This is an archive of the discontinued LLVM Phabricator instance.

Differential D48486

AMDGPU: Respect align argument parameter
ClosedPublic

Authored by arsenm on Jun 22 2018, 7:49 AM.

Details

Reviewers
kzhuravl
scott.linder
Summary

This should avoid relying on the pointee type
to get the alignment, particularly since pointee
types are supposed to be removed at some point.

Also fixes not getting the alignment for unsized types.

Diff Detail

Event Timeline

arsenm created this revision.Jun 22 2018, 7:49 AM
Herald added subscribers: t-tye, tpr, dstuttard and 3 others. · View Herald TranscriptJun 22 2018, 7:49 AM
kzhuravl accepted this revision.Jun 22 2018, 9:00 AM

LGTM

This revision is now accepted and ready to land.Jun 22 2018, 9:00 AM
arsenm closed this revision.Jun 25 2018, 7:33 AM

r335478

Revision Contents

PathSize
lib/
Target/
AMDGPU/
MCTargetDesc/
1 line
28 lines
test/
CodeGen/
AMDGPU/
115 lines

Diff 152477

lib/Target/AMDGPU/MCTargetDesc/AMDGPUHSAMetadataStreamer.h

Show First 20 Lines • Show All 62 Lines • ▼ Show 20 Linesprivate:
void emitKernelAttrs(const Function &Func); void emitKernelAttrs(const Function &Func);
void emitKernelArgs(const Function &Func); void emitKernelArgs(const Function &Func);
void emitKernelArg(const Argument &Arg); void emitKernelArg(const Argument &Arg);
void emitKernelArg(const DataLayout &DL, Type *Ty, ValueKind ValueKind, void emitKernelArg(const DataLayout &DL, Type *Ty, ValueKind ValueKind,
unsigned PointeeAlign = 0,
StringRef Name = "", StringRef TypeName = "", StringRef Name = "", StringRef TypeName = "",
StringRef BaseTypeName = "", StringRef AccQual = "", StringRef BaseTypeName = "", StringRef AccQual = "",
StringRef TypeQual = ""); StringRef TypeQual = "");
public: public:
MetadataStreamer() = default; MetadataStreamer() = default;
~MetadataStreamer() = default; ~MetadataStreamer() = default;
Show All 18 Lines

lib/Target/AMDGPU/MCTargetDesc/AMDGPUHSAMetadataStreamer.cpp

Show First 20 Lines • Show All 318 Lines • ▼ Show 20 Lines if (Node && ArgNo < Node->getNumOperands())
AccQual = cast<MDString>(Node->getOperand(ArgNo))->getString(); AccQual = cast<MDString>(Node->getOperand(ArgNo))->getString();
} }
StringRef TypeQual; StringRef TypeQual;
Node = Func->getMetadata("kernel_arg_type_qual"); Node = Func->getMetadata("kernel_arg_type_qual");
if (Node && ArgNo < Node->getNumOperands()) if (Node && ArgNo < Node->getNumOperands())
TypeQual = cast<MDString>(Node->getOperand(ArgNo))->getString(); TypeQual = cast<MDString>(Node->getOperand(ArgNo))->getString();
emitKernelArg(Func->getParent()->getDataLayout(), Arg.getType(), Type *Ty = Arg.getType();
getValueKind(Arg.getType(), TypeQual, BaseTypeName), Name, const DataLayout &DL = Func->getParent()->getDataLayout();
TypeName, BaseTypeName, AccQual, TypeQual);
unsigned PointeeAlign = 0;
if (auto PtrTy = dyn_cast<PointerType>(Ty)) {
if (PtrTy->getAddressSpace() == AMDGPUASI.LOCAL_ADDRESS) {
PointeeAlign = Arg.getParamAlignment();
if (PointeeAlign == 0)
PointeeAlign = DL.getABITypeAlignment(PtrTy->getElementType());
}
}
emitKernelArg(DL, Ty, getValueKind(Arg.getType(), TypeQual, BaseTypeName),
PointeeAlign, Name, TypeName, BaseTypeName, AccQual, TypeQual);
} }
void MetadataStreamer::emitKernelArg(const DataLayout &DL, Type *Ty, void MetadataStreamer::emitKernelArg(const DataLayout &DL, Type *Ty,
ValueKind ValueKind, StringRef Name, ValueKind ValueKind,
unsigned PointeeAlign,
StringRef Name,
StringRef TypeName, StringRef BaseTypeName, StringRef TypeName, StringRef BaseTypeName,
StringRef AccQual, StringRef TypeQual) { StringRef AccQual, StringRef TypeQual) {
HSAMetadata.mKernels.back().mArgs.push_back(Kernel::Arg::Metadata()); HSAMetadata.mKernels.back().mArgs.push_back(Kernel::Arg::Metadata());
auto &Arg = HSAMetadata.mKernels.back().mArgs.back(); auto &Arg = HSAMetadata.mKernels.back().mArgs.back();
Arg.mName = Name; Arg.mName = Name;
Arg.mTypeName = TypeName; Arg.mTypeName = TypeName;
Arg.mSize = DL.getTypeAllocSize(Ty); Arg.mSize = DL.getTypeAllocSize(Ty);
Arg.mAlign = DL.getABITypeAlignment(Ty); Arg.mAlign = DL.getABITypeAlignment(Ty);
Arg.mValueKind = ValueKind; Arg.mValueKind = ValueKind;
Arg.mValueType = getValueType(Ty, BaseTypeName); Arg.mValueType = getValueType(Ty, BaseTypeName);
Arg.mPointeeAlign = PointeeAlign;
if (auto PtrTy = dyn_cast<PointerType>(Ty)) {
auto ElTy = PtrTy->getElementType();
if (PtrTy->getAddressSpace() == AMDGPUASI.LOCAL_ADDRESS && ElTy->isSized())
Arg.mPointeeAlign = DL.getABITypeAlignment(ElTy);
}
if (auto PtrTy = dyn_cast<PointerType>(Ty)) if (auto PtrTy = dyn_cast<PointerType>(Ty))
Arg.mAddrSpaceQual = getAddressSpaceQualifer(PtrTy->getAddressSpace()); Arg.mAddrSpaceQual = getAddressSpaceQualifer(PtrTy->getAddressSpace());
Arg.mAccQual = getAccessQualifier(AccQual); Arg.mAccQual = getAccessQualifier(AccQual);
// TODO: Emit Arg.mActualAccQual. // TODO: Emit Arg.mActualAccQual.
▲ Show 20 Lines • Show All 53 Lines • Show Last 20 Lines

test/CodeGen/AMDGPU/hsa-metadata-from-llvm-ir-full.ll

Show First 20 Lines • Show All 1,242 Lines • ▼ Show 20 Lines
; CHECK-NEXT: TypeName: 'char16 addrspace(5)*' ; CHECK-NEXT: TypeName: 'char16 addrspace(5)*'
; CHECK-NEXT: Size: 4 ; CHECK-NEXT: Size: 4
; CHECK-NEXT: Align: 4 ; CHECK-NEXT: Align: 4
; CHECK-NEXT: ValueKind: DynamicSharedPointer ; CHECK-NEXT: ValueKind: DynamicSharedPointer
; CHECK-NEXT: ValueType: I8 ; CHECK-NEXT: ValueType: I8
; CHECK-NEXT: PointeeAlign: 16 ; CHECK-NEXT: PointeeAlign: 16
; CHECK-NEXT: AddrSpaceQual: Local ; CHECK-NEXT: AddrSpaceQual: Local
; CHECK-NEXT: AccQual: Default ; CHECK-NEXT: AccQual: Default
; CHECK-NEXT: - Name: h
; CHECK-NEXT: Size: 4
; CHECK-NEXT: Align: 4
; CHECK-NEXT: ValueKind: DynamicSharedPointer
; CHECK-NEXT: ValueType: Struct
; CHECK-NEXT: PointeeAlign: 1
; CHECK-NEXT: AddrSpaceQual: Local
; CHECK-NEXT: - Size: 8 ; CHECK-NEXT: - Size: 8
; CHECK-NEXT: Align: 8 ; CHECK-NEXT: Align: 8
; CHECK-NEXT: ValueKind: HiddenGlobalOffsetX ; CHECK-NEXT: ValueKind: HiddenGlobalOffsetX
; CHECK-NEXT: ValueType: I64 ; CHECK-NEXT: ValueType: I64
; CHECK-NEXT: - Size: 8 ; CHECK-NEXT: - Size: 8
; CHECK-NEXT: Align: 8 ; CHECK-NEXT: Align: 8
; CHECK-NEXT: ValueKind: HiddenGlobalOffsetY ; CHECK-NEXT: ValueKind: HiddenGlobalOffsetY
; CHECK-NEXT: ValueType: I64 ; CHECK-NEXT: ValueType: I64
; CHECK-NEXT: - Size: 8 ; CHECK-NEXT: - Size: 8
; CHECK-NEXT: Align: 8 ; CHECK-NEXT: Align: 8
; CHECK-NEXT: ValueKind: HiddenGlobalOffsetZ ; CHECK-NEXT: ValueKind: HiddenGlobalOffsetZ
; CHECK-NEXT: ValueType: I64 ; CHECK-NEXT: ValueType: I64
; CHECK-NEXT: - Size: 8 ; CHECK-NEXT: - Size: 8
; CHECK-NEXT: Align: 8 ; CHECK-NEXT: Align: 8
; CHECK-NEXT: ValueKind: HiddenPrintfBuffer ; CHECK-NEXT: ValueKind: HiddenPrintfBuffer
; CHECK-NEXT: ValueType: I8 ; CHECK-NEXT: ValueType: I8
; CHECK-NEXT: AddrSpaceQual: Global ; CHECK-NEXT: AddrSpaceQual: Global
define amdgpu_kernel void @test_pointee_align(i64 addrspace(1)* %a, define amdgpu_kernel void @test_pointee_align(i64 addrspace(1)* %a,
i8 addrspace(3)* %b, i8 addrspace(3)* %b,
<2 x i8> addrspace(3)* %c, <2 x i8> addrspace(3)* %c,
<3 x i8> addrspace(3)* %d, <3 x i8> addrspace(3)* %d,
<4 x i8> addrspace(3)* %e, <4 x i8> addrspace(3)* %e,
<8 x i8> addrspace(3)* %f, <8 x i8> addrspace(3)* %f,
<16 x i8> addrspace(3)* %g) <16 x i8> addrspace(3)* %g,
{} addrspace(3)* %h)
!kernel_arg_addr_space !91 !kernel_arg_access_qual !92 !kernel_arg_type !93 !kernel_arg_addr_space !91 !kernel_arg_access_qual !92 !kernel_arg_type !93
!kernel_arg_base_type !93 !kernel_arg_type_qual !94 { !kernel_arg_base_type !93 !kernel_arg_type_qual !94 {
ret void ret void
} }
; CHECK: - Name: test_pointee_align_attribute
; CHECK-NEXT: SymbolName: 'test_pointee_align_attribute@kd'
; CHECK-NEXT: Language: OpenCL C
; CHECK-NEXT: LanguageVersion: [ 2, 0 ]
; CHECK-NEXT: Args:
; CHECK-NEXT: - Name: a
; CHECK-NEXT: TypeName: 'long addrspace(5)*'
; CHECK-NEXT: Size: 8
; CHECK-NEXT: Align: 8
; CHECK-NEXT: ValueKind: GlobalBuffer
; CHECK-NEXT: ValueType: I64
; CHECK-NEXT: AddrSpaceQual: Global
; CHECK-NEXT: AccQual: Default
; CHECK-NEXT: - Name: b
; CHECK-NEXT: TypeName: 'char addrspace(5)*'
; CHECK-NEXT: Size: 4
; CHECK-NEXT: Align: 4
; CHECK-NEXT: ValueKind: DynamicSharedPointer
; CHECK-NEXT: ValueType: I8
; CHECK-NEXT: PointeeAlign: 8
; CHECK-NEXT: AddrSpaceQual: Local
; CHECK-NEXT: AccQual: Default
; CHECK-NEXT: - Name: c
; CHECK-NEXT: TypeName: 'char2 addrspace(5)*'
; CHECK-NEXT: Size: 4
; CHECK-NEXT: Align: 4
; CHECK-NEXT: ValueKind: DynamicSharedPointer
; CHECK-NEXT: ValueType: I8
; CHECK-NEXT: PointeeAlign: 32
; CHECK-NEXT: AddrSpaceQual: Local
; CHECK-NEXT: AccQual: Default
; CHECK-NEXT: - Name: d
; CHECK-NEXT: TypeName: 'char3 addrspace(5)*'
; CHECK-NEXT: Size: 4
; CHECK-NEXT: Align: 4
; CHECK-NEXT: ValueKind: DynamicSharedPointer
; CHECK-NEXT: ValueType: I8
; CHECK-NEXT: PointeeAlign: 64
; CHECK-NEXT: AddrSpaceQual: Local
; CHECK-NEXT: AccQual: Default
; CHECK-NEXT: - Name: e
; CHECK-NEXT: TypeName: 'char4 addrspace(5)*'
; CHECK-NEXT: Size: 4
; CHECK-NEXT: Align: 4
; CHECK-NEXT: ValueKind: DynamicSharedPointer
; CHECK-NEXT: ValueType: I8
; CHECK-NEXT: PointeeAlign: 256
; CHECK-NEXT: AddrSpaceQual: Local
; CHECK-NEXT: AccQual: Default
; CHECK-NEXT: - Name: f
; CHECK-NEXT: TypeName: 'char8 addrspace(5)*'
; CHECK-NEXT: Size: 4
; CHECK-NEXT: Align: 4
; CHECK-NEXT: ValueKind: DynamicSharedPointer
; CHECK-NEXT: ValueType: I8
; CHECK-NEXT: PointeeAlign: 128
; CHECK-NEXT: AddrSpaceQual: Local
; CHECK-NEXT: AccQual: Default
; CHECK-NEXT: - Name: g
; CHECK-NEXT: TypeName: 'char16 addrspace(5)*'
; CHECK-NEXT: Size: 4
; CHECK-NEXT: Align: 4
; CHECK-NEXT: ValueKind: DynamicSharedPointer
; CHECK-NEXT: ValueType: I8
; CHECK-NEXT: PointeeAlign: 1024
; CHECK-NEXT: AddrSpaceQual: Local
; CHECK-NEXT: AccQual: Default
; CHECK-NEXT: - Name: h
; CHECK-NEXT: Size: 4
; CHECK-NEXT: Align: 4
; CHECK-NEXT: ValueKind: DynamicSharedPointer
; CHECK-NEXT: ValueType: Struct
; CHECK-NEXT: PointeeAlign: 16
; CHECK-NEXT: AddrSpaceQual: Local
; CHECK-NEXT: - Size: 8
; CHECK-NEXT: Align: 8
; CHECK-NEXT: ValueKind: HiddenGlobalOffsetX
; CHECK-NEXT: ValueType: I64
; CHECK-NEXT: - Size: 8
; CHECK-NEXT: Align: 8
; CHECK-NEXT: ValueKind: HiddenGlobalOffsetY
; CHECK-NEXT: ValueType: I64
; CHECK-NEXT: - Size: 8
; CHECK-NEXT: Align: 8
; CHECK-NEXT: ValueKind: HiddenGlobalOffsetZ
; CHECK-NEXT: ValueType: I64
; CHECK-NEXT: - Size: 8
; CHECK-NEXT: Align: 8
; CHECK-NEXT: ValueKind: HiddenPrintfBuffer
; CHECK-NEXT: ValueType: I8
; CHECK-NEXT: AddrSpaceQual: Global
define amdgpu_kernel void @test_pointee_align_attribute(i64 addrspace(1)* align 16 %a,
i8 addrspace(3)* align 8 %b,
<2 x i8> addrspace(3)* align 32 %c,
<3 x i8> addrspace(3)* align 64 %d,
<4 x i8> addrspace(3)* align 256 %e,
<8 x i8> addrspace(3)* align 128 %f,
<16 x i8> addrspace(3)* align 1024 %g,
{} addrspace(3)* align 16 %h)
!kernel_arg_addr_space !91 !kernel_arg_access_qual !92 !kernel_arg_type !93
!kernel_arg_base_type !93 !kernel_arg_type_qual !94 {
ret void
}
; CHECK: - Name: __test_block_invoke_kernel ; CHECK: - Name: __test_block_invoke_kernel
; CHECK-NEXT: SymbolName: '__test_block_invoke_kernel@kd' ; CHECK-NEXT: SymbolName: '__test_block_invoke_kernel@kd'
; CHECK-NEXT: Language: OpenCL C ; CHECK-NEXT: Language: OpenCL C
; CHECK-NEXT: LanguageVersion: [ 2, 0 ] ; CHECK-NEXT: LanguageVersion: [ 2, 0 ]
; CHECK-NEXT: Attrs: ; CHECK-NEXT: Attrs:
; CHECK-NEXT: RuntimeHandle: __test_block_invoke_kernel_runtime_handle ; CHECK-NEXT: RuntimeHandle: __test_block_invoke_kernel_runtime_handle
; CHECK-NEXT: Args: ; CHECK-NEXT: Args:
; CHECK-NEXT: - Name: arg ; CHECK-NEXT: - Name: arg
▲ Show 20 Lines • Show All 134 Lines • Show Last 20 Lines