New updates for 2.11 (#775)

* New updates.

* Minor profiler updates

Co-authored-by: Aniket Shivam <[email protected]>

CHANGELOG.md

@@ -285,7 +285,7 @@
 
 ## Copyright
 
-Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 SPDX-License-Identifier: BSD-3-Clause
 
 ```

CMakeLists.txt

@@ -1,4 +1,4 @@
-# Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: BSD-3-Clause
 #
 # Redistribution and use in source and binary forms, with or without
@@ -37,6 +37,7 @@ else()
 endif()
 
 message(STATUS "CMake Version: ${CMAKE_VERSION}")
+set(IMPLICIT_CMAKE_CXX_STANDARD OFF CACHE BOOL "Do not explicitly specify -std=c++11 if set")
 
 project(CUTLASS VERSION 2.11.0 LANGUAGES CXX)
 include(${CMAKE_CURRENT_SOURCE_DIR}/CUDA.cmake)
@@ -52,15 +53,19 @@ find_package(Doxygen QUIET)
 #
 # CUTLASS 2.x requires C++11
 #
-set(CMAKE_CXX_STANDARD 11)
-set(CMAKE_CXX_STANDARD_REQUIRED ON)
-set(CMAKE_CXX_EXTENSIONS OFF)
+if (NOT IMPLICIT_CMAKE_CXX_STANDARD)
+  set(CMAKE_CXX_STANDARD 11)
+  set(CMAKE_CXX_STANDARD_REQUIRED ON)
+  set(CMAKE_CXX_EXTENSIONS OFF)
+endif()
 
 if(CUTLASS_NATIVE_CUDA)
   set(CMAKE_CUDA_STANDARD 11)
   set(CMAKE_CUDA_STANDARD_REQUIRED ON)
 else()
-  list(APPEND CUTLASS_CUDA_NVCC_FLAGS --std=c++11)
+  if (NOT IMPLICIT_CMAKE_CXX_STANDARD)
+    list(APPEND CUTLASS_CUDA_NVCC_FLAGS --std=c++11)
+  endif()
 endif()
 
 if(CMAKE_INSTALL_PREFIX_INITIALIZED_TO_DEFAULT)
@@ -351,11 +356,21 @@ if (CMAKE_VERSION VERSION_GREATER_EQUAL 3.18)
 endif()
 
 function(cutlass_apply_cuda_gencode_flags TARGET)
+  set(options)
+  set(oneValueArgs)
+  set(multiValueArgs SM_ARCHS)
+  cmake_parse_arguments(_ "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
+
+  if (__SM_ARCHS)
+    set(ARCHS_ENABLED ${__SM_ARCHS})
+  else()
+    set(ARCHS_ENABLED ${CUTLASS_NVCC_ARCHS_ENABLED})
+  endif()
 
   set(NVCC_FLAGS)
   set(CLANG_FLAGS)
   set(__CMAKE_CUDA_ARCHS)
-  foreach(ARCH ${CUTLASS_NVCC_ARCHS_ENABLED})
+  foreach(ARCH ${ARCHS_ENABLED})
     list(APPEND CLANG_FLAGS --cuda-gpu-arch=sm_${ARCH})
     set(CODES)
     if(CUTLASS_NVCC_EMBED_CUBIN)
@@ -369,21 +384,37 @@ function(cutlass_apply_cuda_gencode_flags TARGET)
     list(JOIN CODES "," CODES_STR)
     list(APPEND NVCC_FLAGS -gencode=arch=compute_${ARCH},code=[${CODES_STR}])
   endforeach()
-
-  if (CUDA_COMPILER MATCHES "[Cc]lang")
-    target_compile_options(
-      ${TARGET}
-      PRIVATE
-      $<$<COMPILE_LANGUAGE:CXX>:${CLANG_FLAGS}>
-      )
-  elseif(CMAKE_VERSION GREATER_EQUAL 3.18)
-    set_property(TARGET ${TARGET} PROPERTY CUDA_ARCHITECTURES ${__CMAKE_CUDA_ARCHS})
+
+  if (NOT __SM_ARCHS)
+    if (CUDA_COMPILER MATCHES "[Cc]lang")
+      target_compile_options(
+        ${TARGET}
+        PRIVATE
+        $<$<COMPILE_LANGUAGE:CXX>:${CLANG_FLAGS}>
+        )
+    elseif(CMAKE_VERSION GREATER_EQUAL 3.18)
+      set_property(TARGET ${TARGET} PROPERTY CUDA_ARCHITECTURES ${__CMAKE_CUDA_ARCHS})
+    else()
+      target_compile_options(
+        ${TARGET}
+        PRIVATE
+        $<$<COMPILE_LANGUAGE:CUDA>:${NVCC_FLAGS}>
+        )
+    endif()
   else()
-    target_compile_options(
-      ${TARGET}
-      PRIVATE
-      $<$<COMPILE_LANGUAGE:CUDA>:${NVCC_FLAGS}>
-      )
+    list(JOIN CLANG_FLAGS " " CLANG_FLAGS_STR)
+    list(JOIN NVCC_FLAGS " " STR_NVCC_FLAGS)
+    if (CUDA_COMPILER MATCHES "[Cc]lang")
+      if(${TARGET} MATCHES ".*\.cpp")
+        set_source_files_properties(${TARGET} PROPERTIES COMPILE_FLAGS ${CLANG_FLAGS_STR})
+      endif()
+    elseif(CMAKE_VERSION GREATER_EQUAL 3.18)
+      set_source_files_properties(${TARGET} PROPERTIES CUDA_ARCHITECTURES ${STR_NVCC_FLAGS})
+    else()
+      if(${TARGET} MATCHES ".*\.cu")
+        set_source_files_properties(${TARGET} PROPERTIES COMPILE_FLAGS ${STR_NVCC_FLAGS})
+      endif()
+    endif()
   endif()
 
 endfunction()
@@ -463,7 +494,10 @@ set(CUTLASS_TOOLS_UTIL_INCLUDE_DIR ${CMAKE_CURRENT_SOURCE_DIR}/tools/util/includ
 include_directories(${CUTLASS_INCLUDE_DIR})
 
 target_compile_features(CUTLASS INTERFACE cxx_std_11)
-target_compile_definitions(CUTLASS INTERFACE CUTLASS_NAMESPACE=${CUTLASS_NAMESPACE})
+
+if (NOT CUTLASS_NAMESPACE STREQUAL "cutlass")
+  target_compile_definitions(CUTLASS INTERFACE CUTLASS_NAMESPACE=${CUTLASS_NAMESPACE})
+endif()
 
 if (NOT DEFINED CUTLASS_REVISION)
 
@@ -782,15 +816,23 @@ endif()
 
 ################################################################################
 
+include(CMakePackageConfigHelpers)
+
+write_basic_package_version_file(
+  ${CMAKE_CURRENT_BINARY_DIR}/NvidiaCutlassConfigVersion.cmake 
+  COMPATIBILITY AnyNewerVersion)
+
 install(
-  FILES ${CMAKE_CURRENT_SOURCE_DIR}/cmake/NvidiaCutlassConfig.cmake
-  DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/
+  FILES 
+    ${CMAKE_CURRENT_SOURCE_DIR}/cmake/NvidiaCutlassConfig.cmake
+    ${CMAKE_CURRENT_BINARY_DIR}/NvidiaCutlassConfigVersion.cmake 
+  DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/NvidiaCutlass/
   )
 
 install(
   EXPORT NvidiaCutlass
   NAMESPACE nvidia::cutlass::
-  DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/
+  DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/NvidiaCutlass/
   FILE NvidiaCutlassTargets.cmake
   )
 

CUDA.cmake

@@ -1,4 +1,4 @@
-# Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: BSD-3-Clause
 #
 # Redistribution and use in source and binary forms, with or without
@@ -80,7 +80,7 @@ find_library(
   lib64
   lib
   NO_DEFAULT_PATH
-  # We aren't going to search any system paths. We want to find the runtime
+  # We aren't going to search any system paths. We want to find the runtime 
   # in the CUDA toolkit we're building against.
   )
 
@@ -95,10 +95,10 @@ if(NOT TARGET cudart AND CUDART_LIBRARY)
     # from the PATH search.
   else()
     add_library(cudart SHARED IMPORTED GLOBAL)
-  endif()
+  endif()  
 
   add_library(nvidia::cudart ALIAS cudart)
-
+  
   set_property(
     TARGET cudart
     PROPERTY IMPORTED_LOCATION
@@ -126,7 +126,7 @@ find_library(
   lib64/stubs
   lib/stubs
   NO_DEFAULT_PATH
-  # We aren't going to search any system paths. We want to find the runtime
+  # We aren't going to search any system paths. We want to find the runtime 
   # in the CUDA toolkit we're building against.
   )
 
@@ -141,10 +141,10 @@ if(NOT TARGET cuda_driver AND CUDA_DRIVER_LIBRARY)
     # from the PATH search.
   else()
     add_library(cuda_driver SHARED IMPORTED GLOBAL)
-  endif()
+  endif()  
 
   add_library(nvidia::cuda_driver ALIAS cuda_driver)
-
+  
   set_property(
     TARGET cuda_driver
     PROPERTY IMPORTED_LOCATION
@@ -170,7 +170,7 @@ find_library(
   lib64
   lib
   NO_DEFAULT_PATH
-  # We aren't going to search any system paths. We want to find the runtime
+  # We aren't going to search any system paths. We want to find the runtime 
   # in the CUDA toolkit we're building against.
   )
 
@@ -185,10 +185,10 @@ if(NOT TARGET nvrtc AND NVRTC_LIBRARY)
     # from the PATH search.
   else()
     add_library(nvrtc SHARED IMPORTED GLOBAL)
-  endif()
-
+  endif()  
+  
   add_library(nvidia::nvrtc ALIAS nvrtc)
-
+  
   set_property(
     TARGET nvrtc
     PROPERTY IMPORTED_LOCATION
@@ -247,15 +247,15 @@ function(cutlass_unify_source_files TARGET_ARGS_VAR)
 
     set(CUDA_FILE_ARGS)
     set(TARGET_SOURCE_ARGS)
-
+    
     foreach(ARG ${__UNPARSED_ARGUMENTS})
       if(${ARG} MATCHES ".*\.cu$")
         list(APPEND CUDA_FILE_ARGS ${ARG})
       else()
         list(APPEND TARGET_SOURCE_ARGS ${ARG})
       endif()
     endforeach()
-
+    
     list(LENGTH CUDA_FILE_ARGS NUM_CUDA_FILE_ARGS)
     while(NUM_CUDA_FILE_ARGS GREATER 0)
       list(SUBLIST CUDA_FILE_ARGS 0 ${__BATCH_SIZE} CUDA_FILE_BATCH)
@@ -287,7 +287,7 @@ function(cutlass_unify_source_files TARGET_ARGS_VAR)
 endfunction()
 function(cutlass_add_library NAME)
 
-  set(options)
+  set(options SKIP_GENCODE_FLAGS)
   set(oneValueArgs EXPORT_NAME)
   set(multiValueArgs)
   cmake_parse_arguments(_ "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
@@ -303,7 +303,9 @@ function(cutlass_add_library NAME)
   endif()
 
   cutlass_apply_standard_compile_options(${NAME})
+  if (NOT __SKIP_GENCODE_FLAGS)
   cutlass_apply_cuda_gencode_flags(${NAME})
+  endif()
 
   target_compile_features(
    ${NAME}

LICENSE.txt

@@ -1,4 +1,4 @@
-Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 SPDX-License-Identifier: BSD-3-Clause
 
 Redistribution and use in source and binary forms, with or without

README.md

@@ -483,7 +483,7 @@ The official list of CUTLASS developers and contributors is available here: [CON
 
 # Copyright
 
-Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 SPDX-License-Identifier: BSD-3-Clause
 
 ```

cmake/nop.cu

@@ -1,5 +1,5 @@
 /***************************************************************************************************
- * Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: BSD-3-Clause
  *
  * Redistribution and use in source and binary forms, with or without

cuBLAS.cmake

@@ -1,4 +1,4 @@
-# Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: BSD-3-Clause
 #
 # Redistribution and use in source and binary forms, with or without

cuDNN.cmake

@@ -1,4 +1,4 @@
-# Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: BSD-3-Clause
 #
 # Redistribution and use in source and binary forms, with or without

examples/00_basic_gemm/CMakeLists.txt

@@ -1,5 +1,5 @@
 
-# Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: BSD-3-Clause
 #
 # Redistribution and use in source and binary forms, with or without

examples/00_basic_gemm/basic_gemm.cu

@@ -1,5 +1,5 @@
 /***************************************************************************************************
- * Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: BSD-3-Clause
  *
  * Redistribution and use in source and binary forms, with or without

examples/01_cutlass_utilities/CMakeLists.txt

@@ -1,5 +1,5 @@
 
-# Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: BSD-3-Clause
 #
 # Redistribution and use in source and binary forms, with or without

examples/01_cutlass_utilities/cutlass_utilities.cu

@@ -1,5 +1,5 @@
 /***************************************************************************************************
- * Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: BSD-3-Clause
  *
  * Redistribution and use in source and binary forms, with or without

examples/02_dump_reg_shmem/CMakeLists.txt

@@ -1,5 +1,5 @@
 
-# Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: BSD-3-Clause
 #
 # Redistribution and use in source and binary forms, with or without

examples/02_dump_reg_shmem/dump_reg_shmem.cu

@@ -1,5 +1,5 @@
 /***************************************************************************************************
- * Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: BSD-3-Clause
  *
  * Redistribution and use in source and binary forms, with or without

examples/03_visualize_layout/CMakeLists.txt

@@ -1,5 +1,5 @@
 
-# Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: BSD-3-Clause
 #
 # Redistribution and use in source and binary forms, with or without

examples/03_visualize_layout/options.h

@@ -1,5 +1,5 @@
 /***************************************************************************************************
- * Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: BSD-3-Clause
  *
  * Redistribution and use in source and binary forms, with or without

examples/03_visualize_layout/register_layout.cu

@@ -1,5 +1,5 @@
 /***************************************************************************************************
- * Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: BSD-3-Clause
  *
  * Redistribution and use in source and binary forms, with or without

examples/03_visualize_layout/register_layout.h

@@ -1,5 +1,5 @@
 /***************************************************************************************************
- * Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: BSD-3-Clause
  *
  * Redistribution and use in source and binary forms, with or without

examples/03_visualize_layout/visualize_layout.cpp

@@ -1,5 +1,5 @@
 /***************************************************************************************************
- * Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: BSD-3-Clause
  *
  * Redistribution and use in source and binary forms, with or without

examples/03_visualize_layout/visualize_layout.h

@@ -1,5 +1,5 @@
 /***************************************************************************************************
- * Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: BSD-3-Clause
  *
  * Redistribution and use in source and binary forms, with or without

examples/04_tile_iterator/CMakeLists.txt

@@ -1,5 +1,5 @@
 
-# Copyright (c) 2017 - 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2017 - 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: BSD-3-Clause
 #
 # Redistribution and use in source and binary forms, with or without