GFX10: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; GFX10: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; GFX10: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C]] ; GFX10: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; GFX10: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C]] ; GFX10: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 1 ; GFX10: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; GFX10: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; GFX10: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; GFX10: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY]](s64), [[C]](s64) ; GFX10: $vgpr0_vgpr1 = COPY [[INT6]](s64) %0:_(s64) = G_FCONSTANT double 1.0 %1:_(s64) = COPY $vgpr0_vgpr1 %2:_(s64) = G_FDIV %0, %1 $vgpr0_vgpr1 = COPY %2 ... --- name: test_fdiv_s64_constant_negative_one_rcp machineFunctionInfo: mode: fp32-input-denormals: false fp32-output-denormals: false fp64-fp16-input-denormals: true fp64-fp16-output-denormals: true body: | bb.0: liveins: $vgpr0_vgpr1 ; SI-LABEL: name: test_fdiv_s64_constant_negative_one_rcp ; SI: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double -1.000000e+00 ; SI: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; SI: [[C1:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; SI: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 0 ; SI: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; SI: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; SI: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C1]] ; SI: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; SI: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C1]] ; SI: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 1 ; SI: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; SI: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; SI: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C]](s64) ; SI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](s64) ; SI: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[INT]](s64) ; SI: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[INT3]](s64) ; SI: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[UV1]](s32), [[UV7]] ; SI: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[UV3]](s32), [[UV5]] ; SI: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP]], [[ICMP1]] ; SI: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[XOR]](s1) ; SI: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY]](s64), [[C]](s64) ; SI: $vgpr0_vgpr1 = COPY [[INT6]](s64) ; VI-LABEL: name: test_fdiv_s64_constant_negative_one_rcp ; VI: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double -1.000000e+00 ; VI: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; VI: [[C1:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; VI: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 0 ; VI: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; VI: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; VI: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C1]] ; VI: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; VI: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C1]] ; VI: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 1 ; VI: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; VI: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; VI: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; VI: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; VI: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY]](s64), [[C]](s64) ; VI: $vgpr0_vgpr1 = COPY [[INT6]](s64) ; GFX9-LABEL: name: test_fdiv_s64_constant_negative_one_rcp ; GFX9: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double -1.000000e+00 ; GFX9: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; GFX9: [[C1:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX9: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 0 ; GFX9: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; GFX9: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; GFX9: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C1]] ; GFX9: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; GFX9: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C1]] ; GFX9: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 1 ; GFX9: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; GFX9: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; GFX9: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; GFX9: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; GFX9: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY]](s64), [[C]](s64) ; GFX9: $vgpr0_vgpr1 = COPY [[INT6]](s64) ; GFX9-UNSAFE-LABEL: name: test_fdiv_s64_constant_negative_one_rcp ; GFX9-UNSAFE: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double -1.000000e+00 ; GFX9-UNSAFE: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; GFX9-UNSAFE: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[COPY]] ; GFX9-UNSAFE: [[C1:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[COPY]](s64) ; GFX9-UNSAFE: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT]], [[C1]] ; GFX9-UNSAFE: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[FMA]], [[INT]], [[INT]] ; GFX9-UNSAFE: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C1]] ; GFX9-UNSAFE: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA2]], [[FMA1]], [[FMA1]] ; GFX9-UNSAFE: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[C]], [[FMA3]] ; GFX9-UNSAFE: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[C]] ; GFX9-UNSAFE: [[FMA5:%[0-9]+]]:_(s64) = G_FMA [[FMA4]], [[FMA3]], [[FMUL]] ; GFX9-UNSAFE: $vgpr0_vgpr1 = COPY [[FMA5]](s64) ; GFX10-LABEL: name: test_fdiv_s64_constant_negative_one_rcp ; GFX10: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double -1.000000e+00 ; GFX10: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 ; GFX10: [[C1:%[0-9]+]]:_(s64) = G_FCONSTANT double 1.000000e+00 ; GFX10: [[INT:%[0-9]+]]:_(s64), [[INT1:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 0 ; GFX10: [[FNEG:%[0-9]+]]:_(s64) = G_FNEG [[INT]] ; GFX10: [[INT2:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[INT]](s64) ; GFX10: [[FMA:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[INT2]], [[C1]] ; GFX10: [[FMA1:%[0-9]+]]:_(s64) = G_FMA [[INT2]], [[FMA]], [[INT2]] ; GFX10: [[FMA2:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMA1]], [[C1]] ; GFX10: [[INT3:%[0-9]+]]:_(s64), [[INT4:%[0-9]+]]:_(s1) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.scale), [[C]](s64), [[COPY]](s64), 1 ; GFX10: [[FMA3:%[0-9]+]]:_(s64) = G_FMA [[FMA1]], [[FMA2]], [[FMA1]] ; GFX10: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[INT3]], [[FMA3]] ; GFX10: [[FMA4:%[0-9]+]]:_(s64) = G_FMA [[FNEG]], [[FMUL]], [[INT3]] ; GFX10: [[INT5:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fmas), [[FMA4]](s64), [[FMA3]](s64), [[FMUL]](s64), [[INT4]](s1) ; GFX10: [[INT6:%[0-9]+]]:_(s64) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[INT5]](s64), [[COPY]](s64), [[C]](s64) ; GFX10: $vgpr0_vgpr1 = COPY [[INT6]](s64) %0:_(s64) = G_FCONSTANT double -1.0 %1:_(s64) = COPY $vgpr0_vgpr1 %2:_(s64) = G_FDIV %0, %1 $vgpr0_vgpr1 = COPY %2 ...