update 3.8 v2 (#2112)
* update 3.8 v2 * update 3.8 --------- Co-authored-by: yuzhai <yuzhai@nvidia.com>
This commit is contained in:
Yujia Zhai
+60
-47
@@ -150,14 +150,14 @@ set(CUTLASS_ENABLE_PERFORMANCE ${CUTLASS_ENABLE_PROFILER} CACHE BOOL "Enable CUT
|
||||
set(CUTLASS_ENABLE_TESTS ${CUTLASS_ENABLE_TESTS_INIT} CACHE BOOL "Enable CUTLASS Tests")
|
||||
set(CUTLASS_ENABLE_GTEST_UNIT_TESTS ${CUTLASS_ENABLE_TESTS} CACHE BOOL "Enable CUTLASS GTest-based Unit Tests")
|
||||
set(CUTLASS_USE_SYSTEM_GOOGLETEST OFF CACHE BOOL "Use system/external installation of GTest")
|
||||
set(CUTLASS_USE_PACKED_TUPLE ON CACHE BOOL "If ON, make cute::tuple be new standard-layout tuple type; if OFF, use the original cute::tuple implementation that is _not_ standard-layout.")
|
||||
if (CUTLASS_USE_PACKED_TUPLE)
|
||||
list(APPEND CUTLASS_CUDA_NVCC_FLAGS -DCUTE_USE_PACKED_TUPLE=1)
|
||||
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DCUTLASS_USE_PACKED_TUPLE=1")
|
||||
message(STATUS "Make cute::tuple be the new standard-layout tuple type")
|
||||
elseif()
|
||||
message(STATUS "Use the original cute::tuple implementation that is _not_ standard-layout")
|
||||
|
||||
if (CUTLASS_ENABLE_TESTS AND CUTLASS_ENABLE_PROFILER)
|
||||
set(CUTLASS_ENABLE_PROFILER_UNIT_TESTS_INIT ON)
|
||||
else()
|
||||
set(CUTLASS_ENABLE_PROFILER_UNIT_TESTS_INIT OFF)
|
||||
endif()
|
||||
set(CUTLASS_ENABLE_PROFILER_UNIT_TESTS ${CUTLASS_ENABLE_PROFILER_UNIT_TESTS_INIT} CACHE BOOL "Enable CUTLASS Profiler-based Unit Tests")
|
||||
set(CUTLASS_ENABLE_SELF_CONTAINED_INCLUDES_CHECK ON CACHE BOOL "Enable CUTLASS check for self-contained header includes")
|
||||
|
||||
################################################################################
|
||||
|
||||
@@ -406,7 +406,7 @@ endif()
|
||||
|
||||
# Warnings-as-error exceptions and warning suppressions for Clang builds
|
||||
if (CUTLASS_CLANG_HOST_COMPILE)
|
||||
|
||||
|
||||
set(FLAGS_TO_ADD
|
||||
"-Wno-error=implicit-int-conversion"
|
||||
"-Wno-error=pass-failed"
|
||||
@@ -414,13 +414,13 @@ if (CUTLASS_CLANG_HOST_COMPILE)
|
||||
"-Wno-sign-conversion"
|
||||
"-Wno-unused-parameter"
|
||||
)
|
||||
|
||||
|
||||
foreach(FLAG ${FLAGS_TO_ADD})
|
||||
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${FLAG}")
|
||||
list(APPEND CUTLASS_CUDA_NVCC_FLAGS "${FLAG}")
|
||||
list(APPEND CUTLASS_CUDA_CLANG_FLAGS "${FLAG}")
|
||||
endforeach()
|
||||
|
||||
|
||||
endif()
|
||||
|
||||
if (NOT MSVC AND CUTLASS_NVCC_KEEP)
|
||||
@@ -486,7 +486,7 @@ if (CUTLASS_CLANG_DEVICE_COMPILE)
|
||||
|
||||
link_libraries(nvidia::cudart)
|
||||
link_libraries(nvidia::cuda_driver)
|
||||
|
||||
|
||||
endif()
|
||||
|
||||
#Report CUDA build flags
|
||||
@@ -561,7 +561,7 @@ function(cutlass_apply_cuda_gencode_flags TARGET)
|
||||
list(APPEND __CMAKE_CUDA_ARCHS ${ARCH}-real)
|
||||
endif()
|
||||
if(CUTLASS_NVCC_EMBED_PTX AND NOT CUTLASS_CLANG_DEVICE_COMPILE)
|
||||
# If we're using clang for device compilation, the ptx is inserted
|
||||
# If we're using clang for device compilation, the ptx is inserted
|
||||
# via another command line option and the `-virtual` flags will cause an error.
|
||||
list(APPEND __CMAKE_CUDA_ARCHS ${ARCH}-virtual)
|
||||
endif()
|
||||
@@ -922,7 +922,7 @@ function(cutlass_add_executable_tests NAME TARGET)
|
||||
if (NOT __DO_NOT_LOWERCASE_TEST_NAME)
|
||||
string(TOLOWER "${TESTCASE_NAME}" TESTCASE_NAME)
|
||||
endif()
|
||||
|
||||
|
||||
# The following rigmarole is needed to deal with spaces and possible quotes in
|
||||
# command line arguments. The options are passed "by reference" as the actual
|
||||
# variable names holding the real options. We then expand these in a way that
|
||||
@@ -1007,46 +1007,51 @@ function(cutlass_generate_profiler_tests NAME)
|
||||
endif()
|
||||
|
||||
file(STRINGS ${CUTLASS_PROFILER_REGRESSION_LIST_FILE} TEST_LIST)
|
||||
|
||||
foreach(TEST IN LISTS TEST_LIST)
|
||||
|
||||
set(TEMP_TEST ${TEST})
|
||||
if ("${TEST}" MATCHES " *cutlass_profiler.*")
|
||||
|
||||
# Generate a flattened name for the test from the test command line.
|
||||
string(REPLACE "," ";" TEST_NAME_LIST ${TEST})
|
||||
list(GET TEST_NAME_LIST 0 TEST)
|
||||
string(REGEX MATCHALL "[a-zA-Z0-9_=]+" TEST_NAME "${TEST}")
|
||||
list(FILTER TEST_NAME EXCLUDE REGEX "cutlass_profiler|mode=trace|providers=cutlass")
|
||||
list(JOIN TEST_NAME "_" TEST_NAME)
|
||||
string(REGEX REPLACE "_verification_required=(true|false)" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "_verification_providers=device" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "batch_count=" "batch" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "cluster_m=" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "_cluster_n=" "x" TEST_NAME "${TEST_NAME}")
|
||||
string(REGEX REPLACE "_cluster_k=[0-9]+" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "cluster_m_fallback=" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "_cluster_n_fallback=" "x" TEST_NAME "${TEST_NAME}")
|
||||
string(REGEX REPLACE "_cluster_k_fallback=[0-9]+" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "runtime_input_datatype_a=" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "runtime_input_datatype_b=" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "=" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "_error_on_no_match" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "_error_if_nothing_is_profiled" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "kernels" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "operation" "" TEST_NAME "${TEST_NAME}")
|
||||
# Generate a flattened name for the test from the test command line.
|
||||
string(REPLACE "," ";" TEST_NAME_LIST ${TEMP_TEST})
|
||||
string(REGEX REPLACE "\\*" "_" TEST_NAME "${TEMP_TEST}")
|
||||
string(REGEX REPLACE "\\\"\\{\\\"\\\"input_params.*\\{.*\\}\\}\\\"" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REGEX REPLACE "\\\"\\{\\\"\\\"input_params.*\\{.*\\}\\}\\\"" "" TEST "${TEST}")
|
||||
string(REGEX REPLACE "," ";" TEST "${TEST}")
|
||||
string(REGEX MATCHALL "[a-zA-Z0-9_=]+" TEST_NAME "${TEST_NAME}")
|
||||
list(FILTER TEST_NAME EXCLUDE REGEX "cutlass_profiler|mode=trace|providers=cutlass")
|
||||
list(JOIN TEST_NAME "_" TEST_NAME)
|
||||
string(REGEX REPLACE "_verification_required=(true|false)" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "_verification_providers=device" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "batch_count=" "batch" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "cluster_m=" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "_cluster_n=" "x" TEST_NAME "${TEST_NAME}")
|
||||
string(REGEX REPLACE "_cluster_k=[0-9]+" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "cluster_m_fallback=" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "_cluster_n_fallback=" "x" TEST_NAME "${TEST_NAME}")
|
||||
string(REGEX REPLACE "_cluster_k_fallback=[0-9]+" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "runtime_input_datatype_a=" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "runtime_input_datatype_b=" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REGEX REPLACE "verification_enabled=(true|false)" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REGEX REPLACE "warmup_iterations=[0-9]+" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REGEX REPLACE "profiling_iterations=[0-9]+" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REGEX REPLACE "sleep_duration=[0-9]+" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REGEX REPLACE "profiling_enabled=(true|false)" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "=" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "_error_on_no_match" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "_error_if_nothing_is_profiled" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "kernels" "" TEST_NAME "${TEST_NAME}")
|
||||
string(REPLACE "operation" "" TEST_NAME "${TEST_NAME}")
|
||||
|
||||
if (__DO_NOT_LOWERCASE_TEST_NAME)
|
||||
string(TEST_NAME_LOWER "${TEST_NAME}")
|
||||
else()
|
||||
string(TOLOWER "${TEST_NAME}" TEST_NAME_LOWER)
|
||||
endif()
|
||||
if (NOT __DO_NOT_LOWERCASE_TEST_NAME)
|
||||
string(TOLOWER "${TEST_NAME}" TEST_NAME)
|
||||
endif()
|
||||
|
||||
# Munge the test command
|
||||
string(REPLACE "cutlass_profiler" "" TEST "${TEST}")
|
||||
set(TEST "${TEST}" ${__CUTLASS_PROFILER_EXTRA_OPTIONS} "--junit-output=${TEST_NAME_LOWER}")
|
||||
set(TEST_COMMAND_${TEST_NAME_LOWER} "${TEST}")
|
||||
list(APPEND TEST_COMMAND_VARS ${TEST_NAME_LOWER})
|
||||
# Munge the test command
|
||||
|
||||
string(REPLACE "cutlass_profiler" "" TEST "${TEST}")
|
||||
set(TEST "${TEST}" ${__CUTLASS_PROFILER_EXTRA_OPTIONS} "--junit-output=${TEST_NAME}")
|
||||
set(TEST_COMMAND_${TEST_NAME} "${TEST}")
|
||||
list(APPEND TEST_COMMAND_VARS ${TEST_NAME})
|
||||
endif()
|
||||
|
||||
endforeach()
|
||||
@@ -1084,6 +1089,14 @@ if (CUTLASS_ENABLE_TESTS)
|
||||
if (CUTLASS_ENABLE_GTEST_UNIT_TESTS)
|
||||
add_dependencies(test_all test_unit)
|
||||
endif()
|
||||
if (CUTLASS_ENABLE_PROFILER_UNIT_TESTS AND CUTLASS_BUILD_FOR_PROFILER_REGRESSIONS)
|
||||
# Generate profiler based unit test
|
||||
cutlass_generate_profiler_tests(
|
||||
tup
|
||||
DEPENDEES test_unit
|
||||
)
|
||||
endif()
|
||||
|
||||
endif()
|
||||
|
||||
if (CUTLASS_INSTALL_TESTS)
|
||||
|
||||
Reference in New Issue
Block a user