Skip to content

GitLab

Commit 06d2c7b1 authored by Jing Zhang's avatar Jing Zhang Committed by root
Browse files

clean

parents b27909a0 3b18f1e3
Hide whitespace changes
Inline Side-by-side
Showing with 108 additions and 189 deletions
example/12_reduce/reduce_example_common.hpp
View file @ 06d2c7b1
// 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 @ 06d2c7b1
// 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 @ 06d2c7b1
// 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/pool2d_fwd_common.hpp
View file @ 06d2c7b1
// 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
...@@ -17,115 +17,11 @@ ...@@ -17,115 +17,11 @@
#include "ck/library/utility/host_tensor.hpp" #include "ck/library/utility/host_tensor.hpp"
#include "ck/library/utility/host_tensor_generator.hpp" #include "ck/library/utility/host_tensor_generator.hpp"
#include "ck/library/utility/literals.hpp" #include "ck/library/utility/literals.hpp"
#include "ck/library/reference_tensor_operation/cpu/reference_pool_fwd.hpp"
template <typename InDataType, template <typename InDataType,
typename OutDataType, typename OutDataType,
typename AccDataType, typename ComputeDataType,
typename IndexDataType,
ck::ReduceTensorOp ReduceOpId,
bool PropagateNan,
bool OutputIndex>
static void pool_host_verify(const Tensor<InDataType>& in,
Tensor<OutDataType>& out,
Tensor<IndexDataType>& out_indices,
const std::array<ck::index_t, 2>& window_spatial_lengths,
const std::array<ck::index_t, 2>& window_strides,
const std::array<ck::index_t, 2>& in_left_pads,
const std::array<ck::index_t, 2>& /*in_right_pads*/)
{
const int32_t reduceLength = window_spatial_lengths[0] * window_spatial_lengths[1];
using ReduceOperation = typename ck::reduce_binary_operator<ReduceOpId>::opType;
auto elementwise_ops =
ck::reduce_unary_operator<ReduceOpId, true, true>::GetElementwiseOperator(reduceLength);
auto in_elementwise_op = std::get<0>(elementwise_ops);
auto acc_elementwise_op = std::get<1>(elementwise_ops);
if constexpr(!OutputIndex)
{
using Accumulation =
ck::detail::AccumulateWithNanCheck<PropagateNan, ReduceOperation, AccDataType>;
auto f_nchw = [&](auto n, auto c, auto ho, auto wo) {
auto accuVal = ReduceOperation::template GetIdentityValue<AccDataType>();
for(ck::index_t y = 0; y < window_spatial_lengths[0]; ++y)
{
ck::index_t hi = ho * window_strides[0] + y - in_left_pads[0];
for(ck::index_t x = 0; x < window_spatial_lengths[1]; ++x)
{
ck::index_t wi = wo * window_strides[1] + x - in_left_pads[1];
if(hi >= 0 && hi < static_cast<ck::index_t>(in.mDesc.GetLengths()[2]) &&
wi >= 0 && wi < static_cast<ck::index_t>(in.mDesc.GetLengths()[3]))
{
AccDataType currVal = static_cast<AccDataType>(in(n, c, hi, wi));
in_elementwise_op(currVal, currVal);
Accumulation::Calculate(accuVal, currVal);
}
}
}
acc_elementwise_op(accuVal, accuVal);
out(n, c, ho, wo) = accuVal;
};
make_ParallelTensorFunctor(f_nchw,
out.mDesc.GetLengths()[0],
out.mDesc.GetLengths()[1],
out.mDesc.GetLengths()[2],
out.mDesc.GetLengths()[3])(std::thread::hardware_concurrency());
}
else
{
using Accumulation = ck::detail::AccumulateWithIndexAndNanCheck<PropagateNan,
ReduceOperation,
AccDataType,
IndexDataType>;
auto f_nchw = [&](auto n, auto c, auto ho, auto wo) {
auto accuVal = ReduceOperation::template GetIdentityValue<AccDataType>();
IndexDataType accuIndex = 0;
for(ck::index_t y = 0; y < window_spatial_lengths[0]; ++y)
{
ck::index_t hi = ho * window_strides[0] + y - in_left_pads[0];
for(ck::index_t x = 0; x < window_spatial_lengths[1]; ++x)
{
ck::index_t wi = wo * window_strides[1] + x - in_left_pads[1];
if(hi >= 0 && hi < in.mDesc.GetLengths()[2] && wi >= 0 &&
wi < in.mDesc.GetLengths()[3])
{
AccDataType currVal = static_cast<AccDataType>(in(n, c, hi, wi));
IndexDataType currIndex = y * window_spatial_lengths[1] + x;
in_elementwise_op(currVal, currVal);
Accumulation::Calculate(accuVal, currVal, accuIndex, currIndex);
}
}
}
acc_elementwise_op(accuVal, accuVal);
out(n, c, ho, wo) = accuVal;
out_indices(n, c, ho, wo) = accuIndex;
};
make_ParallelTensorFunctor(f_nchw,
out.mDesc.GetLengths()[0],
out.mDesc.GetLengths()[1],
out.mDesc.GetLengths()[2],
out.mDesc.GetLengths()[3])(std::thread::hardware_concurrency());
};
}
template <typename InDataType,
typename OutDataType,
typename AccDataType,
typename IndexDataType, typename IndexDataType,
typename InLayout, typename InLayout,
typename OutLayout, typename OutLayout,
...@@ -150,9 +46,10 @@ bool pool_test(bool do_verification, ...@@ -150,9 +46,10 @@ bool pool_test(bool do_verification,
{ {
using DevicePoolFwdInstance = using DevicePoolFwdInstance =
ck::tensor_operation::device::DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C< ck::tensor_operation::device::DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C<
InDataType, // InDataType InDataType, // InDataType
OutDataType, // OutDataType OutDataType, // OutDataType
AccDataType, // AccDataType IndexDataType, // IndexDataType
ComputeDataType, // ComputeDataType
ReduceOpId, ReduceOpId,
OutputIndex, OutputIndex,
64, // BlockSize 64, // BlockSize
...@@ -165,10 +62,10 @@ bool pool_test(bool do_verification, ...@@ -165,10 +62,10 @@ bool pool_test(bool do_verification,
const ck::index_t Ho = (Hi + in_left_pad_h + in_right_pad_h - Y) / window_stride_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 Wo = (Wi + in_left_pad_w + in_right_pad_w - X) / window_stride_w + 1; const ck::index_t Wo = (Wi + in_left_pad_w + in_right_pad_w - X) / window_stride_w + 1;
const std::array<ck::index_t, 2> window_spatial_lengths{{Y, X}}; const std::vector<ck::index_t> window_spatial_lengths{Y, X};
const std::array<ck::index_t, 2> window_strides{{window_stride_h, window_stride_w}}; const std::vector<ck::index_t> window_strides{window_stride_h, window_stride_w};
const std::array<ck::index_t, 2> 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::array<ck::index_t, 2> 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};
// tensor layout // tensor layout
auto f_host_tensor_descriptor = auto f_host_tensor_descriptor =
...@@ -219,14 +116,16 @@ bool pool_test(bool do_verification, ...@@ -219,14 +116,16 @@ bool pool_test(bool do_verification,
static_cast<InDataType*>(in_device_buf.GetDeviceBuffer()), static_cast<InDataType*>(in_device_buf.GetDeviceBuffer()),
static_cast<OutDataType*>(out_device_buf.GetDeviceBuffer()), static_cast<OutDataType*>(out_device_buf.GetDeviceBuffer()),
static_cast<IndexDataType*>(out_indices_device_buf.GetDeviceBuffer()), static_cast<IndexDataType*>(out_indices_device_buf.GetDeviceBuffer()),
N, {N, C, Hi, Wi},
C, {Y, X},
std::array<ck::index_t, 2>{{Hi, Wi}}, {N, C, Ho, Wo},
std::array<ck::index_t, 2>{{Y, X}}, {C * Hi * Wi, 1, Wi * C, C},
std::array<ck::index_t, 2>{{Ho, Wo}}, {C * Ho * Wo, 1, Wo * C, C},
{C * Ho * Wo, 1, Wo * C, C},
window_strides, window_strides,
input_left_pads, input_left_pads,
input_right_pads); input_right_pads,
{2, 3});
if(!pool.IsSupportedArgument(argument_ptr.get())) if(!pool.IsSupportedArgument(argument_ptr.get()))
{ {
...@@ -252,19 +151,28 @@ bool pool_test(bool do_verification, ...@@ -252,19 +151,28 @@ bool pool_test(bool do_verification,
if(do_verification) if(do_verification)
{ {
pool_host_verify<InDataType, using ReferencePoolingFwdInstance =
OutDataType, ck::tensor_operation::host::ReferencePoolingFwd<4,
AccDataType, 2,
IndexDataType, InDataType,
ReduceOpId, OutDataType,
PropagateNan, ComputeDataType,
OutputIndex>(in_n_c_hi_wi, IndexDataType,
out_n_c_ho_wo_host, ReduceOpId,
out_indices_n_c_ho_wo_host, PropagateNan,
window_spatial_lengths, OutputIndex>;
window_strides,
input_left_pads, auto ref_pooling = ReferencePoolingFwdInstance{};
input_right_pads); auto ref_pooling_invoker = ref_pooling.MakeInvoker();
auto ref_pooling_argument = ref_pooling.MakeArgument(in_n_c_hi_wi,
out_n_c_ho_wo_host,
out_indices_n_c_ho_wo_host,
window_spatial_lengths,
window_strides,
input_left_pads,
input_right_pads);
ref_pooling_invoker.Run(ref_pooling_argument);
out_device_buf.FromDevice(out_n_c_ho_wo_device.mData.data()); out_device_buf.FromDevice(out_n_c_ho_wo_device.mData.data());
......
example/13_pool2d_fwd/pool2d_fwd_fp16.cpp
View file @ 06d2c7b1
// 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 <cstdlib>
#include "ck/ck.hpp" #include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp" #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
...@@ -10,9 +9,9 @@ ...@@ -10,9 +9,9 @@
#include "pool2d_fwd_common.hpp" #include "pool2d_fwd_common.hpp"
using InDataType = ck::half_t; using InDataType = ck::half_t;
using OutDataType = ck::half_t; using OutDataType = ck::half_t;
using AccDataType = float; using ComputeDataType = float;
using IndexDataType = int32_t; using IndexDataType = int32_t;
...@@ -91,7 +90,7 @@ int main(int argc, char* argv[]) ...@@ -91,7 +90,7 @@ int main(int argc, char* argv[])
bool pass = pool_test<InDataType, bool pass = pool_test<InDataType,
OutDataType, OutDataType,
AccDataType, ComputeDataType,
IndexDataType, IndexDataType,
InLayout, InLayout,
OutLayout, OutLayout,
......
example/13_pool2d_fwd/pool2d_fwd_fp32.cpp
View file @ 06d2c7b1
// 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 <cstdlib>
#include "ck/ck.hpp" #include "ck/ck.hpp"
#include "ck/utility/reduction_enums.hpp" #include "ck/utility/reduction_enums.hpp"
...@@ -10,9 +9,9 @@ ...@@ -10,9 +9,9 @@
#include "pool2d_fwd_common.hpp" #include "pool2d_fwd_common.hpp"
using InDataType = float; using InDataType = float;
using OutDataType = float; using OutDataType = float;
using AccDataType = float; using ComputeDataType = float;
using IndexDataType = int32_t; using IndexDataType = int32_t;
...@@ -91,7 +90,7 @@ int main(int argc, char* argv[]) ...@@ -91,7 +90,7 @@ int main(int argc, char* argv[])
bool pass = pool_test<InDataType, bool pass = pool_test<InDataType,
OutDataType, OutDataType,
AccDataType, ComputeDataType,
IndexDataType, IndexDataType,
InLayout, InLayout,
OutLayout, OutLayout,
......
example/14_gemm_quantization/CMakeLists.txt
View file @ 06d2c7b1
...@@ -2,5 +2,12 @@ ...@@ -2,5 +2,12 @@
add_example_executable(example_gemm_dl_quantization_int8 gemm_dl_quantization_int8.cpp) add_example_executable(example_gemm_dl_quantization_int8 gemm_dl_quantization_int8.cpp)
# xdlops # xdlops
add_example_executable(example_gemm_xdl_bias_relu_quantization_int8 gemm_xdl_bias_relu_quantization_int8.cpp) list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
add_example_executable(example_gemm_xdl_quantization_int8 gemm_xdl_quantization_int8.cpp) set(target 0)
\ No newline at end of file foreach(gpu IN LISTS GPU_TARGETS)
if(gpu IN_LIST gpu_list AND target EQUAL 0)
add_example_executable(example_gemm_xdl_bias_relu_quantization_int8 gemm_xdl_bias_relu_quantization_int8.cpp)
add_example_executable(example_gemm_xdl_quantization_int8 gemm_xdl_quantization_int8.cpp)
set(target 1)
endif()
endforeach()
\ No newline at end of file
example/14_gemm_quantization/gemm_dl_quantization_int8.cpp
View file @ 06d2c7b1
// 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/14_gemm_quantization/gemm_xdl_bias_relu_quantization_int8.cpp
View file @ 06d2c7b1
// 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/14_gemm_quantization/gemm_xdl_quantization_int8.cpp
View file @ 06d2c7b1
// 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/15_grouped_gemm/grouped_gemm_multiple_d_dl_fp16.cpp
View file @ 06d2c7b1
// 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 <cstddef> #include <cstddef>
......
example/15_grouped_gemm/grouped_gemm_xdl_bfp16.cpp
View file @ 06d2c7b1
// 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/15_grouped_gemm/grouped_gemm_xdl_fp16.cpp
View file @ 06d2c7b1
// 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/15_grouped_gemm/grouped_gemm_xdl_fp32.cpp
View file @ 06d2c7b1
// 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/15_grouped_gemm/grouped_gemm_xdl_int4.cpp
View file @ 06d2c7b1
// 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/15_grouped_gemm/grouped_gemm_xdl_int8.cpp
View file @ 06d2c7b1
// 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/15_grouped_gemm/grouped_gemm_xdl_splitk_fp16.cpp
View file @ 06d2c7b1
// 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>
...@@ -68,8 +68,7 @@ int main(int argc, char* argv[]) ...@@ -68,8 +68,7 @@ int main(int argc, char* argv[])
problem_size.group_count = 16; problem_size.group_count = 16;
problem_size.Ms = {2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2}; problem_size.Ms = {2, 1, 1, 1, 1, 1, 3, 4, 3, 5, 2, 4, 2, 1, 0, 1};
// problem_size.Ms = {2, 1, 1, 1, 1, 1, 3, 4, 3, 5, 2, 4, 2, 1, 0, 1};
for(int i = 0; i < problem_size.group_count; i++) for(int i = 0; i < problem_size.group_count; i++)
{ {
......
example/16_gemm_multi_d_multi_reduces/CMakeLists.txt
View file @ 06d2c7b1
add_custom_target(example_gemm_reduce_xdl) list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
add_custom_target(example_gemm_reduce_xdl_max) set(target 0)
add_custom_target(example_gemm_reduce_xdl_mean_meansquare) foreach(gpu IN LISTS GPU_TARGETS)
add_custom_target(example_gemm_add_add_mean_meansquare_xdl) if(gpu IN_LIST gpu_list AND target EQUAL 0)
add_custom_target(example_gemm_reduce_xdl)
add_example_executable(example_gemm_max_xdl_fp16 gemm_max_xdl_fp16.cpp) add_custom_target(example_gemm_reduce_xdl_max)
add_example_executable(example_gemm_max_xdl_int8 gemm_max_xdl_int8.cpp) add_custom_target(example_gemm_reduce_xdl_mean_meansquare)
add_example_executable(example_gemm_max_xdl_fp32 gemm_max_xdl_fp32.cpp) add_custom_target(example_gemm_add_add_mean_meansquare_xdl)
add_example_executable(example_gemm_max_xdl_bf16 gemm_max_xdl_bf16.cpp)
add_example_executable(example_gemm_max_xdl_fp16 gemm_max_xdl_fp16.cpp)
add_example_executable(example_gemm_add_add_mean_meansquare_xdl_fp16 gemm_add_add_mean_meansquare_xdl_fp16.cpp) add_example_executable(example_gemm_max_xdl_int8 gemm_max_xdl_int8.cpp)
add_example_executable(example_gemm_max_xdl_fp32 gemm_max_xdl_fp32.cpp)
add_example_executable(example_gemm_mean_meansquare_xdl_fp16 gemm_mean_meansquare_xdl_fp16.cpp) add_example_executable(example_gemm_max_xdl_bf16 gemm_max_xdl_bf16.cpp)
add_example_executable(example_gemm_mean_meansquare_xdl_fp32 gemm_mean_meansquare_xdl_fp32.cpp)
add_example_executable(example_gemm_mean_meansquare_xdl_bf16 gemm_mean_meansquare_xdl_bf16.cpp) add_example_executable(example_gemm_add_add_mean_meansquare_xdl_fp16 gemm_add_add_mean_meansquare_xdl_fp16.cpp)
add_example_executable(example_gemm_add_addsquare_xdl_int8 gemm_add_addsquare_xdl_int8.cpp)
add_example_executable(example_gemm_mean_meansquare_xdl_fp16 gemm_mean_meansquare_xdl_fp16.cpp)
add_dependencies(example_gemm_reduce_xdl_max add_example_executable(example_gemm_mean_meansquare_xdl_fp32 gemm_mean_meansquare_xdl_fp32.cpp)
add_example_executable(example_gemm_mean_meansquare_xdl_bf16 gemm_mean_meansquare_xdl_bf16.cpp)
add_example_executable(example_gemm_add_addsquare_xdl_int8 gemm_add_addsquare_xdl_int8.cpp)
add_dependencies(example_gemm_reduce_xdl_max
example_gemm_max_xdl_bf16 example_gemm_max_xdl_bf16
example_gemm_max_xdl_fp16 example_gemm_max_xdl_fp16
example_gemm_max_xdl_fp32 example_gemm_max_xdl_fp32
example_gemm_max_xdl_int8) example_gemm_max_xdl_int8)
add_dependencies(example_gemm_reduce_xdl_mean_meansquare add_dependencies(example_gemm_reduce_xdl_mean_meansquare
example_gemm_mean_meansquare_xdl_fp16 example_gemm_mean_meansquare_xdl_fp16
example_gemm_mean_meansquare_xdl_fp32 example_gemm_mean_meansquare_xdl_fp32
example_gemm_mean_meansquare_xdl_bf16 example_gemm_mean_meansquare_xdl_bf16
example_gemm_add_addsquare_xdl_int8) example_gemm_add_addsquare_xdl_int8)
add_dependencies(example_gemm_add_add_mean_meansquare_xdl example_gemm_add_add_mean_meansquare_xdl_fp16) add_dependencies(example_gemm_add_add_mean_meansquare_xdl example_gemm_add_add_mean_meansquare_xdl_fp16)
add_dependencies(example_gemm_reduce_xdl add_dependencies(example_gemm_reduce_xdl
example_gemm_reduce_xdl_mean_meansquare example_gemm_reduce_xdl_mean_meansquare
example_gemm_reduce_xdl_max example_gemm_reduce_xdl_max
example_gemm_add_add_mean_meansquare_xdl) example_gemm_add_add_mean_meansquare_xdl)
if(USE_BITINT_EXTENSION_INT4) if(USE_BITINT_EXTENSION_INT4)
add_example_executable(example_gemm_max_xdl_int4 gemm_max_xdl_int4.cpp) add_example_executable(example_gemm_max_xdl_int4 gemm_max_xdl_int4.cpp)
add_dependencies(example_gemm_reduce_xdl_max example_gemm_max_xdl_int4) add_dependencies(example_gemm_reduce_xdl_max example_gemm_max_xdl_int4)
endif() endif()
set(target 1)
endif()
endforeach()
example/16_gemm_multi_d_multi_reduces/gemm_add_add_mean_meansquare_xdl_fp16.cpp
View file @ 06d2c7b1
// 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/16_gemm_multi_d_multi_reduces/gemm_add_addsquare_xdl_int8.cpp
View file @ 06d2c7b1
// 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 "gemm_reduce_xdl_common.hpp" #include "gemm_reduce_xdl_common.hpp"
......
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