Skip to content
This repository was archived by the owner on May 9, 2024. It is now read-only.

Add GpuSharedMemory Tests for Intel GPU #537

Open
wants to merge 8 commits into
base: main
Choose a base branch
from
Open

Add GpuSharedMemory Tests for Intel GPU #537

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
2a4bd6e
Add GpuSharedMemory Tests for Intel GPU
lmontigny Jun 21, 2023
abf9f73
Fix style-check header file
lmontigny Jul 4, 2023
140e1d7
Fix style-check backend.h
lmontigny Jul 5, 2023
c822119
Cleaning smem init
lmontigny Jul 5, 2023
0e564de
Reducing GpuSharedMemory tests
lmontigny Jul 10, 2023
d50f030
Cleaning up GpuSharedMemoryTestIntel
lmontigny Jul 20, 2023
88ce6d6
Adding LLVM metadata
lmontigny Jul 21, 2023
9e4ef60
Fixed all the logic to the SM codegen
lmontigny Jul 21, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions omniscidb/QueryEngine/Compiler/Backend.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -234,4 +234,6 @@ void setSharedMemory(ExecutorDeviceType dt,
GPUTarget& gpu_target,
const std::shared_ptr<compiler::Backend>& backend);

void replace_function(llvm::Module* from, llvm::Module* to, const std::string& fname);

} // namespace compiler
9 changes: 7 additions & 2 deletions omniscidb/QueryEngine/GpuSharedMemoryUtils.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -242,10 +242,12 @@ llvm::Value* codegen_smem_dest_slot_ptr(llvm::LLVMContext& context,
auto ptr_type = [&context, &traits](const size_t slot_bytes,
const hdk::ir::Type* type) {
if (slot_bytes == sizeof(int32_t)) {
return traits.smemPointerType(llvm::Type::getInt32Ty(context));
// return traits.smemPointerType(llvm::Type::getInt32Ty(context)); // quickfix
return llvm::Type::getInt32PtrTy(context, /*address_space=*/3);
} else {
CHECK(slot_bytes == sizeof(int64_t));
return traits.smemPointerType(llvm::Type::getInt64Ty(context));
// return traits.smemPointerType(llvm::Type::getInt64Ty(context)); // quickfix
return llvm::Type::getInt64PtrTy(context, /*address_space=*/3);
}
UNREACHABLE() << "Invalid slot size encountered: " << std::to_string(slot_bytes);
return traits.smemPointerType(llvm::Type::getInt32Ty(context));
Expand Down Expand Up @@ -327,6 +329,9 @@ void GpuSharedMemCodeBuilder::codegenInitialization() {
dest_byte_stream,
byte_offset_ll);

// DUMP(traits_, "output_traits_")
// std::cout << "traits_" << traits_;

llvm::Value* init_value_ll = nullptr;
if (slot_size == sizeof(int32_t)) {
init_value_ll =
Expand Down
13 changes: 13 additions & 0 deletions omniscidb/Tests/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,6 +60,7 @@ endif()
add_executable(StringDictionaryBenchmark StringDictionaryBenchmark.cpp)

if(ENABLE_L0)
add_executable(GpuSharedMemoryTestIntel GpuSharedMemoryTestIntel.cpp ResultSetTestUtils.cpp)
add_executable(L0MgrExecuteTest L0MgrExecuteTest.cpp)
add_executable(SpirvBuildTest SpirvBuildTest.cpp)
add_executable(DataMgrWithL0Test DataMgrWithL0Test.cpp)
Expand All @@ -69,11 +70,13 @@ if(ENABLE_L0)
target_link_libraries(SpirvBuildTest gtest ${llvm_libs})
target_link_libraries(DataMgrWithL0Test DataMgr gtest)
target_link_libraries(IntelGPUEnablingTest gtest QueryEngine ArrowQueryRunner)
target_link_libraries(GpuSharedMemoryTestIntel gtest QueryEngine ArrowQueryRunner)

add_test(L0MgrExecuteTest L0MgrExecuteTest ${TEST_ARGS})
add_test(SpirvBuildTest SpirvBuildTest ${TEST_ARGS})
add_test(DataMgrWithL0Test DataMgrWithL0Test ${TEST_ARGS})
add_test(IntelGPUEnablingTest IntelGPUEnablingTest ${TEST_ARGS})
add_test(GpuSharedMemoryTestIntel GpuSharedMemoryTestIntel ${TEST_ARGS})
endif()

add_executable(CostModelTest CostModel/CostModelTest.cpp)
Expand Down Expand Up @@ -134,6 +137,10 @@ else()
target_link_libraries(StringDictionaryBenchmark benchmark gtest StringDictionary Logger Utils $<$<AND:$<CXX_COMPILER_ID:GNU>,$<VERSION_LESS:$<CXX_COMPILER_VERSION>,9.0>>:stdc++fs> ${CMAKE_DL_LIBS} ${Boost_LIBRARIES} ${ZLIB_LIBRARIES})
endif()

if(ENABLE_L0)
target_link_libraries(GpuSharedMemoryTestIntel gtest Logger QueryEngine)
endif()

if(ENABLE_CUDA)
target_link_libraries(GpuSharedMemoryTest gtest Logger QueryEngine)
endif()
Expand Down Expand Up @@ -180,6 +187,7 @@ if(ENABLE_CUDA)
add_test(GpuSharedMemoryTest GpuSharedMemoryTest ${TEST_ARGS})
endif()
if(ENABLE_L0)
add_test(GpuSharedMemoryTestIntel GpuSharedMemoryTestIntel ${TEST_ARGS})
add_test(NAME PuntToCpu COMMAND ArrowBasedExecuteTest "--gtest_filter=Select.Punt*" ${TEST_ARGS})
set_tests_properties(PuntToCpu PROPERTIES LABELS "enabling")
add_test(NAME StreamingTopNFallback COMMAND ArrowBasedExecuteTest "--gtest_filter=Select.TopKHeap:Select.TimeInterval:Select.OverflowAndUnderFlow" ${TEST_ARGS})
Expand Down Expand Up @@ -236,6 +244,10 @@ if(ENABLE_CUDA)
list(APPEND TEST_PROGRAMS GpuSharedMemoryTest)
endif()

if(ENABLE_L0)
list(APPEND TEST_PROGRAMS GpuSharedMemoryTestIntel)
endif()

#if(NOT ${CMAKE_SYSTEM_NAME} STREQUAL "Darwin")
# list(APPEND TEST_PROGRAMS UdfTest)
#endif()
Expand Down Expand Up @@ -323,6 +335,7 @@ add_custom_target(topk_tests

if(ENABLE_L0)
set(ENABLING_TESTS
GpuSharedMemoryTestIntel
ArrowStorageSqlTest # taxi queries
SpirvBuildTest
L0MgrExecuteTest
Expand Down
Loading