New updates for 2.11 (#775)

* New updates.

* Minor profiler updates

Co-authored-by: Aniket Shivam <ashivam@nvidia.com>

tools/library/scripts/generator.py

@@ -50,9 +50,8 @@ def EpilogueAlignment(max_alignment, tile, epilogue_steps = 8):
 #
 def CreateGemmOperator(manifest, layouts, tile_descriptions, data_type, \
   alignment_constraints, complex_transforms = None, epilogue_functor = EpilogueFunctor.LinearCombination, \
-#  swizzling_functor = SwizzlingFunctor.Identity8):
-# Use StreamK decomposition for basic GEMMs
-  swizzling_functor = SwizzlingFunctor.StreamK):
+  swizzling_functor = SwizzlingFunctor.Identity8):
+  # To use StreamK decomposition for basic GEMMs, set `swizzling_functor = SwizzlingFunctor.StreamK`
 
   if complex_transforms is None:
     complex_transforms = [(ComplexTransform.none, ComplexTransform.none),]
@@ -348,8 +347,7 @@ def CreateConv2dOperator(manifest, layout, tile_descriptions, data_type, alignme
   # one exceptional case
   
   # iterator algorithm (analytic and optimized)
-  #iterator_algorithms = [IteratorAlgorithm.Analytic, IteratorAlgorithm.Optimized]
-  iterator_algorithms = [IteratorAlgorithm.Optimized]
+  iterator_algorithms = [IteratorAlgorithm.Analytic, IteratorAlgorithm.Optimized]
 
   # by default, only generate the largest tile size, largest alignment, and optimized iterator
   if manifest.kernel_filter == '':
@@ -546,8 +544,7 @@ def CreateConv3dOperator(manifest, layout, tile_descriptions, data_type, alignme
   alignment_c = min(8, alignment)
   
   # iterator algorithm (analytic and optimized)
-#  iterator_algorithms = [IteratorAlgorithm.Analytic, IteratorAlgorithm.Optimized]
-  iterator_algorithms = [IteratorAlgorithm.Optimized]
+  iterator_algorithms = [IteratorAlgorithm.Analytic, IteratorAlgorithm.Optimized]
 
   # by default, only generate the largest tile size and optimized iterators
   if manifest.kernel_filter == '':

tools/library/scripts/manifest.py

@@ -190,7 +190,7 @@ class Manifest:
     self.args = args
     self.operation_count = 0
     self.operations_by_name = {}
-    
+
     self.kernel_filter = ''
     self.kernel_filter_list = []
     self.kernel_names = []
@@ -206,36 +206,41 @@ class Manifest:
       self.curr_build_dir = args.curr_build_dir
       architectures = args.architectures.split(';') if len(args.architectures) else ['50',]
       self.compute_capabilities = [int(x) for x in architectures]
-      
+    
       if args.filter_by_cc in ['false', 'False', '0']:
         self.filter_by_cc = False
-
-      if args.operations == 'all':
-        self.operations_enabled = []
-      else:
-        operations_list = [
-          OperationKind.Gemm
-          , OperationKind.Conv2d    
-          , OperationKind.Conv3d    
+    
+    if args.operations == 'all':
+      self.operations_enabled = []
+    else:
+      operations_list = [
+        OperationKind.Gemm
+        , OperationKind.Conv2d    
+        , OperationKind.Conv3d    
           , OperationKind.RankK
           , OperationKind.Trmm
           , OperationKind.Symm
-        ] 
-        self.operations_enabled = [x for x in operations_list if OperationKindNames[x] in args.operations.split(',')]
+      ] 
+      self.operations_enabled = [x for x in operations_list if OperationKindNames[x] in args.operations.split(',')]
 
-      if args.kernels == 'all':
-        self.kernel_names = []
-      else:
-        self.kernel_names = [x for x in args.kernels.split(',') if x != '']
+    if args.kernels == 'all':
+      self.kernel_names = []
+    else:
+      self.kernel_names = [x for x in args.kernels.split(',') if x != '']
 
-      self.ignore_kernel_names = [x for x in args.ignore_kernels.split(',') if x != '']
+    self.ignore_kernel_names = [x for x in args.ignore_kernels.split(',') if x != '']
+
+    if args.kernel_filter_file is None:
+        self.kernel_filter_list = []
+    else:
+        self.kernel_filter_list = self.get_kernel_filters(args.kernel_filter_file)
+
+
+    self.operation_count = 0
+    self.operations_by_name = {}
+    self.disable_full_archs_compilation = args.disable_full_archs_compilation
 
-      if args.kernel_filter_file is None:
-          self.kernel_filter_list = []
-      else:
-          self.kernel_filter_list = self.get_kernel_filters(args.kernel_filter_file)
 
-  #
   def get_kernel_filters (self, kernelListFile):
     if os.path.isfile(kernelListFile):
         with open(kernelListFile, 'r') as fileReader:
@@ -392,11 +397,58 @@ class Manifest:
   PRIVATE
 """, { 'target_name': target_name})
 
-      manifest_file.write(target_text)
+      manifest_file.write(target_text + '\n\n')
 
       for source_file in source_files:
         manifest_file.write("    %s\n" % str(source_file.replace('\\', '/')))
-      manifest_file.write(")")
+      manifest_file.write(")\n")
+
+      if self.disable_full_archs_compilation:
+
+        def for_hopper(name):
+            pass
+
+        def for_ampere(name):
+            return "16816" in name or \
+                   "16832" in name or \
+                   "16864" in name or \
+                   ("1688" in name and "tf32" in name)
+
+        def for_turing(name):
+            return ("1688" in name and "tf32" not in name) or \
+                    "8816" in name  
+
+        def for_volta(name):
+            return "884" in name
+
+        def is_cpp(name):
+            return name.endswith(".cpp")
+
+        def get_src_archs_str_given_requested_cuda_archs(archs, source_file):
+            intersected_archs = archs & set(self.compute_capabilities)
+            if intersected_archs == set():
+                raise RuntimeError(
+                      """
+                      Empty archs set for file {} after taking
+                      the intersection of {} (global requested archs) and
+                      {} (per file requested archs)
+                      """.format(source_file, set(self.compute_capabilities), archs))
+            else:
+                return " ".join(map(str, intersected_archs))
+
+        for source_file in source_files:
+            if is_cpp(source_file):
+                continue # skip because source is cpp
+            elif for_ampere(source_file):
+                archs_str = get_src_archs_str_given_requested_cuda_archs({80, 87, 90}, source_file)
+            elif for_turing(source_file):
+                archs_str = get_src_archs_str_given_requested_cuda_archs({75}, source_file)
+            elif for_volta(source_file):
+                archs_str = get_src_archs_str_given_requested_cuda_archs({70, 72}, source_file)
+            else:
+                raise RuntimeError("Per file archs are not set {}, as there is no rule specified for this file pattern".format(source_file))      
+ 
+            manifest_file.write("cutlass_apply_cuda_gencode_flags({} SM_ARCHS {})\n".format(str(source_file.replace('\\', '/')), archs_str))
   #
 
 ###################################################################################################

tools/library/scripts/pycutlass/docs/Makefile

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/docs/source/conf.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/profile/conv/conv2d_f16_sm80.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/profile/gemm/gemm_f32_sm80.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/cpp/compiler.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

tools/library/scripts/pycutlass/src/cpp/cutlass.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

tools/library/scripts/pycutlass/src/cpp/include/arch.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

tools/library/scripts/pycutlass/src/cpp/include/conv/conv_problem_size.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

tools/library/scripts/pycutlass/src/cpp/include/conv/convolution.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

tools/library/scripts/pycutlass/src/cpp/include/conv/host.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

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_generic.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
@@ -30,12 +30,8 @@
  **************************************************************************************************/
 
 /*! \file
-  
-  \brief A file contains the epilogue visitor with CTA row-wise broadcast
 
-  The epilogue rearranges the result of a matrix product through shared memory to match canonical
-  tensor layouts in global memory. Epilogues support conversion and reduction operations.
-                          
+  \brief A generic wrapper around an epilogue visitor operation
 */
 
 #pragma once

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/binary_ops.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
@@ -30,8 +30,8 @@
  **************************************************************************************************/
 
 /*! \file
-  
-  \brief A file contains the binary ops
+
+  \brief Binary operations to be used within the epilogue visitor model.
 */
 
 #pragma once
@@ -44,7 +44,7 @@ namespace cutlass {
 /////////////////////////////////////////////////////////////////////////////////////////////////
 
 
-/// Scalar multiplication
+/// Elementwise addition of two arrays
 template <typename T, int N>
 struct VectorAdd {
 

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/unary_ops.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
@@ -30,8 +30,8 @@
  **************************************************************************************************/
 
 /*! \file
-  
-  \brief A file contains the unary ops
+
+  \brief Unary operations to be used within the epilogue visitor model.
 */
 
 #pragma once

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/visitor_op_accumulator.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
@@ -30,8 +30,8 @@
  **************************************************************************************************/
 
 /*! \file
-  
-  \brief A file contains the epilogue visitor Op with accumulator
+
+  \brief Epilogue visitor operation that simply returns the accumulator
 */
 
 #pragma once

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/visitor_op_binary.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
@@ -30,8 +30,8 @@
  **************************************************************************************************/
 
 /*! \file
-  
-  \brief A file contains the epilogue visitor Op with Binary op
+
+  \brief Epilogue visitor operator performing a binary operation between two visitor nodes
 */
 
 #pragma once

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/visitor_op_column_broadcast.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
@@ -30,8 +30,8 @@
  **************************************************************************************************/
 
 /*! \file
-  
-  \brief A file contains the epilogue visitor Op with broadcasting vector to all columns
+
+  \brief Epilogue visitor operation that broadcasts a vector to all columns
 */
 
 #pragma once

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/visitor_op_column_reduction.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
@@ -30,8 +30,8 @@
  **************************************************************************************************/
 
 /*! \file
-  
-  \brief A file contains the epilogue visitor Op with reduction over columns in CTA
+
+  \brief Epilogue visitor operation that performs a column-wise reduction within a threadblock
 */
 
 #pragma once

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/visitor_op_linear_combination.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
@@ -30,8 +30,8 @@
  **************************************************************************************************/
 
 /*! \file
-  
-  \brief A file contains the epilogue visitor Op with Linear Combination
+
+  \brief Epilogue visitor operation that performs a linear combination of two visitor nodes
 */
 
 #pragma once

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/visitor_op_row_broadcast.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
@@ -30,8 +30,8 @@
  **************************************************************************************************/
 
 /*! \file
-  
-  \brief A file contains the epilogue visitor Op with broadcasting vector to all rows
+
+  \brief Epilogue visitor operation that broadcasts a vector to all rows
 */
 
 #pragma once

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/visitor_op_row_reduction.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
@@ -30,8 +30,8 @@
  **************************************************************************************************/
 
 /*! \file
-  
-  \brief A file contains the epilogue visitor Op with reduction over rows in CTA
+
+  \brief Epilogue visitor operation that performs a column-wise reduction within a threadblock
 */
 
 #pragma once

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/visitor_op_tensor_input.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

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/visitor_op_tensor_output.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

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_op/visitor_op_unary.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
@@ -30,8 +30,8 @@
  **************************************************************************************************/
 
 /*! \file
-  
-  \brief A file contains the epilogue visitor Op with Unary operation
+
+  \brief Epilogue visitor operator performing a unary operation atop a visitor node
 */
 
 #pragma once

tools/library/scripts/pycutlass/src/cpp/include/epilogue/epilogue_visitor_with_layernorm.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
@@ -30,12 +30,11 @@
  **************************************************************************************************/
 
 /*! \file
-    \brief A file contains all functioning classes needed by GemmLayernorm.
+    \brief Epilogue visitor type used for partial computation of a layernorm operation
 
     GemmLayernorm example =  GEMM0 with partial reduction fused in epilogue (EpilogueVisitorLayerNorm)
                           +  lightweight full reduction kernel (ApplyFinalReduction)
-                          +  GEMM1 with elemenwise operations fused in mainloop (GemmLayernormMainloopFusion)
-                          
+                          +  GEMM1 with elementwise operations fused in mainloop (GemmLayernormMainloopFusion)
 */
 
 #pragma once

tools/library/scripts/pycutlass/src/cpp/include/gemm/gemm.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

tools/library/scripts/pycutlass/src/cpp/include/gemm/gemm_universal_with_visitor.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

tools/library/scripts/pycutlass/src/cpp/include/gemm/host.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

tools/library/scripts/pycutlass/src/cpp/include/layout/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

tools/library/scripts/pycutlass/src/cpp/include/layout/matrix.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

tools/library/scripts/pycutlass/src/cpp/include/layout/tensor.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

tools/library/scripts/pycutlass/src/cpp/include/swizzling.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

tools/library/scripts/pycutlass/src/cpp/include/tensor_coord.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

tools/library/scripts/pycutlass/src/cpp/include/tensor_ref_view.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

tools/library/scripts/pycutlass/src/cpp/include/types.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

tools/library/scripts/pycutlass/src/cpp/test/conv/conv_problems.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

tools/library/scripts/pycutlass/src/cpp/test/conv/convolution.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

tools/library/scripts/pycutlass/src/cpp/test/conv/host.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

tools/library/scripts/pycutlass/src/cpp/test/gemm/gemm.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

tools/library/scripts/pycutlass/src/cpp/test/gemm/host.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

tools/library/scripts/pycutlass/src/pycutlass/arguments.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/c_types.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/compiler.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/conv2d_operation.py

@@ -1,6 +1,6 @@
 ################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/epilogue.py

@@ -1,6 +1,6 @@
 ################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/frontend.py

@@ -1,6 +1,6 @@
 ################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/gemm_operation.py

@@ -1,6 +1,6 @@
 ################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/library.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/memory_manager.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/operation.py

@@ -1,6 +1,6 @@
 ################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/parser.py

@@ -1,6 +1,6 @@
 ################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/reduction_operation.py

@@ -1,6 +1,6 @@
 ################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/tensor_ref.py

@@ -1,6 +1,6 @@
 ################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/test/conv2d_testbed.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/test/gemm_grouped_testbed.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/test/gemm_testbed.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/test/profiler.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/type_hint.py

@@ -1,6 +1,6 @@
 ################################################################################
 #
-# 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

tools/library/scripts/pycutlass/src/pycutlass/utils/device.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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
@@ -73,4 +73,4 @@ def device_cc(device: int = 0) -> int:
     deviceProp = check_cuda_errors(cudart.cudaGetDeviceProperties(device))
     major = str(deviceProp.major)
     minor = str(deviceProp.minor)
-    return int(major + minor)
+    return int(major + minor)

tools/library/scripts/pycutlass/src/pycutlass/utils/reference_model.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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

tools/library/scripts/pycutlass/test/frontend/test_frontend.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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
@@ -52,7 +52,6 @@ class Test_Frontend(unittest.TestCase):
         # Stages > 2 is supported only for compute capability 80 and beyond
         stages = 4 if cc >= 80 else 2
 
-
         tile_description = TileDescription(
             [128, 128, 8], stages, [2, 4, 1],
             math_inst

tools/library/scripts/pycutlass/test/unit/test_sm80.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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
@@ -37,11 +37,12 @@ from pycutlass import *
 from pycutlass.test import *
 from pycutlass.utils.device import device_cc
 import unittest
+import xmlrunner
+import argparse
 
 #
 # Create GEMM operation
 #
-
 @unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
 def TestGemmOperator(gemm_kind, math_inst, layout, alignment, tiling, arch, mixed=False,
     epilogue_functor=None, swizzling_functor=cutlass.IdentitySwizzle1, **kwargs):
@@ -447,7 +448,17 @@ class Test_SM80(unittest.TestCase):
         pass
 
 
+def argumentParser():
+    parser = argparse.ArgumentParser(description="Entrypoint for PyCutlass testing on Ampere architecture.")
+    parser.add_argument("-j", "--junit_path", help="The absolute path to the directory for generating a junit xml report", default="")
+    return parser.parse_args()
+
+
 if __name__ == '__main__':
     pycutlass.get_memory_pool(2**20, 2**34)
     pycutlass.compiler.nvcc()
-    unittest.main()
+    args = argumentParser()
+    if args.junit_path:
+        unittest.main(argv=[''], testRunner=xmlrunner.XMLTestRunner(output=args.junit_path))
+    else:
+        unittest.main(argv=[''])

tools/library/scripts/rt.py

@@ -1,6 +1,6 @@
 #################################################################################################
 #
-# 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