[flang][cuda] Adapt TargetRewrite to support gpu.launch_func #119933
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
llvmbot
added
flang
|
@llvm/pr-subscribers-flang-fir-hlfir Author: Valentin Clement (バレンタイン クレメン) (clementval) ChangesThe gpu.func are already supported in the TargetRewrite pass. Update the pass to also support rewriting the gpu.launch_func operation. Full diff: https://github.com/llvm/llvm-project/pull/119933.diff 2 Files Affected:
diff --git a/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp b/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp
index 5a042b34a58c0a..b0b9499557e2b7 100644
--- a/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp
+++ b/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp
@@ -134,10 +134,18 @@ class TargetRewrite : public fir::impl::TargetRewritePassBase<TargetRewrite> {
mod.walk([&](mlir::Operation *op) {
if (auto call = mlir::dyn_cast<fir::CallOp>(op)) {
if (!hasPortableSignature(call.getFunctionType(), op))
- convertCallOp(call);
+ convertCallOp(call, call.getFunctionType());
} else if (auto dispatch = mlir::dyn_cast<fir::DispatchOp>(op)) {
if (!hasPortableSignature(dispatch.getFunctionType(), op))
- convertCallOp(dispatch);
+ convertCallOp(dispatch, dispatch.getFunctionType());
+ } else if (auto gpuLaunchFunc =
+ mlir::dyn_cast<mlir::gpu::LaunchFuncOp>(op)) {
+ llvm::SmallVector<mlir::Type> operandsTypes;
+ for (auto arg : gpuLaunchFunc.getKernelOperands())
+ operandsTypes.push_back(arg.getType());
+ auto fctTy = mlir::FunctionType::get(&context, operandsTypes, {});
+ if (!hasPortableSignature(fctTy, op))
+ convertCallOp(gpuLaunchFunc, fctTy);
} else if (auto addr = mlir::dyn_cast<fir::AddrOfOp>(op)) {
if (mlir::isa<mlir::FunctionType>(addr.getType()) &&
!hasPortableSignature(addr.getType(), op))
@@ -357,8 +365,7 @@ class TargetRewrite : public fir::impl::TargetRewritePassBase<TargetRewrite> {
// Convert fir.call and fir.dispatch Ops.
template <typename A>
- void convertCallOp(A callOp) {
- auto fnTy = callOp.getFunctionType();
+ void convertCallOp(A callOp, mlir::FunctionType fnTy) {
auto loc = callOp.getLoc();
rewriter->setInsertionPoint(callOp);
llvm::SmallVector<mlir::Type> newResTys;
@@ -376,7 +383,7 @@ class TargetRewrite : public fir::impl::TargetRewritePassBase<TargetRewrite> {
newOpers.push_back(callOp.getOperand(0));
dropFront = 1;
}
- } else {
+ } else if constexpr (std::is_same_v<std::decay_t<A>, fir::DispatchOp>) {
dropFront = 1; // First operand is the polymorphic object.
}
@@ -402,10 +409,14 @@ class TargetRewrite : public fir::impl::TargetRewritePassBase<TargetRewrite> {
llvm::SmallVector<mlir::Type> trailingInTys;
llvm::SmallVector<mlir::Value> trailingOpers;
+ llvm::SmallVector<mlir::Value> operands;
unsigned passArgShift = 0;
+ if constexpr (std::is_same_v<std::decay_t<A>, mlir::gpu::LaunchFuncOp>)
+ operands = callOp.getKernelOperands();
+ else
+ operands = callOp.getOperands().drop_front(dropFront);
for (auto e : llvm::enumerate(
- llvm::zip(fnTy.getInputs().drop_front(dropFront),
- callOp.getOperands().drop_front(dropFront)))) {
+ llvm::zip(fnTy.getInputs().drop_front(dropFront), operands))) {
mlir::Type ty = std::get<0>(e.value());
mlir::Value oper = std::get<1>(e.value());
unsigned index = e.index();
@@ -507,7 +518,19 @@ class TargetRewrite : public fir::impl::TargetRewritePassBase<TargetRewrite> {
newOpers.insert(newOpers.end(), trailingOpers.begin(), trailingOpers.end());
llvm::SmallVector<mlir::Value, 1> newCallResults;
- if constexpr (std::is_same_v<std::decay_t<A>, fir::CallOp>) {
+ if constexpr (std::is_same_v<std::decay_t<A>, mlir::gpu::LaunchFuncOp>) {
+ auto newCall = rewriter->create<A>(
+ loc, callOp.getKernel(), callOp.getGridSizeOperandValues(),
+ callOp.getBlockSizeOperandValues(),
+ callOp.getDynamicSharedMemorySize(), newOpers);
+ if (callOp.getClusterSizeX())
+ newCall.getClusterSizeXMutable().assign(callOp.getClusterSizeX());
+ if (callOp.getClusterSizeY())
+ newCall.getClusterSizeYMutable().assign(callOp.getClusterSizeY());
+ if (callOp.getClusterSizeZ())
+ newCall.getClusterSizeZMutable().assign(callOp.getClusterSizeZ());
+ newCallResults.append(newCall.result_begin(), newCall.result_end());
+ } else if constexpr (std::is_same_v<std::decay_t<A>, fir::CallOp>) {
fir::CallOp newCall;
if (callOp.getCallee()) {
newCall =
diff --git a/flang/test/Fir/CUDA/cuda-target-rewrite.mlir b/flang/test/Fir/CUDA/cuda-target-rewrite.mlir
index d88b6776795a0b..0e7534e06c89c9 100644
--- a/flang/test/Fir/CUDA/cuda-target-rewrite.mlir
+++ b/flang/test/Fir/CUDA/cuda-target-rewrite.mlir
@@ -27,3 +27,27 @@ gpu.module @testmod {
// CHECK-LABEL: gpu.func @_QPtest
// CHECK-SAME: (%arg0: f64, %arg1: f64) -> tuple<f64, f64> {
// CHECK: gpu.return %{{.*}} : tuple<f64, f64>
+
+
+// -----
+module attributes {gpu.container_module} {
+
+gpu.module @testmod {
+ gpu.func @_QPtest(%arg0: complex<f64>) -> () kernel {
+ gpu.return
+ }
+}
+
+func.func @main(%arg0: complex<f64>) {
+ %0 = llvm.mlir.constant(0 : i64) : i64
+ %1 = llvm.mlir.constant(0 : i32) : i32
+ gpu.launch_func @testmod::@_QPtest blocks in (%0, %0, %0) threads in (%0, %0, %0) : i64 dynamic_shared_memory_size %1 args(%arg0 : complex<f64>)
+ return
+}
+
+}
+
+// CHECK-LABEL: gpu.func @_QPtest
+// CHECK-SAME: (%arg0: f64, %arg1: f64) kernel {
+// CHECK: gpu.return
+// CHECK: gpu.launch_func @testmod::@_QPtest blocks in (%{{.*}}, %{{.*}}, %{{.*}}) threads in (%{{.*}}, %{{.*}}, %{{.*}}) : i64 dynamic_shared_memory_size %{{.*}} args(%{{.*}} : f64, %{{.*}} : f64)
|
|
@llvm/pr-subscribers-flang-codegen Author: Valentin Clement (バレンタイン クレメン) (clementval) ChangesThe gpu.func are already supported in the TargetRewrite pass. Update the pass to also support rewriting the gpu.launch_func operation. Full diff: https://github.com/llvm/llvm-project/pull/119933.diff 2 Files Affected:
diff --git a/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp b/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp
index 5a042b34a58c0a..b0b9499557e2b7 100644
--- a/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp
+++ b/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp
@@ -134,10 +134,18 @@ class TargetRewrite : public fir::impl::TargetRewritePassBase<TargetRewrite> {
mod.walk([&](mlir::Operation *op) {
if (auto call = mlir::dyn_cast<fir::CallOp>(op)) {
if (!hasPortableSignature(call.getFunctionType(), op))
- convertCallOp(call);
+ convertCallOp(call, call.getFunctionType());
} else if (auto dispatch = mlir::dyn_cast<fir::DispatchOp>(op)) {
if (!hasPortableSignature(dispatch.getFunctionType(), op))
- convertCallOp(dispatch);
+ convertCallOp(dispatch, dispatch.getFunctionType());
+ } else if (auto gpuLaunchFunc =
+ mlir::dyn_cast<mlir::gpu::LaunchFuncOp>(op)) {
+ llvm::SmallVector<mlir::Type> operandsTypes;
+ for (auto arg : gpuLaunchFunc.getKernelOperands())
+ operandsTypes.push_back(arg.getType());
+ auto fctTy = mlir::FunctionType::get(&context, operandsTypes, {});
+ if (!hasPortableSignature(fctTy, op))
+ convertCallOp(gpuLaunchFunc, fctTy);
} else if (auto addr = mlir::dyn_cast<fir::AddrOfOp>(op)) {
if (mlir::isa<mlir::FunctionType>(addr.getType()) &&
!hasPortableSignature(addr.getType(), op))
@@ -357,8 +365,7 @@ class TargetRewrite : public fir::impl::TargetRewritePassBase<TargetRewrite> {
// Convert fir.call and fir.dispatch Ops.
template <typename A>
- void convertCallOp(A callOp) {
- auto fnTy = callOp.getFunctionType();
+ void convertCallOp(A callOp, mlir::FunctionType fnTy) {
auto loc = callOp.getLoc();
rewriter->setInsertionPoint(callOp);
llvm::SmallVector<mlir::Type> newResTys;
@@ -376,7 +383,7 @@ class TargetRewrite : public fir::impl::TargetRewritePassBase<TargetRewrite> {
newOpers.push_back(callOp.getOperand(0));
dropFront = 1;
}
- } else {
+ } else if constexpr (std::is_same_v<std::decay_t<A>, fir::DispatchOp>) {
dropFront = 1; // First operand is the polymorphic object.
}
@@ -402,10 +409,14 @@ class TargetRewrite : public fir::impl::TargetRewritePassBase<TargetRewrite> {
llvm::SmallVector<mlir::Type> trailingInTys;
llvm::SmallVector<mlir::Value> trailingOpers;
+ llvm::SmallVector<mlir::Value> operands;
unsigned passArgShift = 0;
+ if constexpr (std::is_same_v<std::decay_t<A>, mlir::gpu::LaunchFuncOp>)
+ operands = callOp.getKernelOperands();
+ else
+ operands = callOp.getOperands().drop_front(dropFront);
for (auto e : llvm::enumerate(
- llvm::zip(fnTy.getInputs().drop_front(dropFront),
- callOp.getOperands().drop_front(dropFront)))) {
+ llvm::zip(fnTy.getInputs().drop_front(dropFront), operands))) {
mlir::Type ty = std::get<0>(e.value());
mlir::Value oper = std::get<1>(e.value());
unsigned index = e.index();
@@ -507,7 +518,19 @@ class TargetRewrite : public fir::impl::TargetRewritePassBase<TargetRewrite> {
newOpers.insert(newOpers.end(), trailingOpers.begin(), trailingOpers.end());
llvm::SmallVector<mlir::Value, 1> newCallResults;
- if constexpr (std::is_same_v<std::decay_t<A>, fir::CallOp>) {
+ if constexpr (std::is_same_v<std::decay_t<A>, mlir::gpu::LaunchFuncOp>) {
+ auto newCall = rewriter->create<A>(
+ loc, callOp.getKernel(), callOp.getGridSizeOperandValues(),
+ callOp.getBlockSizeOperandValues(),
+ callOp.getDynamicSharedMemorySize(), newOpers);
+ if (callOp.getClusterSizeX())
+ newCall.getClusterSizeXMutable().assign(callOp.getClusterSizeX());
+ if (callOp.getClusterSizeY())
+ newCall.getClusterSizeYMutable().assign(callOp.getClusterSizeY());
+ if (callOp.getClusterSizeZ())
+ newCall.getClusterSizeZMutable().assign(callOp.getClusterSizeZ());
+ newCallResults.append(newCall.result_begin(), newCall.result_end());
+ } else if constexpr (std::is_same_v<std::decay_t<A>, fir::CallOp>) {
fir::CallOp newCall;
if (callOp.getCallee()) {
newCall =
diff --git a/flang/test/Fir/CUDA/cuda-target-rewrite.mlir b/flang/test/Fir/CUDA/cuda-target-rewrite.mlir
index d88b6776795a0b..0e7534e06c89c9 100644
--- a/flang/test/Fir/CUDA/cuda-target-rewrite.mlir
+++ b/flang/test/Fir/CUDA/cuda-target-rewrite.mlir
@@ -27,3 +27,27 @@ gpu.module @testmod {
// CHECK-LABEL: gpu.func @_QPtest
// CHECK-SAME: (%arg0: f64, %arg1: f64) -> tuple<f64, f64> {
// CHECK: gpu.return %{{.*}} : tuple<f64, f64>
+
+
+// -----
+module attributes {gpu.container_module} {
+
+gpu.module @testmod {
+ gpu.func @_QPtest(%arg0: complex<f64>) -> () kernel {
+ gpu.return
+ }
+}
+
+func.func @main(%arg0: complex<f64>) {
+ %0 = llvm.mlir.constant(0 : i64) : i64
+ %1 = llvm.mlir.constant(0 : i32) : i32
+ gpu.launch_func @testmod::@_QPtest blocks in (%0, %0, %0) threads in (%0, %0, %0) : i64 dynamic_shared_memory_size %1 args(%arg0 : complex<f64>)
+ return
+}
+
+}
+
+// CHECK-LABEL: gpu.func @_QPtest
+// CHECK-SAME: (%arg0: f64, %arg1: f64) kernel {
+// CHECK: gpu.return
+// CHECK: gpu.launch_func @testmod::@_QPtest blocks in (%{{.*}}, %{{.*}}, %{{.*}}) threads in (%{{.*}}, %{{.*}}, %{{.*}}) : i64 dynamic_shared_memory_size %{{.*}} args(%{{.*}} : f64, %{{.*}} : f64)
|
jeanPerier
approved these changes
Dec 16, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The gpu.func are already supported in the TargetRewrite pass. Update the pass to also support rewriting the gpu.launch_func operation.