nils.hoelscher
/
llvm-for-llvmta
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
llvm-for-llvmta/test/CodeGen/AMDGPU/GlobalISel/legalize-fmul.mir

614 lines
35 KiB
Plaintext
Raw Normal View History

first commit
2022-04-25 10:02:23 +02:00
# 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_fmul_s32
body: |
bb.0:
liveins: $vgpr0, $vgpr1
; SI-LABEL: name: test_fmul_s32
; SI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[COPY]], [[COPY1]]
; SI: $vgpr0 = COPY [[FMUL]](s32)
; VI-LABEL: name: test_fmul_s32
; VI: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; VI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[COPY]], [[COPY1]]
; VI: $vgpr0 = COPY [[FMUL]](s32)
; GFX9-LABEL: name: test_fmul_s32
; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX9: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[COPY]], [[COPY1]]
; GFX9: $vgpr0 = COPY [[FMUL]](s32)
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
%2:_(s32) = G_FMUL %0, %1
$vgpr0 = COPY %2
...
---
name: test_fmul_s64
body: |
bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; SI-LABEL: name: test_fmul_s64
; SI: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; SI: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[COPY]], [[COPY1]]
; SI: $vgpr0_vgpr1 = COPY [[FMUL]](s64)
; VI-LABEL: name: test_fmul_s64
; VI: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; VI: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[COPY]], [[COPY1]]
; VI: $vgpr0_vgpr1 = COPY [[FMUL]](s64)
; GFX9-LABEL: name: test_fmul_s64
; GFX9: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; GFX9: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[COPY]], [[COPY1]]
; GFX9: $vgpr0_vgpr1 = COPY [[FMUL]](s64)
%0:_(s64) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
%2:_(s64) = G_FMUL %0, %1
$vgpr0_vgpr1 = COPY %2
...
---
name: test_fmul_s16
body: |
bb.0:
liveins: $vgpr0, $vgpr1
; SI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[FPEXT1]]
; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32)
; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC]](s16)
; SI: $vgpr0 = COPY [[ANYEXT]](s32)
; VI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC1]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL]](s16)
; VI: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC1]]
; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL]](s16)
; GFX9: $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_FMUL %2, %3
%5:_(s32) = G_ANYEXT %4
$vgpr0 = COPY %5
...
---
name: test_fmul_v2s32
body: |
bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; SI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[UV]], [[UV2]]
; SI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[UV1]], [[UV3]]
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32)
; SI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
; VI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[UV]], [[UV2]]
; VI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[UV1]], [[UV3]]
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32)
; VI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
; GFX9-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[UV]], [[UV2]]
; GFX9: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[UV1]], [[UV3]]
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32)
; GFX9: $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_FMUL %0, %1
$vgpr0_vgpr1 = COPY %2
...
---
name: test_fmul_v2s32_flags
body: |
bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; SI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = nnan G_FMUL [[UV]], [[UV2]]
; SI: [[FMUL1:%[0-9]+]]:_(s32) = nnan G_FMUL [[UV1]], [[UV3]]
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32)
; SI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
; VI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = nnan G_FMUL [[UV]], [[UV2]]
; VI: [[FMUL1:%[0-9]+]]:_(s32) = nnan G_FMUL [[UV1]], [[UV3]]
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32)
; VI: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
; GFX9-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = nnan G_FMUL [[UV]], [[UV2]]
; GFX9: [[FMUL1:%[0-9]+]]:_(s32) = nnan G_FMUL [[UV1]], [[UV3]]
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32)
; GFX9: $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_FMUL %0, %1
$vgpr0_vgpr1 = COPY %2
...
---
name: test_fmul_v3s32
body: |
bb.0:
liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[UV]], [[UV3]]
; SI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[UV1]], [[UV4]]
; SI: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[UV2]], [[UV5]]
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32), [[FMUL2]](s32)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
; VI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[UV]], [[UV3]]
; VI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[UV1]], [[UV4]]
; VI: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[UV2]], [[UV5]]
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32), [[FMUL2]](s32)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
; GFX9-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[UV]], [[UV3]]
; GFX9: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[UV1]], [[UV4]]
; GFX9: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[UV2]], [[UV5]]
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FMUL]](s32), [[FMUL1]](s32), [[FMUL2]](s32)
; GFX9: $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_FMUL %0, %1
$vgpr0_vgpr1_vgpr2 = COPY %2
...
---
name: test_fmul_v2s64
body: |
bb.0:
liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5_vgpr6_vgpr7
; SI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[UV]], [[UV2]]
; SI: [[FMUL1:%[0-9]+]]:_(s64) = G_FMUL [[UV1]], [[UV3]]
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[FMUL]](s64), [[FMUL1]](s64)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
; VI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[UV]], [[UV2]]
; VI: [[FMUL1:%[0-9]+]]:_(s64) = G_FMUL [[UV1]], [[UV3]]
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[FMUL]](s64), [[FMUL1]](s64)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
; GFX9-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s64) = G_FMUL [[UV]], [[UV2]]
; GFX9: [[FMUL1:%[0-9]+]]:_(s64) = G_FMUL [[UV1]], [[UV3]]
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[FMUL]](s64), [[FMUL1]](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>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7
%2:_(<2 x s64>) = G_FMUL %0, %1
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
...
---
name: test_fmul_v2s16
body: |
bb.0:
liveins: $vgpr0, $vgpr1
; SI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[FPEXT1]]
; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32)
; SI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16)
; SI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16)
; SI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[FPEXT3]]
; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](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_fmul_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: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC2]]
; VI: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[TRUNC3]]
; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FMUL]](s16)
; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FMUL1]](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_fmul_v2s16
; GFX9: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX9: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX9: [[FMUL:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[COPY]], [[COPY1]]
; GFX9: $vgpr0 = COPY [[FMUL]](<2 x s16>)
%0:_(<2 x s16>) = COPY $vgpr0
%1:_(<2 x s16>) = COPY $vgpr1
%2:_(<2 x s16>) = G_FMUL %0, %1
$vgpr0 = COPY %2
...
---
name: test_fmul_v3s16
body: |
bb.0:
liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_fmul_v3s16
; SI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
; SI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 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: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<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 [[UV4]](<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: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[FPEXT1]]
; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32)
; SI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16)
; SI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16)
; SI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[FPEXT3]]
; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32)
; SI: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16)
; SI: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16)
; SI: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[FPEXT5]]
; SI: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32)
; SI: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; SI: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; SI: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; SI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>)
; SI: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
; SI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
; SI: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
; SI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](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: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; SI: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC2]](s16)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32)
; SI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C1]]
; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
; SI: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C1]]
; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32)
; SI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C1]]
; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL2]]
; SI: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
; VI-LABEL: name: test_fmul_v3s16
; VI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
; VI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 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: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<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 [[UV4]](<2 x s16>)
; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
; VI: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC3]]
; VI: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[TRUNC4]]
; VI: [[FMUL2:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC2]], [[TRUNC5]]
; VI: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; VI: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; VI: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; VI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>)
; VI: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
; VI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
; VI: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
; VI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FMUL]](s16)
; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FMUL1]](s16)
; VI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
; VI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]]
; VI: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; VI: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FMUL2]](s16)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32)
; VI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C1]]
; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
; VI: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; VI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C1]]
; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32)
; VI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C1]]
; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL2]]
; VI: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
; GFX9-LABEL: name: test_fmul_v3s16
; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
; GFX9: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32)
; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[COPY3]](s32)
; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32)
; GFX9: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY4]](s32), [[DEF]](s32)
; GFX9: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
; GFX9: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32)
; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[COPY6]](s32)
; GFX9: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32)
; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[DEF]](s32)
; GFX9: [[FMUL:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
; GFX9: [[FMUL1:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; GFX9: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; GFX9: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[FMUL]](<2 x s16>)
; GFX9: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
; GFX9: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[FMUL1]](<2 x s16>)
; GFX9: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
; GFX9: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>)
; GFX9: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
; GFX9: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
; GFX9: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
; GFX9: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32)
; GFX9: [[COPY8:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32)
; GFX9: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; GFX9: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY8]](s32), [[COPY9]](s32)
; GFX9: [[COPY10:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32)
; GFX9: [[COPY11:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32)
; GFX9: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY10]](s32), [[COPY11]](s32)
; GFX9: [[COPY12:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32)
; GFX9: [[COPY13:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32)
; GFX9: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY12]](s32), [[COPY13]](s32)
; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
%1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
%2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0
%4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1
%6:_(<3 x s16>) = G_FMUL %2, %4
%7:_(<3 x s16>) = G_IMPLICIT_DEF
%8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7
$vgpr0_vgpr1_vgpr2 = COPY %8
...
---
name: test_fmul_v4s16
body: |
bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; SI-LABEL: name: test_fmul_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: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[FPEXT1]]
; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32)
; SI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16)
; SI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16)
; SI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[FPEXT3]]
; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32)
; SI: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16)
; SI: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC6]](s16)
; SI: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[FPEXT5]]
; SI: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32)
; SI: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16)
; SI: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC7]](s16)
; SI: [[FMUL3:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT6]], [[FPEXT7]]
; SI: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL3]](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_fmul_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: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC4]]
; VI: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[TRUNC5]]
; VI: [[FMUL2:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC2]], [[TRUNC6]]
; VI: [[FMUL3:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC3]], [[TRUNC7]]
; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FMUL]](s16)
; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FMUL1]](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 [[FMUL2]](s16)
; VI: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[FMUL3]](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_fmul_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: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
; GFX9: [[FMUL:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[UV]], [[UV2]]
; GFX9: [[FMUL1:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[UV1]], [[UV3]]
; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[FMUL]](<2 x s16>), [[FMUL1]](<2 x s16>)
; GFX9: $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_FMUL %0, %1
$vgpr0_vgpr1 = COPY %2
...