Initial PR for a linalg example through level_zero (#129)

Author: nbpatel

CMakeLists.txt

@@ -52,11 +52,13 @@ endif()
 option(DPNP_ENABLE "Use DPNP for some math functions" OFF)
 option(GPU_ENABLE "Enable GPU codegen" OFF)
 option(NUMBA_ENABLE "Enable numba-based python frontend" ON)
+option(TBB_ENABLE "Enable TBB" ON)
 
 message(STATUS "DPNP_ENABLE ${DPNP_ENABLE}")
 message(STATUS "GPU_ENABLE ${GPU_ENABLE}")
 message(STATUS "BUILD_TESTING ${BUILD_TESTING}")
 message(STATUS "NUMBA_ENABLE ${NUMBA_ENABLE}")
+message(STATUS "TBB_ENABLE ${TBB_ENABLE}")
 
 include(CTest)
 
@@ -73,9 +75,9 @@ if(GPU_ENABLE)
 endif()
 
 add_subdirectory(mlir)
+add_subdirectory(dpcomp_runtime)
 
 if(NUMBA_ENABLE)
-    add_subdirectory(dpcomp_runtime)
     add_subdirectory(tools)
     add_subdirectory(numba_dpcomp)
 endif()

dpcomp_gpu_runtime/lib/kernel_api_stubs.cpp

@@ -15,6 +15,8 @@
 #include <cstdint>
 #include <cstdio>
 #include <cstdlib>
+#include <mutex>
+#include <numeric>
 
 #include "dpcomp-gpu-runtime_export.h"
 

dpcomp_runtime/CMakeLists.txt

@@ -16,16 +16,27 @@ project(dpcomp-runtime LANGUAGES CXX C)
 
 include(GenerateExportHeader)
 
-find_package(TBB REQUIRED)
+if(TBB_ENABLE)
+    find_package(TBB REQUIRED)
+endif()
+
 find_package(MLIR REQUIRED CONFIG)
 
-set(SOURCES_LIST
-    lib/context.cpp
-    lib/tbb_parallel.cpp
-    lib/memory.cpp
-    )
-set(HEADERS_LIST
-    )
+if(TBB_ENABLE)
+    set(SOURCES_LIST
+        lib/context.cpp
+        lib/tbb_parallel.cpp
+        lib/memory.cpp
+        )
+    set(HEADERS_LIST
+        )
+else()
+    set(SOURCES_LIST
+        lib/memory.cpp
+        )   
+    set(HEADERS_LIST
+        )        
+endif()
 
 add_library(${PROJECT_NAME} SHARED ${SOURCES_LIST} ${HEADERS_LIST})
 generate_export_header(${PROJECT_NAME})
@@ -35,5 +46,6 @@ target_include_directories(${PROJECT_NAME} SYSTEM PRIVATE
     PRIVATE
     ${PROJECT_BINARY_DIR}
     )
-
-target_link_libraries(${PROJECT_NAME} TBB::tbb)
+if(TBB_ENABLE)
+    target_link_libraries(${PROJECT_NAME} TBB::tbb)
+endif()

dpcomp_runtime/lib/memory.cpp

@@ -25,6 +25,21 @@
 
 #include "dpcomp-runtime_export.h"
 
+template <typename T, int N> struct MemRefDescriptor {
+  T *allocated;
+  T *aligned;
+  int64_t offset;
+  int64_t sizes[N];
+  int64_t strides[N];
+};
+
+/// Fills the given 1D float memref with the given float value.
+extern "C" DPCOMP_RUNTIME_EXPORT void
+_mlir_ciface_fillResource1DFloat(MemRefDescriptor<float, 1> *ptr, // NOLINT
+                                 float value) {
+  std::fill_n(ptr->allocated, ptr->sizes[0], value);
+}
+
 extern "C" DPCOMP_RUNTIME_EXPORT void
 memrefCopy(int64_t elemSize, UnrankedMemRefType<char> *srcArg,
            UnrankedMemRefType<char> *dstArg) {

mlir/CMakeLists.txt

@@ -39,13 +39,21 @@ endif()
 
 add_subdirectory(include/mlir-extensions/dialect/plier)
 add_subdirectory(include/mlir-extensions/dialect/plier_util)
+add_subdirectory(include/mlir-extensions/dialect/gpu_runtime/IR)
+
+if (GPU_ENABLE)
+    add_subdirectory(tools)
+endif()
 
 set(SOURCES_LIST
     lib/Conversion/SCFToAffine/SCFToAffine.cpp
+    lib/Conversion/gpu_to_gpu_runtime.cpp
+    lib/Conversion/gpu_runtime_to_llvm.cpp
     lib/analysis/memory_ssa.cpp
     lib/analysis/memory_ssa_analysis.cpp
     lib/compiler/compiler.cpp
     lib/compiler/pipeline_registry.cpp
+    lib/dialect/gpu_runtime/IR/gpu_runtime_ops.cpp
     lib/dialect/plier/dialect.cpp
     lib/dialect/plier_util/dialect.cpp
     lib/transforms/arg_lowering.cpp
@@ -72,10 +80,13 @@ set(SOURCES_LIST
     )
 set(HEADERS_LIST
     include/mlir-extensions/Conversion/SCFToAffine/SCFToAffine.h
+    include/mlir-extensions/Conversion/gpu_to_gpu_runtime.hpp
+    include/mlir-extensions/Conversion/gpu_runtime_to_llvm.hpp
     include/mlir-extensions/analysis/memory_ssa.hpp
     include/mlir-extensions/analysis/memory_ssa_analysis.hpp
     include/mlir-extensions/compiler/compiler.hpp
     include/mlir-extensions/compiler/pipeline_registry.hpp
+    include/mlir-extensions/dialect/gpu_runtime/IR/gpu_runtime_ops.hpp
     include/mlir-extensions/dialect/plier/dialect.hpp
     include/mlir-extensions/dialect/plier_util/dialect.hpp
     include/mlir-extensions/transforms/arg_lowering.hpp
@@ -140,4 +151,4 @@ target_include_directories(${MLIR_EXTENSIONS_LIB} PUBLIC
     ${PROJECT_BINARY_DIR}/mlir-extensions/include
     )
 
-add_dependencies(${MLIR_EXTENSIONS_LIB} MLIRPlierOpsIncGen MLIRPlierUtilOpsIncGen)
+add_dependencies(${MLIR_EXTENSIONS_LIB} MLIRPlierOpsIncGen MLIRPlierUtilOpsIncGen MLIRGpuRuntimeOpsIncGen)

mlir/include/mlir-extensions/Conversion/gpu_runtime_to_llvm.hpp

@@ -0,0 +1,36 @@
+// Copyright 2022 Intel Corporation
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "mlir-extensions/dialect/gpu_runtime/IR/gpu_runtime_ops.hpp"
+#include "mlir-extensions/transforms/func_utils.hpp"
+
+#include <mlir/Conversion/AsyncToLLVM/AsyncToLLVM.h>
+#include <mlir/Conversion/GPUCommon/GPUCommonPass.h>
+#include <mlir/Conversion/LLVMCommon/ConversionTarget.h>
+#include <mlir/Conversion/LLVMCommon/Pattern.h>
+#include <mlir/Conversion/LLVMCommon/TypeConverter.h>
+#include <mlir/Dialect/GPU/Passes.h>
+#include <mlir/Dialect/LLVMIR/LLVMDialect.h>
+#include <mlir/Pass/PassManager.h>
+#include <mlir/Transforms/DialectConversion.h>
+#include <mlir/Transforms/Passes.h>
+
+namespace gpu_runtime {
+
+std::unique_ptr<mlir::Pass> createEnumerateEventsPass();
+std::unique_ptr<mlir::Pass> createGPUToLLVMPass();
+
+} // namespace gpu_runtime

mlir/include/mlir-extensions/Conversion/gpu_to_gpu_runtime.hpp

@@ -0,0 +1,63 @@
+// Copyright 2022 Intel Corporation
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "mlir-extensions/dialect/gpu_runtime/IR/gpu_runtime_ops.hpp"
+
+#include "mlir/Conversion/ControlFlowToSPIRV/ControlFlowToSPIRV.h"
+#include <mlir/Analysis/BufferViewFlowAnalysis.h>
+#include <mlir/Conversion/AffineToStandard/AffineToStandard.h>
+#include <mlir/Conversion/ArithmeticToSPIRV/ArithmeticToSPIRV.h>
+#include <mlir/Conversion/GPUCommon/GPUCommonPass.h>
+#include <mlir/Conversion/GPUToSPIRV/GPUToSPIRV.h>
+#include <mlir/Conversion/GPUToSPIRV/GPUToSPIRVPass.h>
+#include <mlir/Conversion/MathToSPIRV/MathToSPIRV.h>
+#include <mlir/Conversion/SCFToSPIRV/SCFToSPIRV.h>
+#include <mlir/Conversion/StandardToSPIRV/StandardToSPIRV.h>
+#include <mlir/Dialect/Affine/IR/AffineOps.h>
+#include <mlir/Dialect/Arithmetic/Transforms/Passes.h>
+#include <mlir/Dialect/GPU/ParallelLoopMapper.h>
+#include <mlir/Dialect/GPU/Passes.h>
+#include <mlir/Dialect/GPU/Utils.h>
+#include <mlir/Dialect/LLVMIR/LLVMDialect.h>
+#include <mlir/Dialect/Math/IR/Math.h>
+#include <mlir/Dialect/MemRef/IR/MemRef.h>
+#include <mlir/Dialect/SCF/SCF.h>
+#include <mlir/Dialect/SPIRV/IR/SPIRVDialect.h>
+#include <mlir/Dialect/SPIRV/IR/SPIRVOps.h>
+#include <mlir/Dialect/SPIRV/IR/TargetAndABI.h>
+#include <mlir/Dialect/SPIRV/Transforms/Passes.h>
+#include <mlir/Dialect/SPIRV/Transforms/SPIRVConversion.h>
+#include <mlir/Dialect/StandardOps/IR/Ops.h>
+#include <mlir/Pass/PassManager.h>
+#include <mlir/Target/SPIRV/Serialization.h>
+#include <mlir/Transforms/DialectConversion.h>
+#include <mlir/Transforms/GreedyPatternRewriteDriver.h>
+#include <mlir/Transforms/Passes.h>
+#include <llvm/ADT/SmallBitVector.h>
+
+
+namespace gpu_runtime {
+
+std::unique_ptr<mlir::Pass> createAbiAttrsPass();
+std::unique_ptr<mlir::Pass> createSetSPIRVCapabilitiesPass();
+std::unique_ptr<mlir::Pass> createGPUToSpirvPass();
+std::unique_ptr<mlir::Pass> createInsertGPUAllocsPass();
+std::unique_ptr<mlir::Pass> createUnstrideMemrefsPass();
+std::unique_ptr<mlir::Pass> createSerializeSPIRVPass();
+std::unique_ptr<mlir::Pass> createGPUExPass();
+std::unique_ptr<mlir::Pass> createParallelLoopGPUMappingPass();
+
+} // namespace gpu_runtime

mlir/include/mlir-extensions/dialect/gpu_runtime/IR/CMakeLists.txt

@@ -0,0 +1,13 @@
+include_directories(${MLIR_INCLUDE_DIRS})
+set(dialect GpuRuntimeOps)
+set(dialect_namespace gpu_runtime)
+set(LLVM_TARGET_DEFINITIONS ${dialect}.td)
+mlir_tablegen(${dialect}Enums.h.inc -gen-enum-decls)
+mlir_tablegen(${dialect}Enums.cpp.inc -gen-enum-defs)
+mlir_tablegen(${dialect}.h.inc -gen-op-decls)
+mlir_tablegen(${dialect}.cpp.inc -gen-op-defs)
+mlir_tablegen(${dialect}Dialect.h.inc -gen-dialect-decls -dialect=${dialect_namespace})
+mlir_tablegen(${dialect}Dialect.cpp.inc -gen-dialect-defs -dialect=${dialect_namespace})
+add_public_tablegen_target(MLIR${dialect}IncGen)
+add_dependencies(mlir-headers MLIR${dialect}IncGen)
+