[llvm] r371803 - AMDGPU/GlobalISel: Legalize G_FFLOOR
Matt Arsenault via llvm-commits
llvm-commits at lists.llvm.org
Thu Sep 12 18:48:16 PDT 2019
Author: arsenm
Date: Thu Sep 12 18:48:15 2019
New Revision: 371803
URL: http://llvm.org/viewvc/llvm-project?rev=371803&view=rev
Log:
AMDGPU/GlobalISel: Legalize G_FFLOOR
Added:
llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-ffloor.mir
llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-ffloor.s16.mir
llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/legalize-ffloor.mir
Modified:
llvm/trunk/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
llvm/trunk/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp
Modified: llvm/trunk/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp?rev=371803&r1=371802&r2=371803&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp (original)
+++ llvm/trunk/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp Thu Sep 12 18:48:15 2019
@@ -366,12 +366,12 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo
getActionDefinitionsBuilder({G_FMINIMUM, G_FMAXIMUM}).lower();
if (ST.has16BitInsts()) {
- getActionDefinitionsBuilder(G_FSQRT)
+ getActionDefinitionsBuilder({G_FSQRT, G_FFLOOR})
.legalFor({S32, S64, S16})
.scalarize(0)
.clampScalar(0, S16, S64);
} else {
- getActionDefinitionsBuilder(G_FSQRT)
+ getActionDefinitionsBuilder({G_FSQRT, G_FFLOOR})
.legalFor({S32, S64})
.scalarize(0)
.clampScalar(0, S32, S64);
Modified: llvm/trunk/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp?rev=371803&r1=371802&r2=371803&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp (original)
+++ llvm/trunk/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp Thu Sep 12 18:48:15 2019
@@ -1782,6 +1782,7 @@ AMDGPURegisterBankInfo::getInstrMapping(
case AMDGPU::G_FMA:
case AMDGPU::G_FMAD:
case AMDGPU::G_FSQRT:
+ case AMDGPU::G_FFLOOR:
case AMDGPU::G_SITOFP:
case AMDGPU::G_UITOFP:
case AMDGPU::G_FPTRUNC:
Added: llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-ffloor.mir
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-ffloor.mir?rev=371803&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-ffloor.mir (added)
+++ llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-ffloor.mir Thu Sep 12 18:48:15 2019
@@ -0,0 +1,150 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -march=amdgcn -mcpu=tahiti -run-pass=instruction-select -global-isel-abort=0 -verify-machineinstrs -o - %s | FileCheck %s
+
+---
+name: ffloor_s32_vv
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+ bb.0:
+ liveins: $vgpr0
+
+ ; CHECK-LABEL: name: ffloor_s32_vv
+ ; CHECK: liveins: $vgpr0
+ ; CHECK: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; CHECK: [[V_FLOOR_F32_e64_:%[0-9]+]]:vgpr_32 = V_FLOOR_F32_e64 0, [[COPY]], 0, 0, implicit $exec
+ ; CHECK: $vgpr0 = COPY [[V_FLOOR_F32_e64_]]
+ %0:vgpr(s32) = COPY $vgpr0
+ %1:vgpr(s32) = G_FFLOOR %0
+ $vgpr0 = COPY %1
+...
+
+---
+name: ffloor_s32_vs
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+ bb.0:
+ liveins: $sgpr0
+
+ ; CHECK-LABEL: name: ffloor_s32_vs
+ ; CHECK: liveins: $sgpr0
+ ; CHECK: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
+ ; CHECK: [[V_FLOOR_F32_e64_:%[0-9]+]]:vgpr_32 = V_FLOOR_F32_e64 0, [[COPY]], 0, 0, implicit $exec
+ ; CHECK: $vgpr0 = COPY [[V_FLOOR_F32_e64_]]
+ %0:sgpr(s32) = COPY $sgpr0
+ %1:vgpr(s32) = G_FFLOOR %0
+ $vgpr0 = COPY %1
+...
+
+---
+name: ffloor_s64_vv
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+ bb.0:
+ liveins: $vgpr0_vgpr1
+
+ ; CHECK-LABEL: name: ffloor_s64_vv
+ ; CHECK: liveins: $vgpr0_vgpr1
+ ; CHECK: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+ ; CHECK: [[V_CMP_CLASS_F64_e64_:%[0-9]+]]:sreg_64 = V_CMP_CLASS_F64_e64 0, [[COPY]], 3, implicit $exec
+ ; CHECK: [[V_MOV_B:%[0-9]+]]:vreg_64 = V_MOV_B64_PSEUDO 4607182418800017407, implicit $exec
+ ; CHECK: [[V_FRACT_F64_e64_:%[0-9]+]]:vreg_64 = V_FRACT_F64_e64 0, [[COPY]], 0, 0, implicit $exec
+ ; CHECK: [[V_MIN_F64_:%[0-9]+]]:vreg_64 = V_MIN_F64 0, [[V_FRACT_F64_e64_]], 0, [[V_MOV_B]], 0, 0, implicit $exec
+ ; CHECK: [[V_CNDMA:%[0-9]+]]:vreg_64 = V_CNDMASK_B64_PSEUDO [[V_MIN_F64_]], [[COPY]], [[V_CMP_CLASS_F64_e64_]], implicit $exec
+ ; CHECK: [[V_ADD_F64_:%[0-9]+]]:vreg_64 = V_ADD_F64 0, [[COPY]], 1, [[V_CNDMA]], 0, 0, implicit $exec
+ ; CHECK: $vgpr0_vgpr1 = COPY [[V_ADD_F64_]]
+ %0:vgpr(s64) = COPY $vgpr0_vgpr1
+ %1:vgpr(s64) = G_FFLOOR %0
+ $vgpr0_vgpr1 = COPY %1
+...
+
+# FIXME: Constant bus restriction
+# ---
+# name: ffloor_s64_vs
+# legalized: true
+# regBankSelected: true
+# tracksRegLiveness: true
+
+# body: |
+# bb.0:
+# liveins: $sgpr0_sgpr1
+
+# %0:sgpr(s64) = COPY $sgpr0_sgpr1
+# %1:vgpr(s64) = G_FFLOOR %0
+# $vgpr0_vgpr1 = COPY %1
+# ...
+
+---
+name: ffloor_fneg_s32_vs
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+ bb.0:
+ liveins: $sgpr0
+
+ ; CHECK-LABEL: name: ffloor_fneg_s32_vs
+ ; CHECK: liveins: $sgpr0
+ ; CHECK: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
+ ; CHECK: [[V_FLOOR_F32_e64_:%[0-9]+]]:vgpr_32 = V_FLOOR_F32_e64 1, [[COPY]], 0, 0, implicit $exec
+ ; CHECK: $vgpr0 = COPY [[V_FLOOR_F32_e64_]]
+ %0:sgpr(s32) = COPY $sgpr0
+ %1:sgpr(s32) = G_FNEG %0
+ %2:vgpr(s32) = G_FFLOOR %1
+ $vgpr0 = COPY %2
+...
+
+---
+name: ffloor_fneg_s32_vv
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+ bb.0:
+ liveins: $vgpr0
+ ; CHECK-LABEL: name: ffloor_fneg_s32_vv
+ ; CHECK: liveins: $vgpr0
+ ; CHECK: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; CHECK: [[V_FLOOR_F32_e64_:%[0-9]+]]:vgpr_32 = V_FLOOR_F32_e64 1, [[COPY]], 0, 0, implicit $exec
+ ; CHECK: $vgpr0 = COPY [[V_FLOOR_F32_e64_]]
+ %0:vgpr(s32) = COPY $vgpr0
+ %1:vgpr(s32) = G_FNEG %0
+ %2:vgpr(s32) = G_FFLOOR %1
+ $vgpr0 = COPY %2
+...
+
+---
+name: ffloor_fneg_s64_vv
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+ bb.0:
+ liveins: $vgpr0_vgpr1
+
+ ; CHECK-LABEL: name: ffloor_fneg_s64_vv
+ ; CHECK: liveins: $vgpr0_vgpr1
+ ; CHECK: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+ ; CHECK: [[V_CMP_CLASS_F64_e64_:%[0-9]+]]:sreg_64 = V_CMP_CLASS_F64_e64 0, [[COPY]], 3, implicit $exec
+ ; CHECK: [[V_MOV_B:%[0-9]+]]:vreg_64 = V_MOV_B64_PSEUDO 4607182418800017407, implicit $exec
+ ; CHECK: [[V_FRACT_F64_e64_:%[0-9]+]]:vreg_64 = V_FRACT_F64_e64 1, [[COPY]], 0, 0, implicit $exec
+ ; CHECK: [[V_MIN_F64_:%[0-9]+]]:vreg_64 = V_MIN_F64 0, [[V_FRACT_F64_e64_]], 0, [[V_MOV_B]], 0, 0, implicit $exec
+ ; CHECK: [[V_CNDMA:%[0-9]+]]:vreg_64 = V_CNDMASK_B64_PSEUDO [[V_MIN_F64_]], [[COPY]], [[V_CMP_CLASS_F64_e64_]], implicit $exec
+ ; CHECK: [[V_ADD_F64_:%[0-9]+]]:vreg_64 = V_ADD_F64 1, [[COPY]], 1, [[V_CNDMA]], 0, 0, implicit $exec
+ ; CHECK: $vgpr0_vgpr1 = COPY [[V_ADD_F64_]]
+ %0:vgpr(s64) = COPY $vgpr0_vgpr1
+ %1:vgpr(s64) = G_FNEG %0
+ %2:vgpr(s64) = G_FFLOOR %1
+ $vgpr0_vgpr1 = COPY %2
+...
Added: llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-ffloor.s16.mir
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-ffloor.s16.mir?rev=371803&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-ffloor.s16.mir (added)
+++ llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-ffloor.s16.mir Thu Sep 12 18:48:15 2019
@@ -0,0 +1,101 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -march=amdgcn -mcpu=fiji -run-pass=instruction-select -global-isel-abort=0 -verify-machineinstrs -o - %s | FileCheck -check-prefix=VI %s
+
+---
+name: ffloor_s16_ss
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+ bb.0:
+ liveins: $sgpr0
+
+ ; VI-LABEL: name: ffloor_s16_ss
+ ; VI: liveins: $sgpr0
+ ; VI: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0
+ ; VI: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY]](s32)
+ ; VI: [[FFLOOR:%[0-9]+]]:sreg_32_xm0(s16) = G_FFLOOR [[TRUNC]]
+ ; VI: [[COPY1:%[0-9]+]]:sreg_32_xm0(s32) = COPY [[FFLOOR]](s16)
+ ; VI: $sgpr0 = COPY [[COPY1]](s32)
+ %0:sgpr(s32) = COPY $sgpr0
+ %1:sgpr(s16) = G_TRUNC %0
+ %2:sgpr(s16) = G_FFLOOR %1
+ %3:sgpr(s32) = G_ANYEXT %2
+ $sgpr0 = COPY %3
+...
+
+---
+name: ffloor_s16_vv
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+ bb.0:
+ liveins: $vgpr0
+
+ ; VI-LABEL: name: ffloor_s16_vv
+ ; VI: liveins: $vgpr0
+ ; VI: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; VI: [[V_FLOOR_F16_e64_:%[0-9]+]]:vgpr_32 = V_FLOOR_F16_e64 0, [[COPY]], 0, 0, implicit $exec
+ ; VI: $vgpr0 = COPY [[V_FLOOR_F16_e64_]]
+ %0:vgpr(s32) = COPY $vgpr0
+ %1:vgpr(s16) = G_TRUNC %0
+ %2:vgpr(s16) = G_FFLOOR %1
+ %3:vgpr(s32) = G_ANYEXT %2
+ $vgpr0 = COPY %3
+...
+
+---
+name: ffloor_s16_vs
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+ bb.0:
+ liveins: $sgpr0
+
+ ; VI-LABEL: name: ffloor_s16_vs
+ ; VI: liveins: $sgpr0
+ ; VI: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
+ ; VI: [[V_FLOOR_F16_e64_:%[0-9]+]]:vgpr_32 = V_FLOOR_F16_e64 0, [[COPY]], 0, 0, implicit $exec
+ ; VI: $vgpr0 = COPY [[V_FLOOR_F16_e64_]]
+ %0:sgpr(s32) = COPY $sgpr0
+ %1:sgpr(s16) = G_TRUNC %0
+ %2:vgpr(s16) = G_FFLOOR %1
+ %3:vgpr(s32) = G_ANYEXT %2
+ $vgpr0 = COPY %3
+...
+
+---
+name: ffloor_fneg_s16_vv
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+ bb.0:
+ liveins: $vgpr0
+
+ ; SI-LABEL: name: ffloor_fneg_s16_vv
+ ; SI: liveins: $vgpr0
+ ; SI: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
+ ; SI: [[TRUNC:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY]](s32)
+ ; SI: [[FNEG:%[0-9]+]]:vgpr(s16) = G_FNEG [[TRUNC]]
+ ; SI: [[FFLOOR:%[0-9]+]]:vgpr(s16) = G_FFLOOR [[FNEG]]
+ ; SI: [[ANYEXT:%[0-9]+]]:vgpr(s32) = G_ANYEXT [[FFLOOR]](s16)
+ ; SI: $vgpr0 = COPY [[ANYEXT]](s32)
+ ; VI-LABEL: name: ffloor_fneg_s16_vv
+ ; VI: liveins: $vgpr0
+ ; VI: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; VI: [[V_FLOOR_F16_e64_:%[0-9]+]]:vgpr_32 = V_FLOOR_F16_e64 1, [[COPY]], 0, 0, implicit $exec
+ ; VI: $vgpr0 = COPY [[V_FLOOR_F16_e64_]]
+ %0:vgpr(s32) = COPY $vgpr0
+ %1:vgpr(s16) = G_TRUNC %0
+ %2:vgpr(s16) = G_FNEG %1
+ %3:vgpr(s16) = G_FFLOOR %2
+ %4:vgpr(s32) = G_ANYEXT %3
+ $vgpr0 = COPY %4
+...
Added: llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/legalize-ffloor.mir
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/legalize-ffloor.mir?rev=371803&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/legalize-ffloor.mir (added)
+++ llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/legalize-ffloor.mir Thu Sep 12 18:48:15 2019
@@ -0,0 +1,388 @@
+# 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 %s -o - | FileCheck -check-prefix=GFX9 %s
+
+---
+name: test_ffloor_s32
+body: |
+ bb.0:
+ liveins: $vgpr0
+
+ ; SI-LABEL: name: test_ffloor_s32
+ ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+ ; SI: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[COPY]]
+ ; SI: $vgpr0 = COPY [[FFLOOR]](s32)
+ ; VI-LABEL: name: test_ffloor_s32
+ ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+ ; VI: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[COPY]]
+ ; VI: $vgpr0 = COPY [[FFLOOR]](s32)
+ ; GFX9-LABEL: name: test_ffloor_s32
+ ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+ ; GFX9: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[COPY]]
+ ; GFX9: $vgpr0 = COPY [[FFLOOR]](s32)
+ %0:_(s32) = COPY $vgpr0
+ %1:_(s32) = G_FFLOOR %0
+ $vgpr0 = COPY %1
+...
+
+---
+name: test_ffloor_s64
+body: |
+ bb.0:
+ liveins: $vgpr0
+
+ ; SI-LABEL: name: test_ffloor_s64
+ ; SI: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
+ ; SI: [[FFLOOR:%[0-9]+]]:_(s64) = G_FFLOOR [[COPY]]
+ ; SI: $vgpr0_vgpr1 = COPY [[FFLOOR]](s64)
+ ; VI-LABEL: name: test_ffloor_s64
+ ; VI: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
+ ; VI: [[FFLOOR:%[0-9]+]]:_(s64) = G_FFLOOR [[COPY]]
+ ; VI: $vgpr0_vgpr1 = COPY [[FFLOOR]](s64)
+ ; GFX9-LABEL: name: test_ffloor_s64
+ ; GFX9: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
+ ; GFX9: [[FFLOOR:%[0-9]+]]:_(s64) = G_FFLOOR [[COPY]]
+ ; GFX9: $vgpr0_vgpr1 = COPY [[FFLOOR]](s64)
+ %0:_(s64) = COPY $vgpr0_vgpr1
+ %1:_(s64) = G_FFLOOR %0
+ $vgpr0_vgpr1 = COPY %1
+
+...
+---
+name: test_ffloor_s16
+body: |
+ bb.0:
+ liveins: $vgpr0
+
+ ; SI-LABEL: name: test_ffloor_s16
+ ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+ ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32)
+ ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)
+ ; SI: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[FPEXT]]
+ ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FFLOOR]](s32)
+ ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC]](s16)
+ ; SI: $vgpr0 = COPY [[ANYEXT]](s32)
+ ; VI-LABEL: name: test_ffloor_s16
+ ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+ ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32)
+ ; VI: [[FFLOOR:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC]]
+ ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FFLOOR]](s16)
+ ; VI: $vgpr0 = COPY [[ANYEXT]](s32)
+ ; GFX9-LABEL: name: test_ffloor_s16
+ ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+ ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32)
+ ; GFX9: [[FFLOOR:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC]]
+ ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FFLOOR]](s16)
+ ; GFX9: $vgpr0 = COPY [[ANYEXT]](s32)
+ %0:_(s32) = COPY $vgpr0
+ %1:_(s16) = G_TRUNC %0
+ %2:_(s16) = G_FFLOOR %1
+ %3:_(s32) = G_ANYEXT %2
+ $vgpr0 = COPY %3
+...
+
+---
+name: test_ffloor_v2s32
+body: |
+ bb.0:
+ liveins: $vgpr0_vgpr1
+
+ ; SI-LABEL: name: test_ffloor_v2s32
+ ; SI: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+ ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+ ; SI: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[UV]]
+ ; SI: [[FFLOOR1:%[0-9]+]]:_(s32) = G_FFLOOR [[UV1]]
+ ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FFLOOR]](s32), [[FFLOOR1]](s32)
+ ; SI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ; VI-LABEL: name: test_ffloor_v2s32
+ ; VI: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+ ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+ ; VI: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[UV]]
+ ; VI: [[FFLOOR1:%[0-9]+]]:_(s32) = G_FFLOOR [[UV1]]
+ ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FFLOOR]](s32), [[FFLOOR1]](s32)
+ ; VI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ; GFX9-LABEL: name: test_ffloor_v2s32
+ ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+ ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+ ; GFX9: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[UV]]
+ ; GFX9: [[FFLOOR1:%[0-9]+]]:_(s32) = G_FFLOOR [[UV1]]
+ ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FFLOOR]](s32), [[FFLOOR1]](s32)
+ ; GFX9: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
+ %1:_(<2 x s32>) = G_FFLOOR %0
+ $vgpr0_vgpr1 = COPY %1
+...
+
+---
+name: test_ffloor_v3s32
+body: |
+ bb.0:
+ liveins: $vgpr0_vgpr1_vgpr2
+
+ ; SI-LABEL: name: test_ffloor_v3s32
+ ; SI: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+ ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+ ; SI: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[UV]]
+ ; SI: [[FFLOOR1:%[0-9]+]]:_(s32) = G_FFLOOR [[UV1]]
+ ; SI: [[FFLOOR2:%[0-9]+]]:_(s32) = G_FFLOOR [[UV2]]
+ ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FFLOOR]](s32), [[FFLOOR1]](s32), [[FFLOOR2]](s32)
+ ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
+ ; VI-LABEL: name: test_ffloor_v3s32
+ ; VI: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+ ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+ ; VI: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[UV]]
+ ; VI: [[FFLOOR1:%[0-9]+]]:_(s32) = G_FFLOOR [[UV1]]
+ ; VI: [[FFLOOR2:%[0-9]+]]:_(s32) = G_FFLOOR [[UV2]]
+ ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FFLOOR]](s32), [[FFLOOR1]](s32), [[FFLOOR2]](s32)
+ ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX9-LABEL: name: test_ffloor_v3s32
+ ; GFX9: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+ ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+ ; GFX9: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[UV]]
+ ; GFX9: [[FFLOOR1:%[0-9]+]]:_(s32) = G_FFLOOR [[UV1]]
+ ; GFX9: [[FFLOOR2:%[0-9]+]]:_(s32) = G_FFLOOR [[UV2]]
+ ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FFLOOR]](s32), [[FFLOOR1]](s32), [[FFLOOR2]](s32)
+ ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
+ %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+ %1:_(<3 x s32>) = G_FFLOOR %0
+ $vgpr0_vgpr1_vgpr2 = COPY %1
+...
+
+---
+name: test_ffloor_v2s64
+body: |
+ bb.0:
+ liveins: $vgpr0_vgpr1_vgpr2_vgpr3
+
+ ; SI-LABEL: name: test_ffloor_v2s64
+ ; SI: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
+ ; SI: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>)
+ ; SI: [[FFLOOR:%[0-9]+]]:_(s64) = G_FFLOOR [[UV]]
+ ; SI: [[FFLOOR1:%[0-9]+]]:_(s64) = G_FFLOOR [[UV1]]
+ ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[FFLOOR]](s64), [[FFLOOR1]](s64)
+ ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ; VI-LABEL: name: test_ffloor_v2s64
+ ; VI: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
+ ; VI: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>)
+ ; VI: [[FFLOOR:%[0-9]+]]:_(s64) = G_FFLOOR [[UV]]
+ ; VI: [[FFLOOR1:%[0-9]+]]:_(s64) = G_FFLOOR [[UV1]]
+ ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[FFLOOR]](s64), [[FFLOOR1]](s64)
+ ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ; GFX9-LABEL: name: test_ffloor_v2s64
+ ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
+ ; GFX9: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>)
+ ; GFX9: [[FFLOOR:%[0-9]+]]:_(s64) = G_FFLOOR [[UV]]
+ ; GFX9: [[FFLOOR1:%[0-9]+]]:_(s64) = G_FFLOOR [[UV1]]
+ ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[FFLOOR]](s64), [[FFLOOR1]](s64)
+ ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ %0:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
+ %1:_(<2 x s64>) = G_FFLOOR %0
+ $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
+...
+
+---
+name: test_ffloor_v2s16
+body: |
+ bb.0:
+ liveins: $vgpr0
+
+ ; SI-LABEL: name: test_ffloor_v2s16
+ ; SI: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
+ ; 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: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)
+ ; SI: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[FPEXT]]
+ ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FFLOOR]](s32)
+ ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16)
+ ; SI: [[FFLOOR1:%[0-9]+]]:_(s32) = G_FFLOOR [[FPEXT1]]
+ ; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FFLOOR1]](s32)
+ ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[FPTRUNC]](s16), [[FPTRUNC1]](s16)
+ ; SI: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; VI-LABEL: name: test_ffloor_v2s16
+ ; VI: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
+ ; 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: [[FFLOOR:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC]]
+ ; VI: [[FFLOOR1:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC1]]
+ ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[FFLOOR]](s16), [[FFLOOR1]](s16)
+ ; VI: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX9-LABEL: name: test_ffloor_v2s16
+ ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
+ ; 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: [[FFLOOR:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC]]
+ ; GFX9: [[FFLOOR1:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC1]]
+ ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[FFLOOR]](s16), [[FFLOOR1]](s16)
+ ; GFX9: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ %0:_(<2 x s16>) = COPY $vgpr0
+ %1:_(<2 x s16>) = G_FFLOOR %0
+ $vgpr0 = COPY %1
+...
+
+---
+name: test_ffloor_v3s16
+body: |
+ bb.0:
+
+ ; SI-LABEL: name: test_ffloor_v3s16
+ ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+ ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0
+ ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+ ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0
+ ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<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: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)
+ ; SI: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[FPEXT]]
+ ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FFLOOR]](s32)
+ ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16)
+ ; SI: [[FFLOOR1:%[0-9]+]]:_(s32) = G_FFLOOR [[FPEXT1]]
+ ; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FFLOOR1]](s32)
+ ; SI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16)
+ ; SI: [[FFLOOR2:%[0-9]+]]:_(s32) = G_FFLOOR [[FPEXT2]]
+ ; SI: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FFLOOR2]](s32)
+ ; 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: [[TRUNC3:%[0-9]+]]:_(<3 x s16>) = G_TRUNC [[BUILD_VECTOR]](<3 x s32>)
+ ; SI: S_NOP 0, implicit [[TRUNC3]](<3 x s16>)
+ ; VI-LABEL: name: test_ffloor_v3s16
+ ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+ ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0
+ ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+ ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0
+ ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<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: [[FFLOOR:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC]]
+ ; VI: [[FFLOOR1:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC1]]
+ ; VI: [[FFLOOR2:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC2]]
+ ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FFLOOR]](s16)
+ ; VI: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FFLOOR1]](s16)
+ ; VI: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FFLOOR2]](s16)
+ ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32)
+ ; VI: [[TRUNC3:%[0-9]+]]:_(<3 x s16>) = G_TRUNC [[BUILD_VECTOR]](<3 x s32>)
+ ; VI: S_NOP 0, implicit [[TRUNC3]](<3 x s16>)
+ ; GFX9-LABEL: name: test_ffloor_v3s16
+ ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+ ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0
+ ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+ ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0
+ ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<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: [[FFLOOR:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC]]
+ ; GFX9: [[FFLOOR1:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC1]]
+ ; GFX9: [[FFLOOR2:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC2]]
+ ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FFLOOR]](s16)
+ ; GFX9: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FFLOOR1]](s16)
+ ; GFX9: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FFLOOR2]](s16)
+ ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32)
+ ; GFX9: [[TRUNC3:%[0-9]+]]:_(<3 x s16>) = G_TRUNC [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX9: S_NOP 0, implicit [[TRUNC3]](<3 x s16>)
+ %0:_(<3 x s16>) = G_IMPLICIT_DEF
+ %1:_(<3 x s16>) = G_FFLOOR %0
+ S_NOP 0, implicit %1
+...
+
+---
+name: test_ffloor_v4s16
+body: |
+ bb.0:
+ liveins: $vgpr0_vgpr1
+
+ ; SI-LABEL: name: test_ffloor_v4s16
+ ; SI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
+ ; 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: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)
+ ; SI: [[FFLOOR:%[0-9]+]]:_(s32) = G_FFLOOR [[FPEXT]]
+ ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FFLOOR]](s32)
+ ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16)
+ ; SI: [[FFLOOR1:%[0-9]+]]:_(s32) = G_FFLOOR [[FPEXT1]]
+ ; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FFLOOR1]](s32)
+ ; SI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16)
+ ; SI: [[FFLOOR2:%[0-9]+]]:_(s32) = G_FFLOOR [[FPEXT2]]
+ ; SI: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FFLOOR2]](s32)
+ ; SI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16)
+ ; SI: [[FFLOOR3:%[0-9]+]]:_(s32) = G_FFLOOR [[FPEXT3]]
+ ; SI: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FFLOOR3]](s32)
+ ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[FPTRUNC]](s16), [[FPTRUNC1]](s16), [[FPTRUNC2]](s16), [[FPTRUNC3]](s16)
+ ; SI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<4 x s16>)
+ ; VI-LABEL: name: test_ffloor_v4s16
+ ; VI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
+ ; 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: [[FFLOOR:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC]]
+ ; VI: [[FFLOOR1:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC1]]
+ ; VI: [[FFLOOR2:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC2]]
+ ; VI: [[FFLOOR3:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC3]]
+ ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[FFLOOR]](s16), [[FFLOOR1]](s16), [[FFLOOR2]](s16), [[FFLOOR3]](s16)
+ ; VI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<4 x s16>)
+ ; GFX9-LABEL: name: test_ffloor_v4s16
+ ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
+ ; 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: [[FFLOOR:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC]]
+ ; GFX9: [[FFLOOR1:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC1]]
+ ; GFX9: [[FFLOOR2:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC2]]
+ ; GFX9: [[FFLOOR3:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC3]]
+ ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[FFLOOR]](s16), [[FFLOOR1]](s16), [[FFLOOR2]](s16), [[FFLOOR3]](s16)
+ ; GFX9: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<4 x s16>)
+ %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
+ %1:_(<4 x s16>) = G_FFLOOR %0
+ $vgpr0_vgpr1 = COPY %1
+...
More information about the llvm-commits
mailing list