leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 25 Packages Projects Releases Wiki Activity

New updates for 2.11 (#775)

Browse Source 
* New updates.

* Minor profiler updates

Co-authored-by: Aniket Shivam <ashivam@nvidia.com>
This commit is contained in:
ANIKET SHIVAM
2023-01-20 16:32:57 -05:00
committed by GitHub
co-authored by Aniket Shivam
parent d49bef88f9
commit 66d9cddc83
1341 changed files with 2003 additions and 1732 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/cutlass/aligned_buffer.h
+1 -1
View File
@@ -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
include/cutlass/arch/arch.h
+1 -1
View File
@@ -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
include/cutlass/arch/cache_operation.h
+1 -1
View File
@@ -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
include/cutlass/arch/memory.h
+1 -1
View File
@@ -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
include/cutlass/arch/memory_sm75.h
+1 -1
View File
@@ -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
include/cutlass/arch/memory_sm80.h
+5 -5
View File
@@ -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
@@ -54,7 +54,7 @@ namespace arch {
/// Initiates an asynchronous copy from global memory to shared memory.
///
/// LDGSTS
/// cp.async
///
template <
/// Size of the access in bytes
@@ -66,7 +66,7 @@ struct cp_async;
/// Initiates an asynchronous copy from global memory to shared memory. Rather than predicate
/// the entire transfer, zeros are written to SMEM if the guard predicate is false.
///
/// LDGSTS
/// cp.async
///
template <
/// Size of the access in bytes
@@ -78,7 +78,7 @@ struct cp_async_zfill;
/// Initiates an asynchronous copy from global memory to shared memory. Rather than predicate
/// the entire transfer, nans (0x7eff) are written to SMEM if the guard predicate is false.
///
/// LDGSTS
/// cp.async
///
template <
/// Size of the access in bytes
@@ -90,7 +90,7 @@ struct cp_async_nan;
/// Either 0 or 1 are written to SMEM based on input element type
/// Used for diagonal elements of triangular matrix of BLAS3 functions
///
/// STS
/// st.shared
///
template <
/// Type of Element
include/cutlass/arch/mma.h
+1 -1
View File
@@ -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
include/cutlass/arch/mma_sm50.h
+1 -1
View File
@@ -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
include/cutlass/arch/mma_sm60.h
+1 -1
View File
@@ -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
include/cutlass/arch/mma_sm61.h
+1 -1
View File
@@ -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
include/cutlass/arch/mma_sm70.h
+1 -1
View File
@@ -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
include/cutlass/arch/mma_sm75.h
+2 -2
View File
@@ -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
@@ -1065,7 +1065,7 @@ struct Mma<
int const *C = reinterpret_cast<int const *>(&c);
int *D = reinterpret_cast<int *>(&d);
asm volatile("_mma.m8n8k32.row.col.u4.s4.sat {%0,%1}, %2, %3, {%4,%5};\n"
asm volatile("mma.sync.aligned.m8n8k32.row.col.satfinite.s32.u4.s4.s32 {%0,%1}, {%2}, {%3}, {%4,%5};\n"
: "=r"(D[0]), "=r"(D[1])
: "r"(A), "r"(B), "r"(C[0]), "r"(C[1]));
include/cutlass/arch/mma_sm80.h
+3 -3
View File
@@ -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
@@ -2166,13 +2166,13 @@ struct Mma<
"r"(C[0]), "r"(C[1]), "r"(C[2]), "r"(C[3]));
#else
CUTLASS_UNUSED(a);
CUTLASS_UNUSED(b);
CUTLASS_UNUSED(c);
CUTLASS_UNUSED(d);
assert(0);
#endif // defined(CUTLASS_ARCH_MMA_SM80_ENABLED)
}
};
include/cutlass/arch/mma_sm90.h
+1 -1
View File
@@ -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
include/cutlass/arch/mma_sparse_sm80.h
+1 -1
View File
@@ -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
include/cutlass/arch/simd.h
+1 -1
View File
@@ -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
include/cutlass/arch/simd_sm60.h
+1 -1
View File
@@ -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
include/cutlass/arch/simd_sm61.h
+1 -1
View File
@@ -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
include/cutlass/arch/wmma.h
+1 -1
View File
@@ -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
include/cutlass/arch/wmma_sm70.h
+1 -1
View File
@@ -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
include/cutlass/arch/wmma_sm72.h
+1 -1
View File
@@ -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
include/cutlass/arch/wmma_sm75.h
+1 -1
View File
@@ -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
include/cutlass/array.h
+1 -1
View File
@@ -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
include/cutlass/array_planar_complex.h
+1 -1
View File
@@ -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
include/cutlass/array_subbyte.h
+1 -1
View File
@@ -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
include/cutlass/barrier.h
+1 -1
View File
@@ -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
include/cutlass/bfloat16.h
+1 -1
View File
@@ -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
include/cutlass/blas3.h
+1 -1
View File
@@ -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
include/cutlass/block_striped.h
+1 -1
View File
@@ -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
include/cutlass/complex.h
+1 -1
View File
@@ -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
include/cutlass/constants.h
+1 -1
View File
@@ -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
include/cutlass/conv/conv2d_problem_size.h
+1 -1
View File
@@ -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
include/cutlass/conv/conv3d_problem_size.h
+1 -1
View File
@@ -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
include/cutlass/conv/convolution.h
+1 -1
View File
@@ -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
include/cutlass/conv/device/direct_convolution.h
+1 -1
View File
@@ -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
include/cutlass/conv/device/implicit_gemm_convolution.h
+1 -10
View File
@@ -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
@@ -162,15 +162,6 @@ public:
if (kConvolutionalOperator == conv::Operator::kDgrad &&
kStrideSupport == conv::StrideSupport::kStrided) {
// Unity stride (1x1) is supported by strided dgrad but disabled for performance
// reasons. For unity stride, use strided dgrad optimized unity stride specialization.
// Note that unit tests strided dgrad for unity stride to make sure that strided
// dgrad implemetnation is functionaly sound.
// Strided dgrad implementation also support mixed strides, i.e., (1x2) and (2x1)
if(args.problem_size.stride_h == 1 && args.problem_size.stride_w == 1) {
return Status::kErrorNotSupported;
}
// split-k (serial or parallel) is not supported for strided dgrad
if(args.problem_size.split_k_slices > 1) {
return Status::kErrorNotSupported;
include/cutlass/conv/device/implicit_gemm_convolution_fusion.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv2d.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv2d_dgrad.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv2d_fprop.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv2d_fprop_fusion.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv2d_fprop_with_broadcast.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv2d_fprop_with_reduction.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv2d_group_fprop.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv2d_wgrad.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv2d_wgrad_fusion.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv3d_dgrad.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv3d_fprop.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv3d_fprop_fusion.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_conv3d_wgrad.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/default_depthwise_fprop.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/direct_convolution.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/implicit_gemm_convolution.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/implicit_gemm_convolution_fusion.h
+1 -1
View File
@@ -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
include/cutlass/conv/kernel/implicit_gemm_convolution_strided_dgrad.h
+37 -37
View File
@@ -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
@@ -389,16 +389,15 @@ struct ImplicitGemmConvolutionStridedDgrad {
// Construct the semaphore.
int block_idx = threadblock_tile_idx.m() + threadblock_tile_idx.n() * params.grid_tiled_shape.m();
Semaphore semaphore(params.semaphore + block_idx, thread_idx);
// Compute logical position within grid
threadblock_tile_idx =
threadblock_swizzle.get_tile_offset(params.grid_tiled_shape);
// If performing a reduction via split-K, fetch the initial synchronization
if (params.split_k_mode == SplitKMode::kSerial && params.grid_tiled_shape.k() > 1) {
// Fetch the synchronization lock initially but do not block.
semaphore.fetch();
@@ -421,51 +420,51 @@ struct ImplicitGemmConvolutionStridedDgrad {
start_r, start_s,
threadblock_offset
);
// Tile iterator reading from source accumulator tensor
typename Epilogue::OutputTileIterator iterator_C(
params.iterator_C,
params.ptr_C,
ConvOutputIteratorParameter::extent(params.problem_size),
thread_idx,
params.stride_h_divmod, params.stride_w_divmod,
start_r, start_s,
threadblock_offset
);
// Construct the epilogue
Epilogue epilogue(
shared_storage.epilogue,
thread_idx,
warp_idx,
shared_storage.epilogue,
thread_idx,
warp_idx,
lane_idx);
// Wait on the semaphore - this latency may have been covered by iterator construction
if (params.split_k_mode == SplitKMode::kSerial && params.grid_tiled_shape.k() > 1) {
// For subsequent threadblocks, the source matrix is held in the 'D' tensor.
if (threadblock_tile_idx.k()) {
iterator_C = iterator_D;
if (output_op.is_source_needed())
{
// Tile iterator reading from source accumulator tensor
typename Epilogue::OutputTileIterator iterator_C(
params.iterator_C,
params.ptr_C,
ConvOutputIteratorParameter::extent(params.problem_size),
thread_idx,
params.stride_h_divmod, params.stride_w_divmod,
start_r, start_s,
threadblock_offset);
// Wait on the semaphore - this latency may have been covered by iterator construction
if (params.split_k_mode == SplitKMode::kSerial && params.grid_tiled_shape.k() > 1) {
// For subsequent threadblocks, the source matrix is held in the 'D' tensor.
if (threadblock_tile_idx.k()) {
iterator_C = iterator_D;
}
semaphore.wait(threadblock_tile_idx.k());
}
semaphore.wait(threadblock_tile_idx.k());
// Run epilogue with addend source iterator
epilogue(output_op, iterator_D, accumulators, iterator_C);
}
// Each split-k-slice writes to a unique tensor location
else if (params.split_k_mode == SplitKMode::kParallel) {
iterator_D.add_pointer_offset(threadblock_tile_idx.k() *
cutlass::conv::implicit_gemm_tensor_c_size(ConvOperator, params.problem_size));
else
{
// Run epilogue without addend source iterator
epilogue(output_op, iterator_D, accumulators);
}
// Run efficient epilogue
epilogue(output_op, iterator_D, accumulators, iterator_C);
//
// Release the semaphore
//
if (params.split_k_mode == SplitKMode::kSerial && params.grid_tiled_shape.k() > 1) {
if (params.split_k_mode == SplitKMode::kSerial && params.grid_tiled_shape.k() > 1) {
int lock = 0;
if (params.grid_tiled_shape.k() == threadblock_tile_idx.k() + 1) {
@@ -477,10 +476,11 @@ struct ImplicitGemmConvolutionStridedDgrad {
// Otherwise, the semaphore is incremented
lock = threadblock_tile_idx.k() + 1;
}
semaphore.release(lock);
}
}
}
};
/////////////////////////////////////////////////////////////////////////////////////////////////
include/cutlass/conv/kernel/implicit_gemm_convolution_with_fused_epilogue.h
+1 -1
View File
@@ -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
include/cutlass/conv/thread/depthwise_mma.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_dgrad_filter_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_dgrad_filter_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_dgrad_output_gradient_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_dgrad_output_gradient_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_fprop_activation_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_fprop_activation_tile_access_iterator_few_channels.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_fprop_activation_tile_access_iterator_fixed_channels.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_fprop_activation_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_fprop_filter_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_fprop_filter_tile_access_iterator_few_channels.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_fprop_filter_tile_access_iterator_fixed_channels.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_fprop_filter_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_params.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_tile_iterator.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_wgrad_activation_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_wgrad_activation_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_wgrad_output_gradient_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv2d_wgrad_output_gradient_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_dgrad_filter_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_dgrad_filter_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_dgrad_output_gradient_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_dgrad_output_gradient_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_fprop_activation_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_fprop_activation_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_fprop_filter_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_fprop_filter_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_params.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_wgrad_activation_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_wgrad_activation_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_wgrad_output_gradient_tile_access_iterator_analytic.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/conv3d_wgrad_output_gradient_tile_access_iterator_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/depthwise_direct_conv_params.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/depthwise_fprop_activation_tile_access_iterator_direct_conv_fixed_stride_dilation.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/depthwise_fprop_activation_tile_access_iterator_direct_conv_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/depthwise_fprop_direct_conv_multistage.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/depthwise_fprop_filter_tile_access_iterator_direct_conv_optimized.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/depthwise_fprop_pipelined.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/depthwise_mma_base.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/depthwise_mma_core_with_lane_access_size.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/implicit_gemm_fprop_fusion_multistage.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/implicit_gemm_multistage.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/implicit_gemm_pipelined.h
+1 -1
View File
@@ -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
include/cutlass/conv/threadblock/implicit_gemm_wgrad_fusion_multistage.h
+1 -1
View File
@@ -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

Some files were not shown because too many files have changed in this diff Show More