[MLIR][GPU] Generalize gpu.printf op lowering to LLVM call pattern. #164297
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
other than AMD gpu OpenCL runtime.
Member
|
@llvm/pr-subscribers-mlir-gpu Author: Sang Ik Lee (silee2) ChangesExisting pattern for lowering gpu.printf op to LLVM call uses fixed function name and calling convention. Full diff: https://github.com/llvm/llvm-project/pull/164297.diff 5 Files Affected:
diff --git a/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.cpp b/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.cpp
index 2285d2695db4e..eb662a1b056de 100644
--- a/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.cpp
+++ b/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.cpp
@@ -507,7 +507,8 @@ LogicalResult GPUPrintfOpToLLVMCallLowering::matchAndRewrite(
LLVM::LLVMFunctionType::get(rewriter.getI32Type(), {ptrType},
/*isVarArg=*/true);
LLVM::LLVMFuncOp printfDecl =
- getOrDefineFunction(moduleOp, loc, rewriter, "printf", printfType);
+ getOrDefineFunction(moduleOp, loc, rewriter, funcName, printfType);
+ printfDecl.setCConv(callingConvention);
// Create the global op or find an existing one.
LLVM::GlobalOp global = getOrCreateStringConstant(
@@ -530,7 +531,8 @@ LogicalResult GPUPrintfOpToLLVMCallLowering::matchAndRewrite(
printfArgs.push_back(stringStart);
printfArgs.append(argsRange.begin(), argsRange.end());
- LLVM::CallOp::create(rewriter, loc, printfDecl, printfArgs);
+ auto call = LLVM::CallOp::create(rewriter, loc, printfDecl, printfArgs);
+ call.setCConv(callingConvention);
rewriter.eraseOp(gpuPrintfOp);
return success();
}
diff --git a/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.h b/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.h
index 66d3bb40a8f5a..adf5ba2feb591 100644
--- a/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.h
+++ b/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.h
@@ -10,6 +10,7 @@
#include "mlir/Conversion/LLVMCommon/Pattern.h"
#include "mlir/Dialect/GPU/IR/GPUDialect.h"
+#include "mlir/Dialect/LLVMIR/LLVMAttrs.h"
#include "mlir/Dialect/LLVMIR/LLVMDialect.h"
namespace mlir {
@@ -142,13 +143,17 @@ struct GPUPrintfOpToHIPLowering : public ConvertOpToLLVMPattern<gpu::PrintfOp> {
/// This pass will add a declaration of printf() to the GPUModule if needed
/// and separate out the format strings into global constants. For some
/// runtimes, such as OpenCL on AMD, this is sufficient setup, as the compiler
-/// will lower printf calls to appropriate device-side code
+/// will lower printf calls to appropriate device-side code.
+/// callingConvention and funcName can be adjusted as needed.
struct GPUPrintfOpToLLVMCallLowering
: public ConvertOpToLLVMPattern<gpu::PrintfOp> {
- GPUPrintfOpToLLVMCallLowering(const LLVMTypeConverter &converter,
- int addressSpace = 0)
+ GPUPrintfOpToLLVMCallLowering(
+ const LLVMTypeConverter &converter, int addressSpace = 0,
+ LLVM::cconv::CConv callingConvention = LLVM::cconv::CConv::C,
+ StringRef funcName = "printf")
: ConvertOpToLLVMPattern<gpu::PrintfOp>(converter),
- addressSpace(addressSpace) {}
+ addressSpace(addressSpace), callingConvention(callingConvention),
+ funcName(funcName) {}
LogicalResult
matchAndRewrite(gpu::PrintfOp gpuPrintfOp, gpu::PrintfOpAdaptor adaptor,
@@ -156,6 +161,8 @@ struct GPUPrintfOpToLLVMCallLowering
private:
int addressSpace;
+ LLVM::cconv::CConv callingConvention;
+ StringRef funcName;
};
/// Lowering of gpu.printf to a vprintf standard library.
diff --git a/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp b/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
index c2363a1a40294..25f1e1b184d61 100644
--- a/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
+++ b/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
@@ -470,10 +470,13 @@ struct GPUToLLVMSPVConversionPass final
gpu::GPUFuncOp, gpu::GlobalIdOp, gpu::GridDimOp,
gpu::LaneIdOp, gpu::NumSubgroupsOp, gpu::ReturnOp,
gpu::ShuffleOp, gpu::SubgroupIdOp, gpu::SubgroupSizeOp,
- gpu::ThreadIdOp>();
+ gpu::ThreadIdOp, gpu::PrintfOp>();
populateGpuToLLVMSPVConversionPatterns(converter, patterns);
populateGpuMemorySpaceAttributeConversions(converter);
+ patterns.add<GPUPrintfOpToLLVMCallLowering>(converter, /*addressSpace=*/2,
+ LLVM::cconv::CConv::SPIR_FUNC,
+ "_Z6printfPU3AS2Kcz");
if (failed(applyPartialConversion(getOperation(), target,
std::move(patterns))))
diff --git a/mlir/test/Conversion/GPUToLLVMSPV/printf.mlir b/mlir/test/Conversion/GPUToLLVMSPV/printf.mlir
new file mode 100644
index 0000000000000..74017e8354cf1
--- /dev/null
+++ b/mlir/test/Conversion/GPUToLLVMSPV/printf.mlir
@@ -0,0 +1,16 @@
+// RUN: mlir-opt %s -convert-gpu-to-llvm-spv | FileCheck %s
+
+gpu.module @test_module {
+ // CHECK: llvm.mlir.global internal constant @[[$PRINT_GLOBAL:[A-Za-z0-9_]+]]("Hello: %d\0A\00") {addr_space = 2 : i32}
+ // CHECK: llvm.func spir_funccc @_Z6printfPU3AS2Kcz(!llvm.ptr<2>, ...) -> i32
+ // CHECK-LABEL: llvm.func spir_funccc @test_printf
+ // CHECK: (%[[ARG0:.*]]: i32)
+ gpu.func @test_printf(%arg0: i32) {
+ // CHECK: %[[IMM0:.*]] = llvm.mlir.addressof @[[$PRINT_GLOBAL]] : !llvm.ptr<2>
+ // CHECK-NEXT: %[[IMM2:.*]] = llvm.getelementptr %[[IMM0]][0, 0] : (!llvm.ptr<2>) -> !llvm.ptr<2>, !llvm.array<11 x i8>
+ // CHECK-NEXT: %{{.*}} = llvm.call spir_funccc @_Z6printfPU3AS2Kcz(%[[IMM2]], %[[ARG0]]) vararg(!llvm.func<i32 (ptr<2>, ...)>) : (!llvm.ptr<2>, i32) -> i32
+ gpu.printf "Hello: %d\n", %arg0 : i32
+ gpu.return
+ }
+}
+
diff --git a/mlir/test/Integration/Dialect/XeVM/GPU/gpu_printf.mlir b/mlir/test/Integration/Dialect/XeVM/GPU/gpu_printf.mlir
new file mode 100644
index 0000000000000..edf8775c72418
--- /dev/null
+++ b/mlir/test/Integration/Dialect/XeVM/GPU/gpu_printf.mlir
@@ -0,0 +1,30 @@
+// RUN: mlir-opt %s \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(cse,func.func(gpu-async-region),xevm-attach-target,gpu.module(convert-gpu-to-llvm-spv{use-64bit-index=true},convert-xevm-to-llvm,cse))' \
+// RUN: | mlir-opt -convert-scf-to-cf -convert-cf-to-llvm -convert-vector-to-llvm -convert-arith-to-llvm \
+// RUN: | mlir-opt -gpu-to-llvm -reconcile-unrealized-casts -cse -gpu-module-to-binary \
+// RUN: | mlir-runner \
+// RUN: --shared-libs=%mlir_sycl_runtime \
+// RUN: --shared-libs=%mlir_runner_utils \
+// RUN: --shared-libs=%mlir_c_runner_utils \
+// RUN: --entry-point-result=void \
+// RUN: | FileCheck %s
+
+module @test attributes {gpu.container_module} {
+ gpu.module @test_module {
+ gpu.func @test_printf(%arg0: i32, %arg1: f32) kernel {
+ gpu.printf "Hello: %d\n", %arg0 : i32
+ gpu.printf "Hello: %f\n", %arg1 : f32
+ gpu.return
+ }
+ }
+
+ func.func @main() attributes {llvm.emit_c_interface} {
+ %c1 = arith.constant 1 : index
+ %c11 = arith.constant 11 : i32
+ %c4 = arith.constant 4.0 : f32
+ // CHECK: Hello: 11
+ // CHECK: Hello: 4.000000
+ gpu.launch_func @test_module::@test_printf blocks in (%c1, %c1, %c1) threads in (%c1, %c1, %c1) args(%c11 : i32, %c4 : f32)
+ return
+ }
+}
|
Member
|
@llvm/pr-subscribers-mlir Author: Sang Ik Lee (silee2) ChangesExisting pattern for lowering gpu.printf op to LLVM call uses fixed function name and calling convention. Full diff: https://github.com/llvm/llvm-project/pull/164297.diff 5 Files Affected:
diff --git a/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.cpp b/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.cpp
index 2285d2695db4e..eb662a1b056de 100644
--- a/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.cpp
+++ b/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.cpp
@@ -507,7 +507,8 @@ LogicalResult GPUPrintfOpToLLVMCallLowering::matchAndRewrite(
LLVM::LLVMFunctionType::get(rewriter.getI32Type(), {ptrType},
/*isVarArg=*/true);
LLVM::LLVMFuncOp printfDecl =
- getOrDefineFunction(moduleOp, loc, rewriter, "printf", printfType);
+ getOrDefineFunction(moduleOp, loc, rewriter, funcName, printfType);
+ printfDecl.setCConv(callingConvention);
// Create the global op or find an existing one.
LLVM::GlobalOp global = getOrCreateStringConstant(
@@ -530,7 +531,8 @@ LogicalResult GPUPrintfOpToLLVMCallLowering::matchAndRewrite(
printfArgs.push_back(stringStart);
printfArgs.append(argsRange.begin(), argsRange.end());
- LLVM::CallOp::create(rewriter, loc, printfDecl, printfArgs);
+ auto call = LLVM::CallOp::create(rewriter, loc, printfDecl, printfArgs);
+ call.setCConv(callingConvention);
rewriter.eraseOp(gpuPrintfOp);
return success();
}
diff --git a/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.h b/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.h
index 66d3bb40a8f5a..adf5ba2feb591 100644
--- a/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.h
+++ b/mlir/lib/Conversion/GPUCommon/GPUOpsLowering.h
@@ -10,6 +10,7 @@
#include "mlir/Conversion/LLVMCommon/Pattern.h"
#include "mlir/Dialect/GPU/IR/GPUDialect.h"
+#include "mlir/Dialect/LLVMIR/LLVMAttrs.h"
#include "mlir/Dialect/LLVMIR/LLVMDialect.h"
namespace mlir {
@@ -142,13 +143,17 @@ struct GPUPrintfOpToHIPLowering : public ConvertOpToLLVMPattern<gpu::PrintfOp> {
/// This pass will add a declaration of printf() to the GPUModule if needed
/// and separate out the format strings into global constants. For some
/// runtimes, such as OpenCL on AMD, this is sufficient setup, as the compiler
-/// will lower printf calls to appropriate device-side code
+/// will lower printf calls to appropriate device-side code.
+/// callingConvention and funcName can be adjusted as needed.
struct GPUPrintfOpToLLVMCallLowering
: public ConvertOpToLLVMPattern<gpu::PrintfOp> {
- GPUPrintfOpToLLVMCallLowering(const LLVMTypeConverter &converter,
- int addressSpace = 0)
+ GPUPrintfOpToLLVMCallLowering(
+ const LLVMTypeConverter &converter, int addressSpace = 0,
+ LLVM::cconv::CConv callingConvention = LLVM::cconv::CConv::C,
+ StringRef funcName = "printf")
: ConvertOpToLLVMPattern<gpu::PrintfOp>(converter),
- addressSpace(addressSpace) {}
+ addressSpace(addressSpace), callingConvention(callingConvention),
+ funcName(funcName) {}
LogicalResult
matchAndRewrite(gpu::PrintfOp gpuPrintfOp, gpu::PrintfOpAdaptor adaptor,
@@ -156,6 +161,8 @@ struct GPUPrintfOpToLLVMCallLowering
private:
int addressSpace;
+ LLVM::cconv::CConv callingConvention;
+ StringRef funcName;
};
/// Lowering of gpu.printf to a vprintf standard library.
diff --git a/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp b/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
index c2363a1a40294..25f1e1b184d61 100644
--- a/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
+++ b/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
@@ -470,10 +470,13 @@ struct GPUToLLVMSPVConversionPass final
gpu::GPUFuncOp, gpu::GlobalIdOp, gpu::GridDimOp,
gpu::LaneIdOp, gpu::NumSubgroupsOp, gpu::ReturnOp,
gpu::ShuffleOp, gpu::SubgroupIdOp, gpu::SubgroupSizeOp,
- gpu::ThreadIdOp>();
+ gpu::ThreadIdOp, gpu::PrintfOp>();
populateGpuToLLVMSPVConversionPatterns(converter, patterns);
populateGpuMemorySpaceAttributeConversions(converter);
+ patterns.add<GPUPrintfOpToLLVMCallLowering>(converter, /*addressSpace=*/2,
+ LLVM::cconv::CConv::SPIR_FUNC,
+ "_Z6printfPU3AS2Kcz");
if (failed(applyPartialConversion(getOperation(), target,
std::move(patterns))))
diff --git a/mlir/test/Conversion/GPUToLLVMSPV/printf.mlir b/mlir/test/Conversion/GPUToLLVMSPV/printf.mlir
new file mode 100644
index 0000000000000..74017e8354cf1
--- /dev/null
+++ b/mlir/test/Conversion/GPUToLLVMSPV/printf.mlir
@@ -0,0 +1,16 @@
+// RUN: mlir-opt %s -convert-gpu-to-llvm-spv | FileCheck %s
+
+gpu.module @test_module {
+ // CHECK: llvm.mlir.global internal constant @[[$PRINT_GLOBAL:[A-Za-z0-9_]+]]("Hello: %d\0A\00") {addr_space = 2 : i32}
+ // CHECK: llvm.func spir_funccc @_Z6printfPU3AS2Kcz(!llvm.ptr<2>, ...) -> i32
+ // CHECK-LABEL: llvm.func spir_funccc @test_printf
+ // CHECK: (%[[ARG0:.*]]: i32)
+ gpu.func @test_printf(%arg0: i32) {
+ // CHECK: %[[IMM0:.*]] = llvm.mlir.addressof @[[$PRINT_GLOBAL]] : !llvm.ptr<2>
+ // CHECK-NEXT: %[[IMM2:.*]] = llvm.getelementptr %[[IMM0]][0, 0] : (!llvm.ptr<2>) -> !llvm.ptr<2>, !llvm.array<11 x i8>
+ // CHECK-NEXT: %{{.*}} = llvm.call spir_funccc @_Z6printfPU3AS2Kcz(%[[IMM2]], %[[ARG0]]) vararg(!llvm.func<i32 (ptr<2>, ...)>) : (!llvm.ptr<2>, i32) -> i32
+ gpu.printf "Hello: %d\n", %arg0 : i32
+ gpu.return
+ }
+}
+
diff --git a/mlir/test/Integration/Dialect/XeVM/GPU/gpu_printf.mlir b/mlir/test/Integration/Dialect/XeVM/GPU/gpu_printf.mlir
new file mode 100644
index 0000000000000..edf8775c72418
--- /dev/null
+++ b/mlir/test/Integration/Dialect/XeVM/GPU/gpu_printf.mlir
@@ -0,0 +1,30 @@
+// RUN: mlir-opt %s \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(cse,func.func(gpu-async-region),xevm-attach-target,gpu.module(convert-gpu-to-llvm-spv{use-64bit-index=true},convert-xevm-to-llvm,cse))' \
+// RUN: | mlir-opt -convert-scf-to-cf -convert-cf-to-llvm -convert-vector-to-llvm -convert-arith-to-llvm \
+// RUN: | mlir-opt -gpu-to-llvm -reconcile-unrealized-casts -cse -gpu-module-to-binary \
+// RUN: | mlir-runner \
+// RUN: --shared-libs=%mlir_sycl_runtime \
+// RUN: --shared-libs=%mlir_runner_utils \
+// RUN: --shared-libs=%mlir_c_runner_utils \
+// RUN: --entry-point-result=void \
+// RUN: | FileCheck %s
+
+module @test attributes {gpu.container_module} {
+ gpu.module @test_module {
+ gpu.func @test_printf(%arg0: i32, %arg1: f32) kernel {
+ gpu.printf "Hello: %d\n", %arg0 : i32
+ gpu.printf "Hello: %f\n", %arg1 : f32
+ gpu.return
+ }
+ }
+
+ func.func @main() attributes {llvm.emit_c_interface} {
+ %c1 = arith.constant 1 : index
+ %c11 = arith.constant 11 : i32
+ %c4 = arith.constant 4.0 : f32
+ // CHECK: Hello: 11
+ // CHECK: Hello: 4.000000
+ gpu.launch_func @test_module::@test_printf blocks in (%c1, %c1, %c1) threads in (%c1, %c1, %c1) args(%c11 : i32, %c4 : f32)
+ return
+ }
+}
|
charithaintc
approved these changes
Oct 20, 2025
fabianmcg
reviewed
Oct 21, 2025
Contributor
fabianmcg
left a comment
fabianmcg
left a comment
There was a problem hiding this comment.
Other than that, LGTM.
| /// runtimes, such as OpenCL on AMD, this is sufficient setup, as the compiler | ||
| /// will lower printf calls to appropriate device-side code | ||
| /// will lower printf calls to appropriate device-side code. | ||
| /// callingConvention and funcName can be adjusted as needed. |
Contributor
There was a problem hiding this comment.
Please expand the comment here explaining why these needs to be customizable. This means, saying callingConvention, and funcName are customizable as not all backends use the same, for example, the SPV LLVM backend uses....
Contributor
Author
There was a problem hiding this comment.
Updated comment as suggested.
…support customization.
fabianmcg
approved these changes
Oct 22, 2025
dvbuka
pushed a commit
to dvbuka/llvm-project
that referenced
this pull request
Oct 27, 2025
…lvm#164297) Existing pattern for lowering gpu.printf op to LLVM call uses fixed function name and calling convention. Those two should be exposed as pass option to allow supporting Intel Compute Runtime for GPU. Also adds gpu.printf op pattern to GPU to LLVMSPV pass. It may appear out of place, but integration test is added to XeVM integration test as that is the current best folder for testing with Intel Compute Runtime. Test should be moved in the future if a better test folder is added.
Lukacma
pushed a commit
to Lukacma/llvm-project
that referenced
this pull request
Oct 29, 2025
…lvm#164297) Existing pattern for lowering gpu.printf op to LLVM call uses fixed function name and calling convention. Those two should be exposed as pass option to allow supporting Intel Compute Runtime for GPU. Also adds gpu.printf op pattern to GPU to LLVMSPV pass. It may appear out of place, but integration test is added to XeVM integration test as that is the current best folder for testing with Intel Compute Runtime. Test should be moved in the future if a better test folder is added.
aokblast
pushed a commit
to aokblast/llvm-project
that referenced
this pull request
Oct 30, 2025
…lvm#164297) Existing pattern for lowering gpu.printf op to LLVM call uses fixed function name and calling convention. Those two should be exposed as pass option to allow supporting Intel Compute Runtime for GPU. Also adds gpu.printf op pattern to GPU to LLVMSPV pass. It may appear out of place, but integration test is added to XeVM integration test as that is the current best folder for testing with Intel Compute Runtime. Test should be moved in the future if a better test folder is added.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
Existing pattern for lowering gpu.printf op to LLVM call uses fixed function name and calling convention.
Those two should be exposed as pass option to allow supporting Intel Compute Runtime for GPU.
Also adds gpu.printf op pattern to GPU to LLVMSPV pass.
It may appear out of place, but integration test is added to XeVM integration test as that is the current best folder for testing with Intel Compute Runtime.
Test should be moved in the future if a better test folder is added.