[tuner]: add c/python binding for querying mma intrinsic (iree-org#19218

) After this PR: iree-org#19199 add Python bindings to these two utility functions to querying mma intrinsic instructions from input module. --------- Signed-off-by: Bangtian Liu <[email protected]>
nod-ai · Nov 20, 2024 · e1ce3fa · e1ce3fa
1 parent 1654ce6
commit e1ce3fa
Show file tree

Hide file tree

Showing 10 changed files with 114 additions and 2 deletions.
diff --git a/compiler/bindings/c/iree/compiler/dialects/iree_codegen.h b/compiler/bindings/c/iree/compiler/dialects/iree_codegen.h
@@ -66,6 +66,14 @@ MLIR_CAPI_EXPORTED MlirAttribute ireeCodegenCompilationInfoAttrGet(
 MLIR_CAPI_EXPORTED ireeCodegenCompilationInfoParameters
 ireeCodegenCompilationInfoAttrGetParameters(MlirAttribute attr);
 
+MLIR_CAPI_EXPORTED void
+ireeCodegenGetExecutableVariantOps(MlirModule module, size_t *numOps,
+                                   MlirOperation *executableOps);
+
+MLIR_CAPI_EXPORTED void ireeCodegenQueryMMAIntrinsics(MlirOperation op,
+                                                      size_t *numIntrinsics,
+                                                      uint32_t *mmaIntrinsics);
+
 #ifdef __cplusplus
 }
 #endif

diff --git a/compiler/bindings/python/IREECompilerDialectsModule.cpp b/compiler/bindings/python/IREECompilerDialectsModule.cpp
@@ -21,6 +21,33 @@ static const char *kGpuModuleImportPath =
 namespace py = pybind11;
 using namespace mlir::python::adaptors;
 
+static std::vector<MlirOperation>
+ireeCodegenGetExecutableVariantOpsBinding(MlirModule module) {
+  size_t numOps = 0;
+  ireeCodegenGetExecutableVariantOps(module, &numOps, nullptr);
+  std::vector<MlirOperation> ops(numOps);
+  ireeCodegenGetExecutableVariantOps(module, &numOps, ops.data());
+
+  return ops;
+}
+
+static std::vector<py::object>
+ireeCodegenQueryMMAIntrinsicsBinding(MlirOperation op) {
+  size_t numMMAs = 0;
+  ireeCodegenQueryMMAIntrinsics(op, &numMMAs, nullptr);
+  std::vector<uint32_t> mmaIntrinsics(numMMAs);
+  ireeCodegenQueryMMAIntrinsics(op, &numMMAs, mmaIntrinsics.data());
+
+  py::object mmaIntrinsicEnum =
+      py::module_::import(kGpuModuleImportPath).attr("MMAIntrinsic");
+  std::vector<py::object> mmaList(numMMAs);
+  for (size_t i = 0; i < numMMAs; ++i) {
+    mmaList[i] = mmaIntrinsicEnum(mmaIntrinsics[i]);
+  }
+
+  return mmaList;
+}
+
 PYBIND11_MODULE(_ireeCompilerDialects, m) {
   m.doc() = "iree-compiler dialects python extension";
 
@@ -326,4 +353,22 @@ PYBIND11_MODULE(_ireeCompilerDialects, m) {
           "Gets an #iree_gpu.lowering_config from parameters.")
       .def_property_readonly("attributes",
                              ireeGPULoweringConfigAttrGetAttributes);
+
+  //===-------------------------------------------------------------------===//
+  // Binding to utility function getExecutableVariantOps
+  //===-------------------------------------------------------------------===//
+
+  iree_codegen_module.def(
+      "get_executable_variant_ops", &ireeCodegenGetExecutableVariantOpsBinding,
+      "Gets the executable variant operations from a module.",
+      py::arg("module"));
+
+  //===-------------------------------------------------------------------===//
+  // Binding to utility function queryMMAIntrinsics
+  //===-------------------------------------------------------------------===//
+
+  iree_codegen_module.def(
+      "query_mma_intrinsics", &ireeCodegenQueryMMAIntrinsicsBinding,
+      "Queries the MMA intrinsics from an executable variant op.",
+      py::arg("op"));
 }
diff --git a/compiler/src/iree/compiler/API/Internal/BUILD.bazel b/compiler/src/iree/compiler/API/Internal/BUILD.bazel
@@ -137,6 +137,7 @@ iree_compiler_cc_library(
     deps = [
         "//compiler/bindings/c:headers",
         "//compiler/src/iree/compiler/Codegen/Dialect/Codegen/IR:IREECodegenDialect",
+        "//compiler/src/iree/compiler/Codegen/Utils",
         "@llvm-project//mlir:CAPIIR",
         "@llvm-project//mlir:CAPIIRHeaders",
         "@llvm-project//mlir:IR",

diff --git a/compiler/src/iree/compiler/API/Internal/CMakeLists.txt b/compiler/src/iree/compiler/API/Internal/CMakeLists.txt
@@ -116,6 +116,7 @@ iree_cc_library(
     MLIRCAPIIR
     MLIRIR
     iree::compiler::Codegen::Dialect::Codegen::IR::IREECodegenDialect
+    iree::compiler::Codegen::Utils
     iree::compiler::bindings::c::headers
   PUBLIC
 )

diff --git a/compiler/src/iree/compiler/API/Internal/IREECodegenDialectCAPI.cpp b/compiler/src/iree/compiler/API/Internal/IREECodegenDialectCAPI.cpp
@@ -10,6 +10,7 @@
 #include <type_traits>
 #include "iree/compiler/Codegen/Dialect/Codegen/IR/IREECodegenAttrs.h"
 #include "iree/compiler/Codegen/Dialect/Codegen/IR/IREECodegenInterfaces.h"
+#include "iree/compiler/Codegen/Utils/GPUUtils.h"
 #include "iree/compiler/dialects/iree_codegen.h"
 #include "mlir-c/BuiltinAttributes.h"
 #include "mlir-c/IR.h"
@@ -24,6 +25,8 @@ using mlir::iree_compiler::IREE::Codegen::DispatchLoweringPassPipeline;
 using mlir::iree_compiler::IREE::Codegen::DispatchLoweringPassPipelineAttr;
 using mlir::iree_compiler::IREE::Codegen::LoweringConfigAttrInterface;
 using mlir::iree_compiler::IREE::Codegen::TranslationInfoAttr;
+using mlir::iree_compiler::IREE::GPU::MMAIntrinsic;
+using mlir::iree_compiler::IREE::HAL::ExecutableVariantOp;
 
 bool ireeAttributeIsACodegenDispatchLoweringPassPipelineAttr(
     MlirAttribute attr) {
@@ -149,3 +152,49 @@ ireeCodegenCompilationInfoAttrGetParameters(MlirAttribute attr) {
   parameters.translationInfo = wrap(compilationInfo.getTranslationInfo());
   return parameters;
 }
+
+void ireeCodegenGetExecutableVariantOps(MlirModule module, size_t *numOps,
+                                        MlirOperation *executableOps) {
+  assert(!mlirModuleIsNull(module) && "module cannot be nullptr");
+  assert(numOps && "numOps cannot be nullptr");
+
+  mlir::ModuleOp moduleOp = unwrap(module);
+  llvm::SmallVector<ExecutableVariantOp> executableVariantOps =
+      mlir::iree_compiler::getExecutableVariantOps(moduleOp);
+
+  if (!executableOps) {
+    *numOps = executableVariantOps.size();
+    return;
+  }
+
+  assert(
+      *numOps == executableVariantOps.size() &&
+      "*numOps must match the number of elements in the executableVariantOps");
+
+  for (size_t i = 0, e = executableVariantOps.size(); i < e; ++i) {
+    executableOps[i] = wrap(executableVariantOps[i]);
+  }
+}
+
+void ireeCodegenQueryMMAIntrinsics(MlirOperation op, size_t *numIntrinsics,
+                                   uint32_t *mmaIntrinsics) {
+  assert(numIntrinsics && "numIntrinsics cannot be nullptr");
+
+  mlir::Operation *mlirOp = unwrap(op);
+  auto variantOp = llvm::dyn_cast_if_present<ExecutableVariantOp>(mlirOp);
+  assert(variantOp && "operation is not a ExecutableVariantOp");
+
+  llvm::SmallVector<MMAIntrinsic> intrinsics =
+      mlir::iree_compiler::queryMMAIntrinsics(variantOp);
+  if (!mmaIntrinsics) {
+    *numIntrinsics = intrinsics.size();
+    return;
+  }
+
+  assert(*numIntrinsics == intrinsics.size() &&
+         "*numIntrinsics must match the number of elements in the intrinsics");
+
+  for (size_t i = 0, e = intrinsics.size(); i < e; ++i) {
+    mmaIntrinsics[i] = static_cast<uint32_t>(intrinsics[i]);
+  }
+}
diff --git a/compiler/src/iree/compiler/API/api_exports.c b/compiler/src/iree/compiler/API/api_exports.c
@@ -24,6 +24,8 @@ extern void ireeCodegenCompilationInfoAttrGetTypeID();
 extern void ireeCodegenDispatchLoweringPassPipelineAttrGet();
 extern void ireeCodegenDispatchLoweringPassPipelineAttrGetTypeID();
 extern void ireeCodegenDispatchLoweringPassPipelineAttrGetValue();
+extern void ireeCodegenGetExecutableVariantOps();
+extern void ireeCodegenQueryMMAIntrinsics();
 extern void ireeCodegenTranslationInfoAttrGet();
 extern void ireeCodegenTranslationInfoAttrGetParameters();
 extern void ireeCodegenTranslationInfoAttrGetTypeID();

diff --git a/compiler/src/iree/compiler/API/api_exports.def b/compiler/src/iree/compiler/API/api_exports.def
@@ -14,6 +14,8 @@ EXPORTS
   ireeCodegenDispatchLoweringPassPipelineAttrGet
   ireeCodegenDispatchLoweringPassPipelineAttrGetTypeID
   ireeCodegenDispatchLoweringPassPipelineAttrGetValue
+  ireeCodegenGetExecutableVariantOps
+  ireeCodegenQueryMMAIntrinsics
   ireeCodegenTranslationInfoAttrGet
   ireeCodegenTranslationInfoAttrGetParameters
   ireeCodegenTranslationInfoAttrGetTypeID

diff --git a/compiler/src/iree/compiler/API/api_exports.ld b/compiler/src/iree/compiler/API/api_exports.ld
@@ -15,6 +15,8 @@ VER_0 {
     ireeCodegenDispatchLoweringPassPipelineAttrGet;
     ireeCodegenDispatchLoweringPassPipelineAttrGetTypeID;
     ireeCodegenDispatchLoweringPassPipelineAttrGetValue;
+    ireeCodegenGetExecutableVariantOps;
+    ireeCodegenQueryMMAIntrinsics;
     ireeCodegenTranslationInfoAttrGet;
     ireeCodegenTranslationInfoAttrGetParameters;
     ireeCodegenTranslationInfoAttrGetTypeID;

diff --git a/compiler/src/iree/compiler/API/api_exports.macos.lst b/compiler/src/iree/compiler/API/api_exports.macos.lst
@@ -13,6 +13,8 @@ _ireeCodegenCompilationInfoAttrGetTypeID
 _ireeCodegenDispatchLoweringPassPipelineAttrGet
 _ireeCodegenDispatchLoweringPassPipelineAttrGetTypeID
 _ireeCodegenDispatchLoweringPassPipelineAttrGetValue
+_ireeCodegenGetExecutableVariantOps
+_ireeCodegenQueryMMAIntrinsics
 _ireeCodegenTranslationInfoAttrGet
 _ireeCodegenTranslationInfoAttrGetParameters
 _ireeCodegenTranslationInfoAttrGetTypeID

diff --git a/compiler/src/iree/compiler/Codegen/Utils/GPUUtils.cpp b/compiler/src/iree/compiler/Codegen/Utils/GPUUtils.cpp
@@ -1030,7 +1030,7 @@ std::optional<int> getGPUSubgroupSize(mlir::FunctionOpInterface func) {
 
 SmallVector<IREE::HAL::ExecutableVariantOp>
 getExecutableVariantOps(mlir::ModuleOp moduleOp) {
-  llvm::SmallVector<IREE::HAL::ExecutableVariantOp> executableVariantOps;
+  SmallVector<IREE::HAL::ExecutableVariantOp> executableVariantOps;
   moduleOp.walk([&](IREE::HAL::ExecutableVariantOp executableOp) {
     executableVariantOps.push_back(executableOp);
   });
@@ -1039,7 +1039,7 @@ getExecutableVariantOps(mlir::ModuleOp moduleOp) {
 
 SmallVector<IREE::GPU::MMAIntrinsic>
 queryMMAIntrinsics(IREE::HAL::ExecutableVariantOp executableOp) {
-  llvm::SmallVector<IREE::GPU::MMAIntrinsic> mmaIntrinsics;
+  SmallVector<IREE::GPU::MMAIntrinsic> mmaIntrinsics;
   if (IREE::GPU::TargetAttr target = getGPUTargetAttr(executableOp)) {
     mmaIntrinsics = llvm::map_to_vector(
         target.getWgp().getMma(),