mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-15 10:37:44 +00:00
4a141b5a04
* gemm+activation
* move C pointwise operation into threadwise copy
* add pointwise operation to A/B matrix
* update ckProfiler
* adding bias add
* adding bias add
* adding bias add
* added bias add; worked around compiler issues
* clean up
* clean up
* Update README.md
* Update README.md
* Update README.md
* clean up
* add conv_xdl example
* adding conv_xdl_bias_relu_add example
* add conv+bias+relu+add, but has register spill issue
* tweak
* tweak
* refactor
* Update README.md
update readme for example/2_gemm_xdl_bias_relu_add
* clean up
* Update README.md
update readme for example/3_conv_xdl
* Update README.md
[ROCm/composable_kernel commit: 41cdd3801a]
35 lines
1.1 KiB
C++
35 lines
1.1 KiB
C++
#pragma once
|
|
#include "host_tensor.hpp"
|
|
|
|
template <typename AType,
|
|
typename BType,
|
|
typename CType,
|
|
typename AElementwiseOperation,
|
|
typename BElementwiseOperation,
|
|
typename CElementwiseOperation>
|
|
void host_gemm_mk_kn_mn(const Tensor<AType>& a_m_k,
|
|
const Tensor<BType>& b_k_n,
|
|
Tensor<CType>& c_m_n,
|
|
const AElementwiseOperation& a_element_op,
|
|
const BElementwiseOperation& b_element_op,
|
|
const CElementwiseOperation& c_element_op)
|
|
{
|
|
auto f_mk_kn_mn = [&](auto m, auto n) {
|
|
const int K = a_m_k.mDesc.GetLengths()[1];
|
|
|
|
double v = 0;
|
|
|
|
for(int k = 0; k < K; ++k)
|
|
{
|
|
v += static_cast<const double>(a_element_op(a_m_k(m, k))) *
|
|
static_cast<const double>(b_element_op(b_k_n(k, n)));
|
|
}
|
|
|
|
c_m_n(m, n) = c_element_op(v);
|
|
};
|
|
|
|
make_ParallelTensorFunctor(f_mk_kn_mn,
|
|
c_m_n.mDesc.GetLengths()[0],
|
|
c_m_n.mDesc.GetLengths()[1])(std::thread::hardware_concurrency());
|
|
}
|