[RUNTIME] Major code cleanup (#711)

This PR does the following:
- CUDA utilities (e.g., cuGetInfo) won't be compiled as part of libtriton.so anymore.
- Refactoring driver/llvm.cc to split it between PTX codegen and python.
- By extension this will also deprecate include/external so Triton won't have to live with a copy of some CUDA/Hip headers anymore.
- `triton-translate` becomes a `triton.tools.aot` Python utility that re-uses functions from the triton.compile sub-module.
This commit is contained in:
Philippe Tillet
2022-09-26 16:38:06 -07:00
committed by GitHub
parent 8bb09f83ee
commit 1e91ed30d0
28 changed files with 509 additions and 31483 deletions

View File

@@ -26,35 +26,35 @@ target_link_libraries(triton-opt PRIVATE
mlir_check_all_link_libraries(triton-opt)
add_llvm_executable(triton-translate triton-translate.cpp PARTIAL_SOURCES_INTENDED)
llvm_update_compile_flags(triton-translate)
target_link_libraries(triton-translate PRIVATE
TritonAnalysis
TritonTransforms
TritonGPUTransforms
TritonLLVMIR
TritonDriver
${dialect_libs}
${conversion_libs}
# tests
TritonTestAnalysis
# add_llvm_executable(triton-translate triton-translate.cpp PARTIAL_SOURCES_INTENDED)
#llvm_update_compile_flags(triton-translate)
# target_link_libraries(triton-translate PRIVATE
# TritonAnalysis
# TritonTransforms
# TritonGPUTransforms
# TritonLLVMIR
# TritonDriver
# ${dialect_libs}
# ${conversion_libs}
# # tests
# TritonTestAnalysis
LLVMCore
LLVMSupport
LLVMOption
LLVMCodeGen
LLVMAsmParser
# LLVMCore
# LLVMSupport
# LLVMOption
# LLVMCodeGen
# LLVMAsmParser
# MLIR core
MLIROptLib
MLIRIR
MLIRPass
MLIRSupport
MLIRTransforms
MLIRExecutionEngine
MLIRMathToLLVM
MLIRTransformUtils
MLIRLLVMToLLVMIRTranslation
MLIRNVVMToLLVMIRTranslation
)
mlir_check_all_link_libraries(triton-translate)
# # MLIR core
# MLIROptLib
# MLIRIR
# MLIRPass
# MLIRSupport
# MLIRTransforms
# MLIRExecutionEngine
# MLIRMathToLLVM
# MLIRTransformUtils
# MLIRLLVMToLLVMIRTranslation
# MLIRNVVMToLLVMIRTranslation
# )
# mlir_check_all_link_libraries(triton-translate)