Skip to content

GitLab

Commit e24f37fb authored by aska-0096's avatar aska-0096
Browse files

Merge branch 'develop' of... 

Merge branch 'develop' of https://github.com/ROCmSoftwarePlatform/composable_kernel into add_int8_wmma_example_instance
parents 9697ad4e 85e2e1e2
Hide whitespace changes
Inline Side-by-side
Showing with 103 additions and 75 deletions
example/09_convnd_fwd/convnd_fwd_xdl_int8.cpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include "convnd_fwd_common.hpp" #include "convnd_fwd_common.hpp"
......
example/09_convnd_fwd/run_convnd_fwd_dl_example.inc
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once #pragma once
......
example/09_convnd_fwd/run_convnd_fwd_example.inc
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once #pragma once
......
example/10_convnd_fwd_multiple_d_multiple_reduce/CMakeLists.txt
View file @ e24f37fb
if(GPU_TARGETS MATCHES "gfx908" OR GPU_TARGETS MATCHES "gfx90a" OR GPU_TARGETS MATCHES "gfx940") list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
set(target 0)
foreach(gpu IN LISTS GPU_TARGETS)
if(gpu IN_LIST gpu_list AND target EQUAL 0)
add_custom_target(example_convnd_fwd_reduce_xdl) add_custom_target(example_convnd_fwd_reduce_xdl)
add_example_executable(example_convnd_fwd_max_xdl_int8 convnd_fwd_max_xdl_int8.cpp) if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_example_executable_no_testing(example_convnd_fwd_max_xdl_bf16 convnd_fwd_max_xdl_bf16.cpp) add_example_executable(example_convnd_fwd_max_xdl_int8 convnd_fwd_max_xdl_int8.cpp)
add_example_executable_no_testing(example_convnd_fwd_max_xdl_fp16 convnd_fwd_max_xdl_fp16.cpp) add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_int8)
add_example_executable(example_convnd_fwd_max_xdl_fp32 convnd_fwd_max_xdl_fp32.cpp) endif()
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_int8) if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_bf16) add_example_executable_no_testing(example_convnd_fwd_max_xdl_bf16 convnd_fwd_max_xdl_bf16.cpp)
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_fp16) add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_bf16)
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_fp32) endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable_no_testing(example_convnd_fwd_max_xdl_fp16 convnd_fwd_max_xdl_fp16.cpp)
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_fp16)
endif()
if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
add_example_executable(example_convnd_fwd_max_xdl_fp32 convnd_fwd_max_xdl_fp32.cpp)
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_fp32)
endif()
if(USE_BITINT_EXTENSION_INT4) if(USE_BITINT_EXTENSION_INT4)
add_example_executable(example_convnd_fwd_max_xdl_int4 convnd_fwd_max_xdl_int4.cpp) add_example_executable(example_convnd_fwd_max_xdl_int4 convnd_fwd_max_xdl_int4.cpp)
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_int4) add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_int4)
endif(USE_BITINT_EXTENSION_INT4) endif(USE_BITINT_EXTENSION_INT4)
endif() set(target 1)
\ No newline at end of file endif()
endforeach()
\ No newline at end of file
example/10_convnd_fwd_multiple_d_multiple_reduce/common.hpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <algorithm> #include <algorithm>
#include <cassert> #include <cassert>
......
example/10_convnd_fwd_multiple_d_multiple_reduce/convnd_fwd_max_xdl_bf16.cpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp" #include "common.hpp"
......
example/10_convnd_fwd_multiple_d_multiple_reduce/convnd_fwd_max_xdl_fp16.cpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp" #include "common.hpp"
......
example/10_convnd_fwd_multiple_d_multiple_reduce/convnd_fwd_max_xdl_fp32.cpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp" #include "common.hpp"
......
example/10_convnd_fwd_multiple_d_multiple_reduce/convnd_fwd_max_xdl_int4.cpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#ifndef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4 #ifndef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
#error Should compile this file with ck::int4_t support #error Should compile this file with ck::int4_t support
......
example/10_convnd_fwd_multiple_d_multiple_reduce/convnd_fwd_max_xdl_int8.cpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp" #include "common.hpp"
......
example/10_convnd_fwd_multiple_d_multiple_reduce/run_convnd_fwd_max_example.inc
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once #pragma once
......
example/12_reduce/reduce_blockwise.cpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <iostream> #include <iostream>
#include <initializer_list> #include <initializer_list>
......
example/12_reduce/reduce_blockwise_impl.hpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once #pragma once
......
example/12_reduce/reduce_blockwise_two_call.cpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <iostream> #include <iostream>
#include <numeric> #include <numeric>
......
example/12_reduce/reduce_example_common.hpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once #pragma once
......
example/12_reduce/reduce_multiblock_atomic_add.cpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <iostream> #include <iostream>
#include <initializer_list> #include <initializer_list>
......
example/12_reduce/reduce_multiblock_atomic_add_impl.hpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once #pragma once
......
example/13_pool2d_fwd/CMakeLists.txt
View file @ e24f37fb
add_example_executable(example_pool2d_fwd_fp16 pool2d_fwd_fp16.cpp) if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_pool2d_fwd_fp32 pool2d_fwd_fp32.cpp) add_example_executable(example_pool2d_fwd_fp16 pool2d_fwd_fp16.cpp)
endif()
if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
add_example_executable(example_pool2d_fwd_fp32 pool2d_fwd_fp32.cpp)
endif()
example/13_pool2d_fwd/pool2d_fwd_common.hpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once #pragma once
...@@ -39,31 +39,35 @@ bool pool_test(bool do_verification, ...@@ -39,31 +39,35 @@ bool pool_test(bool do_verification,
ck::index_t Wi, ck::index_t Wi,
ck::index_t window_stride_h, ck::index_t window_stride_h,
ck::index_t window_stride_w, ck::index_t window_stride_w,
ck::index_t window_dilation_h,
ck::index_t window_dilation_w,
ck::index_t in_left_pad_h, ck::index_t in_left_pad_h,
ck::index_t in_left_pad_w, ck::index_t in_left_pad_w,
ck::index_t in_right_pad_h, ck::index_t in_right_pad_h,
ck::index_t in_right_pad_w) ck::index_t in_right_pad_w)
{ {
using DevicePoolFwdInstance = using DevicePoolFwdInstance =
ck::tensor_operation::device::DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C< ck::tensor_operation::device::DevicePool2dFwd_NHWC_NHWC<InDataType,
InDataType, // InDataType OutDataType,
OutDataType, // OutDataType IndexDataType,
IndexDataType, // IndexDataType ComputeDataType,
ComputeDataType, // ComputeDataType ReduceOpId,
ReduceOpId, OutputIndex,
OutputIndex, 64, // BlockSize
64, // BlockSize 64, // ReduceMThreadClusterSize
64, // ReduceMThreadClusterSize 1, // ReduceKThreadClusterSize
1, // ReduceKThreadClusterSize 4, // ReduceMThreadSliceSize
4, // ReduceMThreadSliceSize 1, // ReduceKThreadSliceSize
1, // ReduceKThreadSliceSize 1>; // InSrcOutDstVectorSize
4>; // InSrcOutDstVectorSize
const ck::index_t Ys = (Y - 1) * window_dilation_h + 1;
const ck::index_t Ho = (Hi + in_left_pad_h + in_right_pad_h - Y) / window_stride_h + 1; const ck::index_t Xs = (X - 1) * window_dilation_w + 1;
const ck::index_t Wo = (Wi + in_left_pad_w + in_right_pad_w - X) / window_stride_w + 1; const ck::index_t Ho = (Hi + in_left_pad_h + in_right_pad_h - Ys) / window_stride_h + 1;
const ck::index_t Wo = (Wi + in_left_pad_w + in_right_pad_w - Xs) / window_stride_w + 1;
const std::vector<ck::index_t> window_spatial_lengths{Y, X}; const std::vector<ck::index_t> window_spatial_lengths{Y, X};
const std::vector<ck::index_t> window_strides{window_stride_h, window_stride_w}; const std::vector<ck::index_t> window_strides{window_stride_h, window_stride_w};
const std::vector<ck::index_t> window_dilations{window_dilation_h, window_dilation_w};
const std::vector<ck::index_t> input_left_pads{in_left_pad_h, in_left_pad_w}; const std::vector<ck::index_t> input_left_pads{in_left_pad_h, in_left_pad_w};
const std::vector<ck::index_t> input_right_pads{in_right_pad_h, in_right_pad_w}; const std::vector<ck::index_t> input_right_pads{in_right_pad_h, in_right_pad_w};
...@@ -123,6 +127,7 @@ bool pool_test(bool do_verification, ...@@ -123,6 +127,7 @@ bool pool_test(bool do_verification,
{C * Ho * Wo, 1, Wo * C, C}, {C * Ho * Wo, 1, Wo * C, C},
{C * Ho * Wo, 1, Wo * C, C}, {C * Ho * Wo, 1, Wo * C, C},
window_strides, window_strides,
window_dilations,
input_left_pads, input_left_pads,
input_right_pads, input_right_pads,
{2, 3}); {2, 3});
...@@ -144,8 +149,8 @@ bool pool_test(bool do_verification, ...@@ -144,8 +149,8 @@ bool pool_test(bool do_verification,
float gb_per_sec = num_btype / 1.E6 / ave_time; float gb_per_sec = num_btype / 1.E6 / ave_time;
std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s" std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec
<< std::endl; << " GB / s " << std::endl;
bool pass = true; bool pass = true;
...@@ -169,6 +174,7 @@ bool pool_test(bool do_verification, ...@@ -169,6 +174,7 @@ bool pool_test(bool do_verification,
out_indices_n_c_ho_wo_host, out_indices_n_c_ho_wo_host,
window_spatial_lengths, window_spatial_lengths,
window_strides, window_strides,
window_dilations,
input_left_pads, input_left_pads,
input_right_pads); input_right_pads);
......
example/13_pool2d_fwd/pool2d_fwd_fp16.cpp
View file @ e24f37fb
// SPDX-License-Identifier: MIT // SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved. // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#include <iostream> #include <iostream>
...@@ -34,18 +34,20 @@ int main(int argc, char* argv[]) ...@@ -34,18 +34,20 @@ int main(int argc, char* argv[])
bool time_kernel; bool time_kernel;
// Pool shape // Pool shape
ck::index_t N = 128; ck::index_t N = 128;
ck::index_t C = 192; ck::index_t C = 192;
ck::index_t Y = 3; ck::index_t Y = 3;
ck::index_t X = 3; ck::index_t X = 3;
ck::index_t Hi = 71; ck::index_t Hi = 71;
ck::index_t Wi = 71; ck::index_t Wi = 71;
ck::index_t window_stride_h = 2; ck::index_t window_stride_h = 2;
ck::index_t window_stride_w = 2; ck::index_t window_stride_w = 2;
ck::index_t in_left_pad_h = 1; ck::index_t window_dilation_h = 1;
ck::index_t in_left_pad_w = 1; ck::index_t window_dilation_w = 1;
ck::index_t in_right_pad_h = 1; ck::index_t in_left_pad_h = 1;
ck::index_t in_right_pad_w = 1; ck::index_t in_left_pad_w = 1;
ck::index_t in_right_pad_h = 1;
ck::index_t in_right_pad_w = 1;
if(argc == 1) if(argc == 1)
{ {
...@@ -59,31 +61,33 @@ int main(int argc, char* argv[]) ...@@ -59,31 +61,33 @@ int main(int argc, char* argv[])
init_method = std::stoi(argv[2]); init_method = std::stoi(argv[2]);
time_kernel = static_cast<bool>(std::stoi(argv[3])); time_kernel = static_cast<bool>(std::stoi(argv[3]));
} }
else if(argc == 16) else if(argc == 18)
{ {
do_verification = std::stoi(argv[1]); do_verification = std::stoi(argv[1]);
init_method = std::stoi(argv[2]); init_method = std::stoi(argv[2]);
time_kernel = static_cast<bool>(std::stoi(argv[3])); time_kernel = static_cast<bool>(std::stoi(argv[3]));
N = std::stoi(argv[4]); N = std::stoi(argv[4]);
C = std::stoi(argv[5]); C = std::stoi(argv[5]);
Y = std::stoi(argv[6]); Y = std::stoi(argv[6]);
X = std::stoi(argv[7]); X = std::stoi(argv[7]);
Hi = std::stoi(argv[8]); Hi = std::stoi(argv[8]);
Wi = std::stoi(argv[9]); Wi = std::stoi(argv[9]);
window_stride_h = std::stoi(argv[10]); window_stride_h = std::stoi(argv[10]);
window_stride_w = std::stoi(argv[11]); window_stride_w = std::stoi(argv[11]);
in_left_pad_h = std::stoi(argv[12]); window_dilation_h = std::stoi(argv[12]);
in_left_pad_w = std::stoi(argv[13]); window_dilation_w = std::stoi(argv[13]);
in_right_pad_h = std::stoi(argv[14]); in_left_pad_h = std::stoi(argv[14]);
in_right_pad_w = std::stoi(argv[15]); in_left_pad_w = std::stoi(argv[15]);
in_right_pad_h = std::stoi(argv[16]);
in_right_pad_w = std::stoi(argv[17]);
} }
else else
{ {
printf("arg1: verification (0=no, 1=yes)\n"); printf("arg1: verification (0=no, 1=yes)\n");
printf("arg2: initialization (0=no init, 1=integer value, 2=decimal value)\n"); printf("arg2: initialization (0=no init, 1=integer value, 2=decimal value)\n");
printf("arg3: time kernel (0=no, 1=yes)\n"); printf("arg3: time kernel (0=no, 1=yes)\n");
printf("arg4 to 15: N, C, Y, X, Hi, Wi, Sy, Sx, LeftPy, LeftPx, RightPy, " printf("arg4 to 15: N, C, Y, X, Hi, Wi, Sy, Sx, Dy, Dx, LeftPy, LeftPx, RightPy, "
"RightPx\n"); "RightPx\n");
exit(0); exit(0);
} }
...@@ -107,6 +111,8 @@ int main(int argc, char* argv[]) ...@@ -107,6 +111,8 @@ int main(int argc, char* argv[])
Wi, Wi,
window_stride_h, window_stride_h,
window_stride_w, window_stride_w,
window_dilation_h,
window_dilation_w,
in_left_pad_h, in_left_pad_h,
in_left_pad_w, in_left_pad_w,
in_right_pad_h, in_right_pad_h,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment