RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 03:02:36 +01:00
Code Issues Projects Releases Wiki Activity

AMDGPU/GlobalISel: Stop using G_EXTRACT in argument lowering

Browse Source 
We really need to put this undef padding stuff into a helper
somewhere, but leave that for when this is moved to generic code.
This commit is contained in:
Matt Arsenault 2020-07-10 16:12:35 -04:00
parent d9a567a83e
commit 68cd67d10b
3 changed files with 238 additions and 269 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

49
lib/Target/AMDGPU/AMDGPUCallLowering.cpp
View File

@ -357,16 +357,6 @@ void AMDGPUCallLowering::splitToValueTypes(
}
}
// Get the appropriate type to make \p OrigTy \p Factor times bigger.
static LLT getMultipleType(LLT OrigTy, int Factor) {
if (OrigTy.isVector()) {
return LLT::vector(OrigTy.getNumElements() * Factor,
OrigTy.getElementType());
}
return LLT::scalar(OrigTy.getSizeInBits() * Factor);
}
// TODO: Move to generic code
static void unpackRegsToOrigType(MachineIRBuilder &B,
ArrayRef<Register> DstRegs,
@ -376,34 +366,51 @@ static void unpackRegsToOrigType(MachineIRBuilder &B,
LLT PartTy) {
assert(DstRegs.size() > 1 && "Nothing to unpack");
const unsigned SrcSize = SrcTy.getSizeInBits();
const unsigned PartSize = PartTy.getSizeInBits();
if (SrcTy.isVector() && !PartTy.isVector() &&
PartSize > SrcTy.getElementType().getSizeInBits()) {
// Vector was scalarized, and the elements extended.
auto UnmergeToEltTy = B.buildUnmerge(SrcTy.getElementType(),
SrcReg);
auto UnmergeToEltTy = B.buildUnmerge(SrcTy.getElementType(), SrcReg);
for (int i = 0, e = DstRegs.size(); i != e; ++i)
B.buildAnyExt(DstRegs[i], UnmergeToEltTy.getReg(i));
return;
}
if (SrcSize % PartSize == 0) {
LLT GCDTy = getGCDType(SrcTy, PartTy);
if (GCDTy == PartTy) {
// If this already evenly divisible, we can create a simple unmerge.
B.buildUnmerge(DstRegs, SrcReg);
return;
}
const int NumRoundedParts = (SrcSize + PartSize - 1) / PartSize;
MachineRegisterInfo &MRI = *B.getMRI();
LLT DstTy = MRI.getType(DstRegs[0]);
LLT LCMTy = getLCMType(SrcTy, PartTy);
LLT BigTy = getMultipleType(PartTy, NumRoundedParts);
auto ImpDef = B.buildUndef(BigTy);
const unsigned LCMSize = LCMTy.getSizeInBits();
const unsigned DstSize = DstTy.getSizeInBits();
const unsigned SrcSize = SrcTy.getSizeInBits();
auto Big = B.buildInsert(BigTy, ImpDef.getReg(0), SrcReg, 0).getReg(0);
Register UnmergeSrc = SrcReg;
if (LCMSize != SrcSize) {
// Widen to the common type.
Register Undef = B.buildUndef(SrcTy).getReg(0);
SmallVector<Register, 8> MergeParts(1, SrcReg);
for (unsigned Size = SrcSize; Size != LCMSize; Size += SrcSize)
MergeParts.push_back(Undef);
int64_t Offset = 0;
for (unsigned i = 0, e = DstRegs.size(); i != e; ++i, Offset += PartSize)
B.buildExtract(DstRegs[i], Big, Offset);
UnmergeSrc = B.buildMerge(LCMTy, MergeParts).getReg(0);
}
// Unmerge to the original registers and pad with dead defs.
SmallVector<Register, 8> UnmergeResults(DstRegs.begin(), DstRegs.end());
for (unsigned Size = DstSize * DstRegs.size(); Size != LCMSize;
Size += DstSize) {
UnmergeResults.push_back(MRI.createGenericVirtualRegister(DstTy));
}
B.buildUnmerge(UnmergeResults, UnmergeSrc);
}
/// Lower the return value for the already existing \p Ret. This assumes that

258
test/CodeGen/AMDGPU/GlobalISel/irtranslator-call.ll
View File

@ -1977,10 +1977,9 @@ define amdgpu_kernel void @test_call_external_void_func_v3i16() #0 {
; CHECK: [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
; CHECK: [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
; CHECK: [[LOAD:%[0-9]+]]:_(<3 x s16>) = G_LOAD [[DEF]](p1) :: (load 6 from `<3 x i16> addrspace(1)* undef`, align 8, addrspace 1)
; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[LOAD]](<3 x s16>), 0
; CHECK: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; CHECK: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; CHECK: [[DEF1:%[0-9]+]]:_(<3 x s16>) = G_IMPLICIT_DEF
; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[LOAD]](<3 x s16>), [[DEF1]](<3 x s16>)
; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $scc
; CHECK: [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v3i16
; CHECK: [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
@ -2001,8 +2000,8 @@ define amdgpu_kernel void @test_call_external_void_func_v3i16() #0 {
; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
; CHECK: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; CHECK: $vgpr0 = COPY [[UV]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[UV1]](<2 x s16>)
; CHECK: [[COPY20:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
; CHECK: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY20]](<4 x s32>)
; CHECK: $sgpr4_sgpr5 = COPY [[COPY10]](p4)
@ -2037,10 +2036,9 @@ define amdgpu_kernel void @test_call_external_void_func_v3f16() #0 {
; CHECK: [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
; CHECK: [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
; CHECK: [[LOAD:%[0-9]+]]:_(<3 x s16>) = G_LOAD [[DEF]](p1) :: (load 6 from `<3 x half> addrspace(1)* undef`, align 8, addrspace 1)
; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[LOAD]](<3 x s16>), 0
; CHECK: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; CHECK: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; CHECK: [[DEF1:%[0-9]+]]:_(<3 x s16>) = G_IMPLICIT_DEF
; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[LOAD]](<3 x s16>), [[DEF1]](<3 x s16>)
; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $scc
; CHECK: [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v3f16
; CHECK: [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
@ -2061,8 +2059,8 @@ define amdgpu_kernel void @test_call_external_void_func_v3f16() #0 {
; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
; CHECK: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; CHECK: $vgpr0 = COPY [[UV]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[UV1]](<2 x s16>)
; CHECK: [[COPY20:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
; CHECK: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY20]](<4 x s32>)
; CHECK: $sgpr4_sgpr5 = COPY [[COPY10]](p4)
@ -2213,11 +2211,9 @@ define amdgpu_kernel void @test_call_external_void_func_v5i16() #0 {
; CHECK: [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
; CHECK: [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
; CHECK: [[LOAD:%[0-9]+]]:_(<5 x s16>) = G_LOAD [[DEF]](p1) :: (load 10 from `<5 x i16> addrspace(1)* undef`, align 16, addrspace 1)
; CHECK: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[LOAD]](<5 x s16>), 0
; CHECK: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<6 x s16>), 0
; CHECK: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<6 x s16>), 32
; CHECK: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<6 x s16>), 64
; CHECK: [[DEF1:%[0-9]+]]:_(<5 x s16>) = G_IMPLICIT_DEF
; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[LOAD]](<5 x s16>), [[DEF1]](<5 x s16>)
; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<10 x s16>)
; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $scc
; CHECK: [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v5i16
; CHECK: [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
@ -2238,9 +2234,9 @@ define amdgpu_kernel void @test_call_external_void_func_v5i16() #0 {
; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
; CHECK: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; CHECK: $vgpr2 = COPY [[EXTRACT2]](<2 x s16>)
; CHECK: $vgpr0 = COPY [[UV]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[UV1]](<2 x s16>)
; CHECK: $vgpr2 = COPY [[UV2]](<2 x s16>)
; CHECK: [[COPY20:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
; CHECK: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY20]](<4 x s32>)
; CHECK: $sgpr4_sgpr5 = COPY [[COPY10]](p4)
@ -2275,12 +2271,9 @@ define amdgpu_kernel void @test_call_external_void_func_v7i16() #0 {
; CHECK: [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
; CHECK: [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
; CHECK: [[LOAD:%[0-9]+]]:_(<7 x s16>) = G_LOAD [[DEF]](p1) :: (load 14 from `<7 x i16> addrspace(1)* undef`, align 16, addrspace 1)
; CHECK: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; CHECK: [[INSERT:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF1]], [[LOAD]](<7 x s16>), 0
; CHECK: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<8 x s16>), 0
; CHECK: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<8 x s16>), 32
; CHECK: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<8 x s16>), 64
; CHECK: [[EXTRACT3:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<8 x s16>), 96
; CHECK: [[DEF1:%[0-9]+]]:_(<7 x s16>) = G_IMPLICIT_DEF
; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<14 x s16>) = G_CONCAT_VECTORS [[LOAD]](<7 x s16>), [[DEF1]](<7 x s16>)
; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<14 x s16>)
; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $scc
; CHECK: [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v7i16
; CHECK: [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
@ -2301,10 +2294,10 @@ define amdgpu_kernel void @test_call_external_void_func_v7i16() #0 {
; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
; CHECK: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; CHECK: $vgpr2 = COPY [[EXTRACT2]](<2 x s16>)
; CHECK: $vgpr3 = COPY [[EXTRACT3]](<2 x s16>)
; CHECK: $vgpr0 = COPY [[UV]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[UV1]](<2 x s16>)
; CHECK: $vgpr2 = COPY [[UV2]](<2 x s16>)
; CHECK: $vgpr3 = COPY [[UV3]](<2 x s16>)
; CHECK: [[COPY20:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
; CHECK: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY20]](<4 x s32>)
; CHECK: $sgpr4_sgpr5 = COPY [[COPY10]](p4)
@ -2339,40 +2332,9 @@ define amdgpu_kernel void @test_call_external_void_func_v63i16() #0 {
; CHECK: [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
; CHECK: [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
; CHECK: [[LOAD:%[0-9]+]]:_(<63 x s16>) = G_LOAD [[DEF]](p1) :: (load 126 from `<63 x i16> addrspace(1)* undef`, align 128, addrspace 1)
; CHECK: [[DEF1:%[0-9]+]]:_(<64 x s16>) = G_IMPLICIT_DEF
; CHECK: [[INSERT:%[0-9]+]]:_(<64 x s16>) = G_INSERT [[DEF1]], [[LOAD]](<63 x s16>), 0
; CHECK: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 0
; CHECK: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 32
; CHECK: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 64
; CHECK: [[EXTRACT3:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 96
; CHECK: [[EXTRACT4:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 128
; CHECK: [[EXTRACT5:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 160
; CHECK: [[EXTRACT6:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 192
; CHECK: [[EXTRACT7:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 224
; CHECK: [[EXTRACT8:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 256
; CHECK: [[EXTRACT9:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 288
; CHECK: [[EXTRACT10:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 320
; CHECK: [[EXTRACT11:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 352
; CHECK: [[EXTRACT12:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 384
; CHECK: [[EXTRACT13:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 416
; CHECK: [[EXTRACT14:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 448
; CHECK: [[EXTRACT15:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 480
; CHECK: [[EXTRACT16:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 512
; CHECK: [[EXTRACT17:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 544
; CHECK: [[EXTRACT18:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 576
; CHECK: [[EXTRACT19:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 608
; CHECK: [[EXTRACT20:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 640
; CHECK: [[EXTRACT21:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 672
; CHECK: [[EXTRACT22:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 704
; CHECK: [[EXTRACT23:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 736
; CHECK: [[EXTRACT24:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 768
; CHECK: [[EXTRACT25:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 800
; CHECK: [[EXTRACT26:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 832
; CHECK: [[EXTRACT27:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 864
; CHECK: [[EXTRACT28:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 896
; CHECK: [[EXTRACT29:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 928
; CHECK: [[EXTRACT30:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 960
; CHECK: [[EXTRACT31:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<64 x s16>), 992
; CHECK: [[DEF1:%[0-9]+]]:_(<63 x s16>) = G_IMPLICIT_DEF
; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<126 x s16>) = G_CONCAT_VECTORS [[LOAD]](<63 x s16>), [[DEF1]](<63 x s16>)
; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>), [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>), [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>), [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>), [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>), [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>), [[UV24:%[0-9]+]]:_(<2 x s16>), [[UV25:%[0-9]+]]:_(<2 x s16>), [[UV26:%[0-9]+]]:_(<2 x s16>), [[UV27:%[0-9]+]]:_(<2 x s16>), [[UV28:%[0-9]+]]:_(<2 x s16>), [[UV29:%[0-9]+]]:_(<2 x s16>), [[UV30:%[0-9]+]]:_(<2 x s16>), [[UV31:%[0-9]+]]:_(<2 x s16>), [[UV32:%[0-9]+]]:_(<2 x s16>), [[UV33:%[0-9]+]]:_(<2 x s16>), [[UV34:%[0-9]+]]:_(<2 x s16>), [[UV35:%[0-9]+]]:_(<2 x s16>), [[UV36:%[0-9]+]]:_(<2 x s16>), [[UV37:%[0-9]+]]:_(<2 x s16>), [[UV38:%[0-9]+]]:_(<2 x s16>), [[UV39:%[0-9]+]]:_(<2 x s16>), [[UV40:%[0-9]+]]:_(<2 x s16>), [[UV41:%[0-9]+]]:_(<2 x s16>), [[UV42:%[0-9]+]]:_(<2 x s16>), [[UV43:%[0-9]+]]:_(<2 x s16>), [[UV44:%[0-9]+]]:_(<2 x s16>), [[UV45:%[0-9]+]]:_(<2 x s16>), [[UV46:%[0-9]+]]:_(<2 x s16>), [[UV47:%[0-9]+]]:_(<2 x s16>), [[UV48:%[0-9]+]]:_(<2 x s16>), [[UV49:%[0-9]+]]:_(<2 x s16>), [[UV50:%[0-9]+]]:_(<2 x s16>), [[UV51:%[0-9]+]]:_(<2 x s16>), [[UV52:%[0-9]+]]:_(<2 x s16>), [[UV53:%[0-9]+]]:_(<2 x s16>), [[UV54:%[0-9]+]]:_(<2 x s16>), [[UV55:%[0-9]+]]:_(<2 x s16>), [[UV56:%[0-9]+]]:_(<2 x s16>), [[UV57:%[0-9]+]]:_(<2 x s16>), [[UV58:%[0-9]+]]:_(<2 x s16>), [[UV59:%[0-9]+]]:_(<2 x s16>), [[UV60:%[0-9]+]]:_(<2 x s16>), [[UV61:%[0-9]+]]:_(<2 x s16>), [[UV62:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<126 x s16>)
; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $scc
; CHECK: [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v63i16
; CHECK: [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
@ -2393,41 +2355,41 @@ define amdgpu_kernel void @test_call_external_void_func_v63i16() #0 {
; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
; CHECK: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; CHECK: $vgpr2 = COPY [[EXTRACT2]](<2 x s16>)
; CHECK: $vgpr3 = COPY [[EXTRACT3]](<2 x s16>)
; CHECK: $vgpr4 = COPY [[EXTRACT4]](<2 x s16>)
; CHECK: $vgpr5 = COPY [[EXTRACT5]](<2 x s16>)
; CHECK: $vgpr6 = COPY [[EXTRACT6]](<2 x s16>)
; CHECK: $vgpr7 = COPY [[EXTRACT7]](<2 x s16>)
; CHECK: $vgpr8 = COPY [[EXTRACT8]](<2 x s16>)
; CHECK: $vgpr9 = COPY [[EXTRACT9]](<2 x s16>)
; CHECK: $vgpr10 = COPY [[EXTRACT10]](<2 x s16>)
; CHECK: $vgpr11 = COPY [[EXTRACT11]](<2 x s16>)
; CHECK: $vgpr12 = COPY [[EXTRACT12]](<2 x s16>)
; CHECK: $vgpr13 = COPY [[EXTRACT13]](<2 x s16>)
; CHECK: $vgpr14 = COPY [[EXTRACT14]](<2 x s16>)
; CHECK: $vgpr15 = COPY [[EXTRACT15]](<2 x s16>)
; CHECK: $vgpr16 = COPY [[EXTRACT16]](<2 x s16>)
; CHECK: $vgpr17 = COPY [[EXTRACT17]](<2 x s16>)
; CHECK: $vgpr18 = COPY [[EXTRACT18]](<2 x s16>)
; CHECK: $vgpr19 = COPY [[EXTRACT19]](<2 x s16>)
; CHECK: $vgpr20 = COPY [[EXTRACT20]](<2 x s16>)
; CHECK: $vgpr21 = COPY [[EXTRACT21]](<2 x s16>)
; CHECK: $vgpr22 = COPY [[EXTRACT22]](<2 x s16>)
; CHECK: $vgpr23 = COPY [[EXTRACT23]](<2 x s16>)
; CHECK: $vgpr24 = COPY [[EXTRACT24]](<2 x s16>)
; CHECK: $vgpr25 = COPY [[EXTRACT25]](<2 x s16>)
; CHECK: $vgpr26 = COPY [[EXTRACT26]](<2 x s16>)
; CHECK: $vgpr27 = COPY [[EXTRACT27]](<2 x s16>)
; CHECK: $vgpr28 = COPY [[EXTRACT28]](<2 x s16>)
; CHECK: $vgpr29 = COPY [[EXTRACT29]](<2 x s16>)
; CHECK: $vgpr30 = COPY [[EXTRACT30]](<2 x s16>)
; CHECK: $vgpr0 = COPY [[UV]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[UV1]](<2 x s16>)
; CHECK: $vgpr2 = COPY [[UV2]](<2 x s16>)
; CHECK: $vgpr3 = COPY [[UV3]](<2 x s16>)
; CHECK: $vgpr4 = COPY [[UV4]](<2 x s16>)
; CHECK: $vgpr5 = COPY [[UV5]](<2 x s16>)
; CHECK: $vgpr6 = COPY [[UV6]](<2 x s16>)
; CHECK: $vgpr7 = COPY [[UV7]](<2 x s16>)
; CHECK: $vgpr8 = COPY [[UV8]](<2 x s16>)
; CHECK: $vgpr9 = COPY [[UV9]](<2 x s16>)
; CHECK: $vgpr10 = COPY [[UV10]](<2 x s16>)
; CHECK: $vgpr11 = COPY [[UV11]](<2 x s16>)
; CHECK: $vgpr12 = COPY [[UV12]](<2 x s16>)
; CHECK: $vgpr13 = COPY [[UV13]](<2 x s16>)
; CHECK: $vgpr14 = COPY [[UV14]](<2 x s16>)
; CHECK: $vgpr15 = COPY [[UV15]](<2 x s16>)
; CHECK: $vgpr16 = COPY [[UV16]](<2 x s16>)
; CHECK: $vgpr17 = COPY [[UV17]](<2 x s16>)
; CHECK: $vgpr18 = COPY [[UV18]](<2 x s16>)
; CHECK: $vgpr19 = COPY [[UV19]](<2 x s16>)
; CHECK: $vgpr20 = COPY [[UV20]](<2 x s16>)
; CHECK: $vgpr21 = COPY [[UV21]](<2 x s16>)
; CHECK: $vgpr22 = COPY [[UV22]](<2 x s16>)
; CHECK: $vgpr23 = COPY [[UV23]](<2 x s16>)
; CHECK: $vgpr24 = COPY [[UV24]](<2 x s16>)
; CHECK: $vgpr25 = COPY [[UV25]](<2 x s16>)
; CHECK: $vgpr26 = COPY [[UV26]](<2 x s16>)
; CHECK: $vgpr27 = COPY [[UV27]](<2 x s16>)
; CHECK: $vgpr28 = COPY [[UV28]](<2 x s16>)
; CHECK: $vgpr29 = COPY [[UV29]](<2 x s16>)
; CHECK: $vgpr30 = COPY [[UV30]](<2 x s16>)
; CHECK: [[COPY20:%[0-9]+]]:_(p5) = COPY $sp_reg
; CHECK: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; CHECK: [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY20]], [[C3]](s32)
; CHECK: G_STORE [[EXTRACT31]](<2 x s16>), [[PTR_ADD1]](p5) :: (store 4 into stack, align 16, addrspace 5)
; CHECK: G_STORE [[UV31]](<2 x s16>), [[PTR_ADD1]](p5) :: (store 4 into stack, align 16, addrspace 5)
; CHECK: [[COPY21:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
; CHECK: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY21]](<4 x s32>)
; CHECK: $sgpr4_sgpr5 = COPY [[COPY10]](p4)
@ -2462,41 +2424,9 @@ define amdgpu_kernel void @test_call_external_void_func_v65i16() #0 {
; CHECK: [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
; CHECK: [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
; CHECK: [[LOAD:%[0-9]+]]:_(<65 x s16>) = G_LOAD [[DEF]](p1) :: (load 130 from `<65 x i16> addrspace(1)* undef`, align 256, addrspace 1)
; CHECK: [[DEF1:%[0-9]+]]:_(<66 x s16>) = G_IMPLICIT_DEF
; CHECK: [[INSERT:%[0-9]+]]:_(<66 x s16>) = G_INSERT [[DEF1]], [[LOAD]](<65 x s16>), 0
; CHECK: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 0
; CHECK: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 32
; CHECK: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 64
; CHECK: [[EXTRACT3:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 96
; CHECK: [[EXTRACT4:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 128
; CHECK: [[EXTRACT5:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 160
; CHECK: [[EXTRACT6:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 192
; CHECK: [[EXTRACT7:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 224
; CHECK: [[EXTRACT8:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 256
; CHECK: [[EXTRACT9:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 288
; CHECK: [[EXTRACT10:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 320
; CHECK: [[EXTRACT11:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 352
; CHECK: [[EXTRACT12:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 384
; CHECK: [[EXTRACT13:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 416
; CHECK: [[EXTRACT14:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 448
; CHECK: [[EXTRACT15:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 480
; CHECK: [[EXTRACT16:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 512
; CHECK: [[EXTRACT17:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 544
; CHECK: [[EXTRACT18:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 576
; CHECK: [[EXTRACT19:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 608
; CHECK: [[EXTRACT20:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 640
; CHECK: [[EXTRACT21:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 672
; CHECK: [[EXTRACT22:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 704
; CHECK: [[EXTRACT23:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 736
; CHECK: [[EXTRACT24:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 768
; CHECK: [[EXTRACT25:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 800
; CHECK: [[EXTRACT26:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 832
; CHECK: [[EXTRACT27:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 864
; CHECK: [[EXTRACT28:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 896
; CHECK: [[EXTRACT29:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 928
; CHECK: [[EXTRACT30:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 960
; CHECK: [[EXTRACT31:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 992
; CHECK: [[EXTRACT32:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<66 x s16>), 1024
; CHECK: [[DEF1:%[0-9]+]]:_(<65 x s16>) = G_IMPLICIT_DEF
; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<130 x s16>) = G_CONCAT_VECTORS [[LOAD]](<65 x s16>), [[DEF1]](<65 x s16>)
; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>), [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>), [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>), [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>), [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>), [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>), [[UV24:%[0-9]+]]:_(<2 x s16>), [[UV25:%[0-9]+]]:_(<2 x s16>), [[UV26:%[0-9]+]]:_(<2 x s16>), [[UV27:%[0-9]+]]:_(<2 x s16>), [[UV28:%[0-9]+]]:_(<2 x s16>), [[UV29:%[0-9]+]]:_(<2 x s16>), [[UV30:%[0-9]+]]:_(<2 x s16>), [[UV31:%[0-9]+]]:_(<2 x s16>), [[UV32:%[0-9]+]]:_(<2 x s16>), [[UV33:%[0-9]+]]:_(<2 x s16>), [[UV34:%[0-9]+]]:_(<2 x s16>), [[UV35:%[0-9]+]]:_(<2 x s16>), [[UV36:%[0-9]+]]:_(<2 x s16>), [[UV37:%[0-9]+]]:_(<2 x s16>), [[UV38:%[0-9]+]]:_(<2 x s16>), [[UV39:%[0-9]+]]:_(<2 x s16>), [[UV40:%[0-9]+]]:_(<2 x s16>), [[UV41:%[0-9]+]]:_(<2 x s16>), [[UV42:%[0-9]+]]:_(<2 x s16>), [[UV43:%[0-9]+]]:_(<2 x s16>), [[UV44:%[0-9]+]]:_(<2 x s16>), [[UV45:%[0-9]+]]:_(<2 x s16>), [[UV46:%[0-9]+]]:_(<2 x s16>), [[UV47:%[0-9]+]]:_(<2 x s16>), [[UV48:%[0-9]+]]:_(<2 x s16>), [[UV49:%[0-9]+]]:_(<2 x s16>), [[UV50:%[0-9]+]]:_(<2 x s16>), [[UV51:%[0-9]+]]:_(<2 x s16>), [[UV52:%[0-9]+]]:_(<2 x s16>), [[UV53:%[0-9]+]]:_(<2 x s16>), [[UV54:%[0-9]+]]:_(<2 x s16>), [[UV55:%[0-9]+]]:_(<2 x s16>), [[UV56:%[0-9]+]]:_(<2 x s16>), [[UV57:%[0-9]+]]:_(<2 x s16>), [[UV58:%[0-9]+]]:_(<2 x s16>), [[UV59:%[0-9]+]]:_(<2 x s16>), [[UV60:%[0-9]+]]:_(<2 x s16>), [[UV61:%[0-9]+]]:_(<2 x s16>), [[UV62:%[0-9]+]]:_(<2 x s16>), [[UV63:%[0-9]+]]:_(<2 x s16>), [[UV64:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<130 x s16>)
; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $scc
; CHECK: [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v65i16
; CHECK: [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
@ -2517,44 +2447,44 @@ define amdgpu_kernel void @test_call_external_void_func_v65i16() #0 {
; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
; CHECK: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; CHECK: $vgpr2 = COPY [[EXTRACT2]](<2 x s16>)
; CHECK: $vgpr3 = COPY [[EXTRACT3]](<2 x s16>)
; CHECK: $vgpr4 = COPY [[EXTRACT4]](<2 x s16>)
; CHECK: $vgpr5 = COPY [[EXTRACT5]](<2 x s16>)
; CHECK: $vgpr6 = COPY [[EXTRACT6]](<2 x s16>)
; CHECK: $vgpr7 = COPY [[EXTRACT7]](<2 x s16>)
; CHECK: $vgpr8 = COPY [[EXTRACT8]](<2 x s16>)
; CHECK: $vgpr9 = COPY [[EXTRACT9]](<2 x s16>)
; CHECK: $vgpr10 = COPY [[EXTRACT10]](<2 x s16>)
; CHECK: $vgpr11 = COPY [[EXTRACT11]](<2 x s16>)
; CHECK: $vgpr12 = COPY [[EXTRACT12]](<2 x s16>)
; CHECK: $vgpr13 = COPY [[EXTRACT13]](<2 x s16>)
; CHECK: $vgpr14 = COPY [[EXTRACT14]](<2 x s16>)
; CHECK: $vgpr15 = COPY [[EXTRACT15]](<2 x s16>)
; CHECK: $vgpr16 = COPY [[EXTRACT16]](<2 x s16>)
; CHECK: $vgpr17 = COPY [[EXTRACT17]](<2 x s16>)
; CHECK: $vgpr18 = COPY [[EXTRACT18]](<2 x s16>)
; CHECK: $vgpr19 = COPY [[EXTRACT19]](<2 x s16>)
; CHECK: $vgpr20 = COPY [[EXTRACT20]](<2 x s16>)
; CHECK: $vgpr21 = COPY [[EXTRACT21]](<2 x s16>)
; CHECK: $vgpr22 = COPY [[EXTRACT22]](<2 x s16>)
; CHECK: $vgpr23 = COPY [[EXTRACT23]](<2 x s16>)
; CHECK: $vgpr24 = COPY [[EXTRACT24]](<2 x s16>)
; CHECK: $vgpr25 = COPY [[EXTRACT25]](<2 x s16>)
; CHECK: $vgpr26 = COPY [[EXTRACT26]](<2 x s16>)
; CHECK: $vgpr27 = COPY [[EXTRACT27]](<2 x s16>)
; CHECK: $vgpr28 = COPY [[EXTRACT28]](<2 x s16>)
; CHECK: $vgpr29 = COPY [[EXTRACT29]](<2 x s16>)
; CHECK: $vgpr30 = COPY [[EXTRACT30]](<2 x s16>)
; CHECK: $vgpr0 = COPY [[UV]](<2 x s16>)
; CHECK: $vgpr1 = COPY [[UV1]](<2 x s16>)
; CHECK: $vgpr2 = COPY [[UV2]](<2 x s16>)
; CHECK: $vgpr3 = COPY [[UV3]](<2 x s16>)
; CHECK: $vgpr4 = COPY [[UV4]](<2 x s16>)
; CHECK: $vgpr5 = COPY [[UV5]](<2 x s16>)
; CHECK: $vgpr6 = COPY [[UV6]](<2 x s16>)
; CHECK: $vgpr7 = COPY [[UV7]](<2 x s16>)
; CHECK: $vgpr8 = COPY [[UV8]](<2 x s16>)
; CHECK: $vgpr9 = COPY [[UV9]](<2 x s16>)
; CHECK: $vgpr10 = COPY [[UV10]](<2 x s16>)
; CHECK: $vgpr11 = COPY [[UV11]](<2 x s16>)
; CHECK: $vgpr12 = COPY [[UV12]](<2 x s16>)
; CHECK: $vgpr13 = COPY [[UV13]](<2 x s16>)
; CHECK: $vgpr14 = COPY [[UV14]](<2 x s16>)
; CHECK: $vgpr15 = COPY [[UV15]](<2 x s16>)
; CHECK: $vgpr16 = COPY [[UV16]](<2 x s16>)
; CHECK: $vgpr17 = COPY [[UV17]](<2 x s16>)
; CHECK: $vgpr18 = COPY [[UV18]](<2 x s16>)
; CHECK: $vgpr19 = COPY [[UV19]](<2 x s16>)
; CHECK: $vgpr20 = COPY [[UV20]](<2 x s16>)
; CHECK: $vgpr21 = COPY [[UV21]](<2 x s16>)
; CHECK: $vgpr22 = COPY [[UV22]](<2 x s16>)
; CHECK: $vgpr23 = COPY [[UV23]](<2 x s16>)
; CHECK: $vgpr24 = COPY [[UV24]](<2 x s16>)
; CHECK: $vgpr25 = COPY [[UV25]](<2 x s16>)
; CHECK: $vgpr26 = COPY [[UV26]](<2 x s16>)
; CHECK: $vgpr27 = COPY [[UV27]](<2 x s16>)
; CHECK: $vgpr28 = COPY [[UV28]](<2 x s16>)
; CHECK: $vgpr29 = COPY [[UV29]](<2 x s16>)
; CHECK: $vgpr30 = COPY [[UV30]](<2 x s16>)
; CHECK: [[COPY20:%[0-9]+]]:_(p5) = COPY $sp_reg
; CHECK: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; CHECK: [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY20]], [[C3]](s32)
; CHECK: G_STORE [[EXTRACT31]](<2 x s16>), [[PTR_ADD1]](p5) :: (store 4 into stack, align 16, addrspace 5)
; CHECK: G_STORE [[UV31]](<2 x s16>), [[PTR_ADD1]](p5) :: (store 4 into stack, align 16, addrspace 5)
; CHECK: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; CHECK: [[PTR_ADD2:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY20]], [[C4]](s32)
; CHECK: G_STORE [[EXTRACT32]](<2 x s16>), [[PTR_ADD2]](p5) :: (store 4 into stack + 4, addrspace 5)
; CHECK: G_STORE [[UV32]](<2 x s16>), [[PTR_ADD2]](p5) :: (store 4 into stack + 4, addrspace 5)
; CHECK: [[COPY21:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
; CHECK: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY21]](<4 x s32>)
; CHECK: $sgpr4_sgpr5 = COPY [[COPY10]](p4)

200
test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.load.2d.d16.ll
View File

@ -132,11 +132,13 @@ define amdgpu_ps <3 x half> @image_load_v3f16(<8 x i32> inreg %rsrc, i32 %s, i32
; UNPACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>)
; UNPACKED: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0
; UNPACKED: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
; UNPACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; UNPACKED: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; UNPACKED: $vgpr0 = COPY [[EXTRACT1]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[EXTRACT2]](<2 x s16>)
; UNPACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
; UNPACKED: [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; UNPACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV3]](<3 x s16>)
; UNPACKED: [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; UNPACKED: $vgpr0 = COPY [[UV7]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[UV8]](<2 x s16>)
; UNPACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
; PACKED-LABEL: name: image_load_v3f16
; PACKED: bb.1 (%ir-block.0):
@ -159,11 +161,13 @@ define amdgpu_ps <3 x half> @image_load_v3f16(<8 x i32> inreg %rsrc, i32 %s, i32
; PACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[DEF]](<2 x s16>)
; PACKED: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; PACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV2]](<3 x s16>), 0
; PACKED: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; PACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; PACKED: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; PACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF2]](<4 x s16>), [[DEF2]](<4 x s16>)
; PACKED: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; PACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV2]](<3 x s16>), [[UV4]](<3 x s16>)
; PACKED: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; PACKED: $vgpr0 = COPY [[UV8]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[UV9]](<2 x s16>)
; PACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
%tex = call <3 x half> @llvm.amdgcn.image.load.2d.v3f16.i32(i32 7, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
ret <3 x half> %tex
@ -380,11 +384,13 @@ define amdgpu_ps <3 x half> @image_load_tfe_v3f16(<8 x i32> inreg %rsrc, i32 %s,
; UNPACKED: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0
; UNPACKED: G_STORE [[UV3]](s32), [[DEF]](p1) :: (store 4 into `i32 addrspace(1)* undef`, addrspace 1)
; UNPACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0
; UNPACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; UNPACKED: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; UNPACKED: $vgpr0 = COPY [[EXTRACT1]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[EXTRACT2]](<2 x s16>)
; UNPACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF2]](<4 x s16>), [[DEF2]](<4 x s16>)
; UNPACKED: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; UNPACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV4]](<3 x s16>)
; UNPACKED: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; UNPACKED: $vgpr0 = COPY [[UV8]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[UV9]](<2 x s16>)
; UNPACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
; PACKED-LABEL: name: image_load_tfe_v3f16
; PACKED: bb.1 (%ir-block.0):
@ -411,11 +417,13 @@ define amdgpu_ps <3 x half> @image_load_tfe_v3f16(<8 x i32> inreg %rsrc, i32 %s,
; PACKED: [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; PACKED: G_STORE [[UV2]](s32), [[DEF]](p1) :: (store 4 into `i32 addrspace(1)* undef`, addrspace 1)
; PACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[UV3]](<3 x s16>), 0
; PACKED: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; PACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; PACKED: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; PACKED: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF3]](<4 x s16>), [[DEF3]](<4 x s16>)
; PACKED: [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; PACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV3]](<3 x s16>), [[UV5]](<3 x s16>)
; PACKED: [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; PACKED: $vgpr0 = COPY [[UV9]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[UV10]](<2 x s16>)
; PACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
%res = call { <3 x half>, i32 } @llvm.amdgcn.image.load.2d.sl_v3f16i32s.i32(i32 7, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
%tex = extractvalue { <3 x half>, i32 } %res, 0
@ -651,11 +659,13 @@ define amdgpu_ps <3 x half> @image_load_v3f16_dmask_1100(<8 x i32> inreg %rsrc,
; UNPACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>)
; UNPACKED: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0
; UNPACKED: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
; UNPACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; UNPACKED: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; UNPACKED: $vgpr0 = COPY [[EXTRACT1]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[EXTRACT2]](<2 x s16>)
; UNPACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
; UNPACKED: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; UNPACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV2]](<3 x s16>)
; UNPACKED: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; UNPACKED: $vgpr0 = COPY [[UV6]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[UV7]](<2 x s16>)
; UNPACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
; PACKED-LABEL: name: image_load_v3f16_dmask_1100
; PACKED: bb.1 (%ir-block.0):
@ -677,11 +687,13 @@ define amdgpu_ps <3 x half> @image_load_v3f16_dmask_1100(<8 x i32> inreg %rsrc,
; PACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>)
; PACKED: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; PACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
; PACKED: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; PACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; PACKED: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; PACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF2]](<4 x s16>), [[DEF2]](<4 x s16>)
; PACKED: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; PACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<3 x s16>), [[UV2]](<3 x s16>)
; PACKED: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; PACKED: $vgpr0 = COPY [[UV6]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[UV7]](<2 x s16>)
; PACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
%tex = call <3 x half> @llvm.amdgcn.image.load.2d.v3f16.i32(i32 3, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
ret <3 x half> %tex
@ -717,11 +729,13 @@ define amdgpu_ps <3 x half> @image_load_v3f16_dmask_1000(<8 x i32> inreg %rsrc,
; UNPACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>)
; UNPACKED: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0
; UNPACKED: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
; UNPACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; UNPACKED: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; UNPACKED: $vgpr0 = COPY [[EXTRACT1]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[EXTRACT2]](<2 x s16>)
; UNPACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
; UNPACKED: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; UNPACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV]](<3 x s16>)
; UNPACKED: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; UNPACKED: $vgpr0 = COPY [[UV4]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[UV5]](<2 x s16>)
; UNPACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
; PACKED-LABEL: name: image_load_v3f16_dmask_1000
; PACKED: bb.1 (%ir-block.0):
@ -743,11 +757,13 @@ define amdgpu_ps <3 x half> @image_load_v3f16_dmask_1000(<8 x i32> inreg %rsrc,
; PACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>)
; PACKED: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; PACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
; PACKED: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; PACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; PACKED: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; PACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF2]](<4 x s16>), [[DEF2]](<4 x s16>)
; PACKED: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; PACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<3 x s16>), [[UV2]](<3 x s16>)
; PACKED: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; PACKED: $vgpr0 = COPY [[UV6]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[UV7]](<2 x s16>)
; PACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
%tex = call <3 x half> @llvm.amdgcn.image.load.2d.v3f16.i32(i32 1, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
ret <3 x half> %tex
@ -768,14 +784,16 @@ define amdgpu_ps <3 x half> @image_load_v3f16_dmask_0000(<8 x i32> inreg %rsrc,
; UNPACKED: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; UNPACKED: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; UNPACKED: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %16(<4 x s16>), %16(<4 x s16>)
; UNPACKED: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
; UNPACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
; UNPACKED: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; UNPACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; UNPACKED: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; UNPACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
; UNPACKED: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
; UNPACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
; UNPACKED: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; UNPACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<3 x s16>), [[UV4]](<3 x s16>)
; UNPACKED: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; UNPACKED: $vgpr0 = COPY [[UV8]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[UV9]](<2 x s16>)
; UNPACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
; PACKED-LABEL: name: image_load_v3f16_dmask_0000
; PACKED: bb.1 (%ir-block.0):
@ -791,14 +809,16 @@ define amdgpu_ps <3 x half> @image_load_v3f16_dmask_0000(<8 x i32> inreg %rsrc,
; PACKED: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; PACKED: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; PACKED: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %16(<4 x s16>), %16(<4 x s16>)
; PACKED: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
; PACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
; PACKED: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; PACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; PACKED: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; PACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
; PACKED: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
; PACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
; PACKED: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; PACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<3 x s16>), [[UV4]](<3 x s16>)
; PACKED: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; PACKED: $vgpr0 = COPY [[UV8]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[UV9]](<2 x s16>)
; PACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
%tex = call <3 x half> @llvm.amdgcn.image.load.2d.v3f16.i32(i32 0, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
ret <3 x half> %tex
@ -1223,11 +1243,13 @@ define amdgpu_ps <3 x half> @image_load_tfe_v3f16_dmask_1100(<8 x i32> inreg %rs
; UNPACKED: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0
; UNPACKED: G_STORE [[UV2]](s32), [[DEF]](p1) :: (store 4 into `i32 addrspace(1)* undef`, addrspace 1)
; UNPACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0
; UNPACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; UNPACKED: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; UNPACKED: $vgpr0 = COPY [[EXTRACT1]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[EXTRACT2]](<2 x s16>)
; UNPACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF2]](<4 x s16>), [[DEF2]](<4 x s16>)
; UNPACKED: [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; UNPACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV3]](<3 x s16>)
; UNPACKED: [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; UNPACKED: $vgpr0 = COPY [[UV7]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[UV8]](<2 x s16>)
; UNPACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
; PACKED-LABEL: name: image_load_tfe_v3f16_dmask_1100
; PACKED: bb.1 (%ir-block.0):
@ -1253,11 +1275,13 @@ define amdgpu_ps <3 x half> @image_load_tfe_v3f16_dmask_1100(<8 x i32> inreg %rs
; PACKED: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; PACKED: G_STORE [[UV1]](s32), [[DEF]](p1) :: (store 4 into `i32 addrspace(1)* undef`, addrspace 1)
; PACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[UV2]](<3 x s16>), 0
; PACKED: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; PACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; PACKED: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; PACKED: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF3]](<4 x s16>), [[DEF3]](<4 x s16>)
; PACKED: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; PACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV2]](<3 x s16>), [[UV4]](<3 x s16>)
; PACKED: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; PACKED: $vgpr0 = COPY [[UV8]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[UV9]](<2 x s16>)
; PACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
%res = call { <3 x half>, i32 } @llvm.amdgcn.image.load.2d.sl_v3f16i32s.i32(i32 3, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
%tex = extractvalue { <3 x half>, i32 } %res, 0
@ -1299,11 +1323,13 @@ define amdgpu_ps <3 x half> @image_load_tfe_v3f16_dmask_1000(<8 x i32> inreg %rs
; UNPACKED: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0
; UNPACKED: G_STORE [[UV1]](s32), [[DEF]](p1) :: (store 4 into `i32 addrspace(1)* undef`, addrspace 1)
; UNPACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0
; UNPACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; UNPACKED: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; UNPACKED: $vgpr0 = COPY [[EXTRACT1]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[EXTRACT2]](<2 x s16>)
; UNPACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF2]](<4 x s16>), [[DEF2]](<4 x s16>)
; UNPACKED: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; UNPACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV2]](<3 x s16>)
; UNPACKED: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; UNPACKED: $vgpr0 = COPY [[UV6]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[UV7]](<2 x s16>)
; UNPACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
; PACKED-LABEL: name: image_load_tfe_v3f16_dmask_1000
; PACKED: bb.1 (%ir-block.0):
@ -1329,11 +1355,13 @@ define amdgpu_ps <3 x half> @image_load_tfe_v3f16_dmask_1000(<8 x i32> inreg %rs
; PACKED: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; PACKED: G_STORE [[UV1]](s32), [[DEF]](p1) :: (store 4 into `i32 addrspace(1)* undef`, addrspace 1)
; PACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[UV2]](<3 x s16>), 0
; PACKED: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; PACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; PACKED: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; PACKED: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF3]](<4 x s16>), [[DEF3]](<4 x s16>)
; PACKED: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; PACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV2]](<3 x s16>), [[UV4]](<3 x s16>)
; PACKED: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; PACKED: $vgpr0 = COPY [[UV8]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[UV9]](<2 x s16>)
; PACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
%res = call { <3 x half>, i32 } @llvm.amdgcn.image.load.2d.sl_v3f16i32s.i32(i32 1, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
%tex = extractvalue { <3 x half>, i32 } %res, 0
@ -1375,11 +1403,13 @@ define amdgpu_ps <3 x half> @image_load_tfe_v3f16_dmask_0000(<8 x i32> inreg %rs
; UNPACKED: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0
; UNPACKED: G_STORE [[UV1]](s32), [[DEF]](p1) :: (store 4 into `i32 addrspace(1)* undef`, addrspace 1)
; UNPACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0
; UNPACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; UNPACKED: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; UNPACKED: $vgpr0 = COPY [[EXTRACT1]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[EXTRACT2]](<2 x s16>)
; UNPACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; UNPACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF2]](<4 x s16>), [[DEF2]](<4 x s16>)
; UNPACKED: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; UNPACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV2]](<3 x s16>)
; UNPACKED: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; UNPACKED: $vgpr0 = COPY [[UV6]](<2 x s16>)
; UNPACKED: $vgpr1 = COPY [[UV7]](<2 x s16>)
; UNPACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
; PACKED-LABEL: name: image_load_tfe_v3f16_dmask_0000
; PACKED: bb.1 (%ir-block.0):
@ -1405,11 +1435,13 @@ define amdgpu_ps <3 x half> @image_load_tfe_v3f16_dmask_0000(<8 x i32> inreg %rs
; PACKED: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; PACKED: G_STORE [[UV1]](s32), [[DEF]](p1) :: (store 4 into `i32 addrspace(1)* undef`, addrspace 1)
; PACKED: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[UV2]](<3 x s16>), 0
; PACKED: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
; PACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32
; PACKED: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>)
; PACKED: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
; PACKED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF3]](<4 x s16>), [[DEF3]](<4 x s16>)
; PACKED: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
; PACKED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV2]](<3 x s16>), [[UV4]](<3 x s16>)
; PACKED: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<6 x s16>)
; PACKED: $vgpr0 = COPY [[UV8]](<2 x s16>)
; PACKED: $vgpr1 = COPY [[UV9]](<2 x s16>)
; PACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
%res = call { <3 x half>, i32 } @llvm.amdgcn.image.load.2d.sl_v3f16i32s.i32(i32 0, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
%tex = extractvalue { <3 x half>, i32 } %res, 0