[llvm-branch-commits] [flang] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
https://github.com/skatrak edited https://github.com/llvm/llvm-project/pull/77761 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [flang] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
https://github.com/skatrak commented: Thanks Krzysztof! This looks ok to me, but I suggest waiting for approval from a second reviewer. https://github.com/llvm/llvm-project/pull/77761 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [flang] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
@@ -0,0 +1,19 @@ +! RUN: bbc -fopenmp -o - %s | FileCheck %s + +! Check that this test can be lowered successfully. +! See https://github.com/llvm/llvm-project/issues/74348 + +! CHECK-LABEL: func.func @_QPsb +! CHECK: omp.parallel skatrak wrote: I think it would be a bit better to also check the general structure inside of omp.parallel. That is, checking the blocks and terminators that are created. https://github.com/llvm/llvm-project/pull/77761 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang] [llvm] [SPARC] Prefer RDPC over CALL to implement GETPCX for 64-bit target (PR #77196)
koachan wrote: Uh oh, seems like I couldn't merge it with the tool either... Guess I'll open a new PR to merge this, would that be okay? @brad0 @s-barannikov https://github.com/llvm/llvm-project/pull/77196 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [mlir] 91eb719 - Revert "[mlir][ExecutionEngine] Add support for global constructors and destructors"
Author: Cullen Rhodes
Date: 2024-01-15T14:10:08Z
New Revision: 91eb719cae57c7787af56401d2e63545cc53bc75
URL:
https://github.com/llvm/llvm-project/commit/91eb719cae57c7787af56401d2e63545cc53bc75
DIFF:
https://github.com/llvm/llvm-project/commit/91eb719cae57c7787af56401d2e63545cc53bc75.diff
LOG: Revert "[mlir][ExecutionEngine] Add support for global constructors and
destructors"
Added:
Modified:
mlir/lib/ExecutionEngine/ExecutionEngine.cpp
Removed:
mlir/test/mlir-cpu-runner/global-constructors.mlir
diff --git a/mlir/lib/ExecutionEngine/ExecutionEngine.cpp
b/mlir/lib/ExecutionEngine/ExecutionEngine.cpp
index d4ad2da045e2c4..267ec236d3fd59 100644
--- a/mlir/lib/ExecutionEngine/ExecutionEngine.cpp
+++ b/mlir/lib/ExecutionEngine/ExecutionEngine.cpp
@@ -219,9 +219,6 @@ ExecutionEngine::ExecutionEngine(bool enableObjectDump,
}
ExecutionEngine::~ExecutionEngine() {
- // Execute the global destructors from the module being processed.
- if (jit)
-llvm::consumeError(jit->deinitialize(jit->getMainJITDylib()));
// Run all dynamic library destroy callbacks to prepare for the shutdown.
for (LibraryDestroyFn destroy : destroyFns)
destroy();
@@ -399,9 +396,6 @@ ExecutionEngine::create(Operation *m, const
ExecutionEngineOptions &options,
};
engine->registerSymbols(runtimeSymbolMap);
- // Execute the global constructors from the module being processed.
- cantFail(engine->jit->initialize(engine->jit->getMainJITDylib()));
-
return std::move(engine);
}
diff --git a/mlir/test/mlir-cpu-runner/global-constructors.mlir
b/mlir/test/mlir-cpu-runner/global-constructors.mlir
deleted file mode 100644
index 3ebcaeccc83e32..00
--- a/mlir/test/mlir-cpu-runner/global-constructors.mlir
+++ /dev/null
@@ -1,32 +0,0 @@
-// RUN: mlir-cpu-runner %s -e entry -entry-point-result=void \
-// RUN: -shared-libs=%mlir_c_runner_utils | \
-// RUN: FileCheck %s
-
-// Test that the `ctor` executes before `entry` and that `dtor` executes last.
-module {
- llvm.func @printNewline()
- llvm.func @printI64(i64)
- llvm.mlir.global_ctors {ctors = [@ctor], priorities = [0 : i32]}
- llvm.mlir.global_dtors {dtors = [@dtor], priorities = [0 : i32]}
- llvm.func @ctor() {
-%0 = llvm.mlir.constant(1 : i64) : i64
-llvm.call @printI64(%0) : (i64) -> ()
-llvm.call @printNewline() : () -> ()
-// CHECK: 1
-llvm.return
- }
- llvm.func @entry() {
-%0 = llvm.mlir.constant(2 : i64) : i64
-llvm.call @printI64(%0) : (i64) -> ()
-llvm.call @printNewline() : () -> ()
-// CHECK: 2
-llvm.return
- }
- llvm.func @dtor() {
-%0 = llvm.mlir.constant(3 : i64) : i64
-llvm.call @printI64(%0) : (i64) -> ()
-llvm.call @printNewline() : () -> ()
-// CHECK: 3
-llvm.return
- }
-}
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang] [llvm] [SPARC] Prefer RDPC over CALL to implement GETPCX for 64-bit target (PR #77196)
s-barannikov wrote: Yes, sure https://github.com/llvm/llvm-project/pull/77196 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [flang] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
kiranchandramohan wrote: @tblah mentions that this patch fixes the unstructured code issue in worksharing loop with collapse. https://github.com/llvm/llvm-project/pull/77329. https://github.com/llvm/llvm-project/pull/77761 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [compiler-rt] [llvm] [clang-tools-extra] [clang] [flang] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
https://github.com/kparzysz updated
https://github.com/llvm/llvm-project/pull/77761
>From 1b5524ae8874e389d373a55417919afa56beb2b5 Mon Sep 17 00:00:00 2001
From: Krzysztof Parzyszek
Date: Mon, 8 Jan 2024 15:53:07 -0600
Subject: [PATCH] [Flang][OpenMP] Restructure recursive lowering in
`createBodyOfOp`
This brings `createBodyOfOp` to its final intended form. First, input
privatization is performed, then the recursive lowering takes place,
and finally the output privatization (lastprivate) is done.
This enables fixing a known issue with infinite loops inside of an
OpenMP region, and the fix is included in this patch.
Fixes https://github.com/llvm/llvm-project/issues/74348.
Recursive lowering [5/5]
---
flang/include/flang/Lower/OpenMP.h| 4 +-
flang/lib/Lower/OpenMP.cpp| 133 --
flang/test/Lower/OpenMP/FIR/sections.f90 | 6 +-
.../OpenMP/infinite-loop-in-construct.f90 | 19 +++
flang/test/Lower/OpenMP/sections.f90 | 6 +-
5 files changed, 119 insertions(+), 49 deletions(-)
create mode 100644 flang/test/Lower/OpenMP/infinite-loop-in-construct.f90
diff --git a/flang/include/flang/Lower/OpenMP.h
b/flang/include/flang/Lower/OpenMP.h
index 6e772c43d8c46e7..477d3e7d9da3a8f 100644
--- a/flang/include/flang/Lower/OpenMP.h
+++ b/flang/include/flang/Lower/OpenMP.h
@@ -50,8 +50,8 @@ struct Variable;
} // namespace pft
// Generate the OpenMP terminator for Operation at Location.
-void genOpenMPTerminator(fir::FirOpBuilder &, mlir::Operation *,
- mlir::Location);
+mlir::Operation *genOpenMPTerminator(fir::FirOpBuilder &, mlir::Operation *,
+ mlir::Location);
void genOpenMPConstruct(AbstractConverter &, Fortran::lower::SymMap &,
semantics::SemanticsContext &, pft::Evaluation &,
diff --git a/flang/lib/Lower/OpenMP.cpp b/flang/lib/Lower/OpenMP.cpp
index 6e93387fbff7cc9..bb92fdce4ac56ef 100644
--- a/flang/lib/Lower/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP.cpp
@@ -383,7 +383,8 @@ void
DataSharingProcessor::insertLastPrivateCompare(mlir::Operation *op) {
// construct
mlir::OpBuilder::InsertPoint unstructuredSectionsIP =
firOpBuilder.saveInsertionPoint();
-firOpBuilder.setInsertionPointToStart(&op->getRegion(0).back());
+mlir::Operation *lastOper =
op->getRegion(0).back().getTerminator();
+firOpBuilder.setInsertionPoint(lastOper);
lastPrivIP = firOpBuilder.saveInsertionPoint();
firOpBuilder.restoreInsertionPoint(unstructuredSectionsIP);
}
@@ -2133,15 +2134,6 @@ static mlir::Type
getLoopVarType(Fortran::lower::AbstractConverter &converter,
return converter.getFirOpBuilder().getIntegerType(loopVarTypeSize);
}
-static void resetBeforeTerminator(fir::FirOpBuilder &firOpBuilder,
- mlir::Operation *storeOp,
- mlir::Block &block) {
- if (storeOp)
-firOpBuilder.setInsertionPointAfter(storeOp);
- else
-firOpBuilder.setInsertionPointToStart(&block);
-}
-
static mlir::Operation *
createAndSetPrivatizedLoopVar(Fortran::lower::AbstractConverter &converter,
mlir::Location loc, mlir::Value indexVal,
@@ -2183,11 +2175,43 @@ static void createBodyOfOp(
const llvm::SmallVector &args = {},
bool outerCombined = false, DataSharingProcessor *dsp = nullptr) {
fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+
+ auto insertMarker = [](fir::FirOpBuilder &builder) {
+mlir::Value undef = builder.create(builder.getUnknownLoc(),
+ builder.getIndexType());
+return undef.getDefiningOp();
+ };
+
+ // Find the block where the OMP terminator should go. In simple cases
+ // it is the single block in the operation's region. When the region
+ // is more complicated, especially with unstructured control flow, there
+ // may be multiple blocks, and some of them may have non-OMP terminators
+ // resulting from lowering of the code contained within the operation.
+ // By OpenMP rules, there should be a single exit point from the region:
+ // here exit means transfering control to the code following the operation.
+ // STOP statement is allowed and does not count as exit for the purpose of
+ // inserting terminators.
+ auto findExitBlock = [&](mlir::Region ®ion) -> mlir::Block * {
+auto isTerminated = [](mlir::Block &block) -> bool {
+ if (block.empty())
+return false;
+ return block.back().hasTrait();
+};
+
+mlir::Block *exit = nullptr;
+for (auto &block : region) {
+ if (!isTerminated(block)) {
+assert(exit == nullptr && "Multiple exit block in OpenMP region");
+exit = █
+ }
+}
+return exit;
+ };
+
// If an argument for the region is provided then create the block with that
// argument. Also up
[llvm-branch-commits] [compiler-rt] [clang-tools-extra] [clang] [flang] [llvm] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
https://github.com/kparzysz updated
https://github.com/llvm/llvm-project/pull/77761
>From 1b5524ae8874e389d373a55417919afa56beb2b5 Mon Sep 17 00:00:00 2001
From: Krzysztof Parzyszek
Date: Mon, 8 Jan 2024 15:53:07 -0600
Subject: [PATCH 1/3] [Flang][OpenMP] Restructure recursive lowering in
`createBodyOfOp`
This brings `createBodyOfOp` to its final intended form. First, input
privatization is performed, then the recursive lowering takes place,
and finally the output privatization (lastprivate) is done.
This enables fixing a known issue with infinite loops inside of an
OpenMP region, and the fix is included in this patch.
Fixes https://github.com/llvm/llvm-project/issues/74348.
Recursive lowering [5/5]
---
flang/include/flang/Lower/OpenMP.h| 4 +-
flang/lib/Lower/OpenMP.cpp| 133 --
flang/test/Lower/OpenMP/FIR/sections.f90 | 6 +-
.../OpenMP/infinite-loop-in-construct.f90 | 19 +++
flang/test/Lower/OpenMP/sections.f90 | 6 +-
5 files changed, 119 insertions(+), 49 deletions(-)
create mode 100644 flang/test/Lower/OpenMP/infinite-loop-in-construct.f90
diff --git a/flang/include/flang/Lower/OpenMP.h
b/flang/include/flang/Lower/OpenMP.h
index 6e772c43d8c46e..477d3e7d9da3a8 100644
--- a/flang/include/flang/Lower/OpenMP.h
+++ b/flang/include/flang/Lower/OpenMP.h
@@ -50,8 +50,8 @@ struct Variable;
} // namespace pft
// Generate the OpenMP terminator for Operation at Location.
-void genOpenMPTerminator(fir::FirOpBuilder &, mlir::Operation *,
- mlir::Location);
+mlir::Operation *genOpenMPTerminator(fir::FirOpBuilder &, mlir::Operation *,
+ mlir::Location);
void genOpenMPConstruct(AbstractConverter &, Fortran::lower::SymMap &,
semantics::SemanticsContext &, pft::Evaluation &,
diff --git a/flang/lib/Lower/OpenMP.cpp b/flang/lib/Lower/OpenMP.cpp
index 6e93387fbff7cc..bb92fdce4ac56e 100644
--- a/flang/lib/Lower/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP.cpp
@@ -383,7 +383,8 @@ void
DataSharingProcessor::insertLastPrivateCompare(mlir::Operation *op) {
// construct
mlir::OpBuilder::InsertPoint unstructuredSectionsIP =
firOpBuilder.saveInsertionPoint();
-firOpBuilder.setInsertionPointToStart(&op->getRegion(0).back());
+mlir::Operation *lastOper =
op->getRegion(0).back().getTerminator();
+firOpBuilder.setInsertionPoint(lastOper);
lastPrivIP = firOpBuilder.saveInsertionPoint();
firOpBuilder.restoreInsertionPoint(unstructuredSectionsIP);
}
@@ -2133,15 +2134,6 @@ static mlir::Type
getLoopVarType(Fortran::lower::AbstractConverter &converter,
return converter.getFirOpBuilder().getIntegerType(loopVarTypeSize);
}
-static void resetBeforeTerminator(fir::FirOpBuilder &firOpBuilder,
- mlir::Operation *storeOp,
- mlir::Block &block) {
- if (storeOp)
-firOpBuilder.setInsertionPointAfter(storeOp);
- else
-firOpBuilder.setInsertionPointToStart(&block);
-}
-
static mlir::Operation *
createAndSetPrivatizedLoopVar(Fortran::lower::AbstractConverter &converter,
mlir::Location loc, mlir::Value indexVal,
@@ -2183,11 +2175,43 @@ static void createBodyOfOp(
const llvm::SmallVector &args = {},
bool outerCombined = false, DataSharingProcessor *dsp = nullptr) {
fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+
+ auto insertMarker = [](fir::FirOpBuilder &builder) {
+mlir::Value undef = builder.create(builder.getUnknownLoc(),
+ builder.getIndexType());
+return undef.getDefiningOp();
+ };
+
+ // Find the block where the OMP terminator should go. In simple cases
+ // it is the single block in the operation's region. When the region
+ // is more complicated, especially with unstructured control flow, there
+ // may be multiple blocks, and some of them may have non-OMP terminators
+ // resulting from lowering of the code contained within the operation.
+ // By OpenMP rules, there should be a single exit point from the region:
+ // here exit means transfering control to the code following the operation.
+ // STOP statement is allowed and does not count as exit for the purpose of
+ // inserting terminators.
+ auto findExitBlock = [&](mlir::Region ®ion) -> mlir::Block * {
+auto isTerminated = [](mlir::Block &block) -> bool {
+ if (block.empty())
+return false;
+ return block.back().hasTrait();
+};
+
+mlir::Block *exit = nullptr;
+for (auto &block : region) {
+ if (!isTerminated(block)) {
+assert(exit == nullptr && "Multiple exit block in OpenMP region");
+exit = █
+ }
+}
+return exit;
+ };
+
// If an argument for the region is provided then create the block with that
// argument. Also up
[llvm-branch-commits] [llvm] [clang] [SPARC] Prefer RDPC over CALL to implement GETPCX for 64-bit target (PR #77196)
brad0 wrote: > Uh oh, seems like I couldn't merge it with the tool either... Guess I'll open > a new PR to merge this, would that be okay? @brad0 @s-barannikov Ya, sure. Whatever path is easier and quickest for you. https://github.com/llvm/llvm-project/pull/77196 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [llvm] 1b922e1 - Revert "[DFAJumpThreading] Extends the bitwidth of state from uint64_t to APInt"
Author: Vitaly Buka
Date: 2024-01-15T17:50:16-08:00
New Revision: 1b922e19e3c0047206e7f6ba62e44d1f17880ec5
URL:
https://github.com/llvm/llvm-project/commit/1b922e19e3c0047206e7f6ba62e44d1f17880ec5
DIFF:
https://github.com/llvm/llvm-project/commit/1b922e19e3c0047206e7f6ba62e44d1f17880ec5.diff
LOG: Revert "[DFAJumpThreading] Extends the bitwidth of state from uint64_t to
APInt"
Added:
Modified:
llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp
llvm/test/Transforms/DFAJumpThreading/dfa-jump-threading-transform.ll
Removed:
diff --git a/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp
b/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp
index 0219f65618d8d8..c5bf913cda301b 100644
--- a/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp
+++ b/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp
@@ -307,7 +307,7 @@ void unfold(DomTreeUpdater *DTU, SelectInstToUnfold
SIToUnfold,
struct ClonedBlock {
BasicBlock *BB;
- APInt State; ///< \p State corresponds to the next value of a switch stmnt.
+ uint64_t State; ///< \p State corresponds to the next value of a switch
stmnt.
};
typedef std::deque PathType;
@@ -344,9 +344,9 @@ inline raw_ostream &operator<<(raw_ostream &OS, const
PathType &Path) {
/// exit state, and the block that determines the next state.
struct ThreadingPath {
/// Exit value is DFA's exit state for the given path.
- APInt getExitValue() const { return ExitVal; }
+ uint64_t getExitValue() const { return ExitVal; }
void setExitValue(const ConstantInt *V) {
-ExitVal = V->getValue();
+ExitVal = V->getZExtValue();
IsExitValSet = true;
}
bool isExitValueSet() const { return IsExitValSet; }
@@ -365,7 +365,7 @@ struct ThreadingPath {
private:
PathType Path;
- APInt ExitVal;
+ uint64_t ExitVal;
const BasicBlock *DBB = nullptr;
bool IsExitValSet = false;
};
@@ -744,7 +744,7 @@ struct TransformDFA {
for (ThreadingPath &TPath : SwitchPaths->getThreadingPaths()) {
PathType PathBBs = TPath.getPath();
- APInt NextState = TPath.getExitValue();
+ uint64_t NextState = TPath.getExitValue();
const BasicBlock *Determinator = TPath.getDeterminatorBB();
// Update Metrics for the Switch block, this is always cloned
@@ -901,7 +901,7 @@ struct TransformDFA {
DuplicateBlockMap &DuplicateMap,
SmallSet &BlocksToClean,
DomTreeUpdater *DTU) {
-APInt NextState = Path.getExitValue();
+uint64_t NextState = Path.getExitValue();
const BasicBlock *Determinator = Path.getDeterminatorBB();
PathType PathBBs = Path.getPath();
@@ -993,14 +993,13 @@ struct TransformDFA {
/// This function also includes updating phi nodes in the successors of the
/// BB, and remapping uses that were defined locally in the cloned BB.
BasicBlock *cloneBlockAndUpdatePredecessor(BasicBlock *BB, BasicBlock
*PrevBB,
- const APInt &NextState,
+ uint64_t NextState,
DuplicateBlockMap &DuplicateMap,
DefMap &NewDefs,
DomTreeUpdater *DTU) {
ValueToValueMapTy VMap;
BasicBlock *NewBB = CloneBasicBlock(
-BB, VMap, ".jt" + std::to_string(NextState.getLimitedValue()),
-BB->getParent());
+BB, VMap, ".jt" + std::to_string(NextState), BB->getParent());
NewBB->moveAfter(BB);
NumCloned++;
@@ -1035,7 +1034,7 @@ struct TransformDFA {
/// This means creating a new incoming value from NewBB with the new
/// instruction wherever there is an incoming value from BB.
void updateSuccessorPhis(BasicBlock *BB, BasicBlock *ClonedBB,
- const APInt &NextState, ValueToValueMapTy &VMap,
+ uint64_t NextState, ValueToValueMapTy &VMap,
DuplicateBlockMap &DuplicateMap) {
std::vector BlocksToUpdate;
@@ -1145,7 +1144,7 @@ struct TransformDFA {
void updateLastSuccessor(ThreadingPath &TPath,
DuplicateBlockMap &DuplicateMap,
DomTreeUpdater *DTU) {
-APInt NextState = TPath.getExitValue();
+uint64_t NextState = TPath.getExitValue();
BasicBlock *BB = TPath.getPath().back();
BasicBlock *LastBlock = getClonedBB(BB, NextState, DuplicateMap);
@@ -1199,7 +1198,7 @@ struct TransformDFA {
/// Checks if BB was already cloned for a particular next state value. If it
/// was then it returns this cloned block, and otherwise null.
- BasicBlock *getClonedBB(BasicBlock *BB, const APInt &NextState,
+ BasicBlock *getClonedBB(BasicBlock *BB, uint64_t NextState,
DuplicateBlockMap &DuplicateMap) {
CloneList ClonedBBs = DuplicateMap[BB];
@@ -1213,10 +1
[llvm-branch-commits] [llvm] [DirectX] Move DXIL ResourceKind and ElementType to DXILABI.h. NFC (PR #78225)
https://github.com/bogner created https://github.com/llvm/llvm-project/pull/78225 None ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [llvm] [DirectX] Move DXIL ResourceKind and ElementType to DXILABI.h. NFC (PR #78225)
llvmbot wrote:
@llvm/pr-subscribers-backend-directx
Author: Justin Bogner (bogner)
Changes
---
Full diff: https://github.com/llvm/llvm-project/pull/78225.diff
4 Files Affected:
- (modified) llvm/include/llvm/Frontend/HLSL/HLSLResource.h (+4-48)
- (modified) llvm/include/llvm/Support/DXILABI.h (+48)
- (modified) llvm/lib/Target/DirectX/DXILResource.cpp (+46-46)
- (modified) llvm/lib/Target/DirectX/DXILResource.h (+11-10)
``diff
diff --git a/llvm/include/llvm/Frontend/HLSL/HLSLResource.h
b/llvm/include/llvm/Frontend/HLSL/HLSLResource.h
index 068b4c66711e467..dcf78ec856a9d64 100644
--- a/llvm/include/llvm/Frontend/HLSL/HLSLResource.h
+++ b/llvm/include/llvm/Frontend/HLSL/HLSLResource.h
@@ -14,6 +14,7 @@
#define LLVM_FRONTEND_HLSL_HLSLRESOURCE_H
#include "llvm/IR/Metadata.h"
+#include "llvm/Support/DXILABI.h"
namespace llvm {
class GlobalVariable;
@@ -29,54 +30,9 @@ enum class ResourceClass : uint8_t {
NumClasses = Invalid,
};
-// The value ordering of this enumeration is part of the DXIL ABI. Elements
-// can only be added to the end, and not removed.
-enum class ResourceKind : uint32_t {
- Invalid = 0,
- Texture1D,
- Texture2D,
- Texture2DMS,
- Texture3D,
- TextureCube,
- Texture1DArray,
- Texture2DArray,
- Texture2DMSArray,
- TextureCubeArray,
- TypedBuffer,
- RawBuffer,
- StructuredBuffer,
- CBuffer,
- Sampler,
- TBuffer,
- RTAccelerationStructure,
- FeedbackTexture2D,
- FeedbackTexture2DArray,
- NumEntries,
-};
-
-// The value ordering of this enumeration is part of the DXIL ABI. Elements
-// can only be added to the end, and not removed.
-enum class ElementType : uint32_t {
- Invalid = 0,
- I1,
- I16,
- U16,
- I32,
- U32,
- I64,
- U64,
- F16,
- F32,
- F64,
- SNormF16,
- UNormF16,
- SNormF32,
- UNormF32,
- SNormF64,
- UNormF64,
- PackedS8x32,
- PackedU8x32,
-};
+// For now we use DXIL ABI enum values directly. This may change in the future.
+using dxil::ResourceKind;
+using dxil::ElementType;
class FrontendResource {
MDNode *Entry;
diff --git a/llvm/include/llvm/Support/DXILABI.h
b/llvm/include/llvm/Support/DXILABI.h
index e2a8fbad16bb3f7..c1d81775b6711e3 100644
--- a/llvm/include/llvm/Support/DXILABI.h
+++ b/llvm/include/llvm/Support/DXILABI.h
@@ -39,6 +39,54 @@ enum class ParameterKind : uint8_t {
DXIL_HANDLE,
};
+/// The kind of resource for an SRV or UAV resource. Sometimes referred to as
+/// "Shape" in the DXIL docs.
+enum class ResourceKind : uint32_t {
+ Invalid = 0,
+ Texture1D,
+ Texture2D,
+ Texture2DMS,
+ Texture3D,
+ TextureCube,
+ Texture1DArray,
+ Texture2DArray,
+ Texture2DMSArray,
+ TextureCubeArray,
+ TypedBuffer,
+ RawBuffer,
+ StructuredBuffer,
+ CBuffer,
+ Sampler,
+ TBuffer,
+ RTAccelerationStructure,
+ FeedbackTexture2D,
+ FeedbackTexture2DArray,
+ NumEntries,
+};
+
+/// The element type of an SRV or UAV resource.
+enum class ElementType : uint32_t {
+ Invalid = 0,
+ I1,
+ I16,
+ U16,
+ I32,
+ U32,
+ I64,
+ U64,
+ F16,
+ F32,
+ F64,
+ SNormF16,
+ UNormF16,
+ SNormF32,
+ UNormF32,
+ SNormF64,
+ UNormF64,
+ PackedS8x32,
+ PackedU8x32,
+};
+
} // namespace dxil
} // namespace llvm
diff --git a/llvm/lib/Target/DirectX/DXILResource.cpp
b/llvm/lib/Target/DirectX/DXILResource.cpp
index 621852f2453ff51..8e5b9867e6661bf 100644
--- a/llvm/lib/Target/DirectX/DXILResource.cpp
+++ b/llvm/lib/Target/DirectX/DXILResource.cpp
@@ -21,7 +21,6 @@
using namespace llvm;
using namespace llvm::dxil;
-using namespace llvm::hlsl;
template void ResourceTable::collect(Module &M) {
NamedMDNode *Entry = M.getNamedMetadata(MDName);
@@ -30,7 +29,7 @@ template void ResourceTable::collect(Module
&M) {
uint32_t Counter = 0;
for (auto *Res : Entry->operands()) {
-Data.push_back(T(Counter++, FrontendResource(cast(Res;
+Data.push_back(T(Counter++, hlsl::FrontendResource(cast(Res;
}
}
@@ -42,7 +41,7 @@ template <> void
ResourceTable::collect(Module &M) {
uint32_t Counter = 0;
for (auto *Res : Entry->operands()) {
Data.push_back(
-ConstantBuffer(Counter++, FrontendResource(cast(Res;
+ConstantBuffer(Counter++, hlsl::FrontendResource(cast(Res;
}
// FIXME: share CBufferDataLayout with CBuffer load lowering.
// See https://github.com/llvm/llvm-project/issues/58381
@@ -56,7 +55,7 @@ void Resources::collect(Module &M) {
CBuffers.collect(M);
}
-ResourceBase::ResourceBase(uint32_t I, FrontendResource R)
+ResourceBase::ResourceBase(uint32_t I, hlsl::FrontendResource R)
: ID(I), GV(R.getGlobalVariable()), Name(""), Space(R.getSpace()),
LowerBound(R.getResourceIndex()), RangeSize(1) {
if (auto *ArrTy = dyn_cast(GV->getValueType()))
@@ -107,83 +106,84 @@ StringRef ResourceBase::getElementTypeName(ElementType
ElTy) {
llvm_unreachable("All ElementType enums are handled in switch");
}
-void ResourceBase::printElementType(Kinds Kind, ElementType ElTy,
+void ResourceBase::printElementT
[llvm-branch-commits] [llvm] [DirectX] Move DXIL ResourceKind and ElementType to DXILABI.h. NFC (PR #78225)
github-actions[bot] wrote:
:warning: C/C++ code formatter, clang-format found issues in your code.
:warning:
You can test this locally with the following command:
``bash
git-clang-format --diff b0fff11ff5d46deb5df3485b87b22bb61e93910f
edca247b1dec13f7549ad7d43881d0c2b8f34d62 --
llvm/include/llvm/Frontend/HLSL/HLSLResource.h
llvm/include/llvm/Support/DXILABI.h llvm/lib/Target/DirectX/DXILResource.cpp
llvm/lib/Target/DirectX/DXILResource.h
``
View the diff from clang-format here.
``diff
diff --git a/llvm/include/llvm/Frontend/HLSL/HLSLResource.h
b/llvm/include/llvm/Frontend/HLSL/HLSLResource.h
index dcf78ec856..edfcbda0a3 100644
--- a/llvm/include/llvm/Frontend/HLSL/HLSLResource.h
+++ b/llvm/include/llvm/Frontend/HLSL/HLSLResource.h
@@ -31,8 +31,8 @@ enum class ResourceClass : uint8_t {
};
// For now we use DXIL ABI enum values directly. This may change in the future.
-using dxil::ResourceKind;
using dxil::ElementType;
+using dxil::ResourceKind;
class FrontendResource {
MDNode *Entry;
``
https://github.com/llvm/llvm-project/pull/78225
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
