/**
* \file dnn/src/rocm/fill/fill.cpp.hip
* MegEngine is Licensed under the Apache License, Version 2.0 (the "License")
*
* Copyright (c) 2014-2021 Megvii Inc. All rights reserved.
*
* Unless required by applicable law or agreed to in writing,
* software distributed under the License is distributed on an
* "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
*/
#include "hcc_detail/hcc_defs_prologue.h"
#include "hip_header.h"
#include "megdnn/dtype.h"
#include "src/rocm/fill/fill.h.hip"
#include "src/rocm/utils.h.hip"
namespace {
template <typename T>
__global__ void kernel(T *dst, T value, uint32_t size) {
int32_t i = threadIdx.x + blockIdx.x * blockDim.x;
if (i < size) {
dst[i] = value;
}
}
} // anonymous namespace
namespace megdnn {
namespace rocm {
namespace fill {
template <typename T>
void exec_internal(T *dst, T value, size_t size, hipStream_t stream) {
hipLaunchKernelGGL(
(kernel<T>),
dim3(DIVUP(size, NR_THREADS)),
dim3(NR_THREADS),
0, stream, dst, value, size);
after_kernel_launch();
}
#define INST(T) template void exec_internal<T>(T *, \
T, size_t, hipStream_t);
#define cb(DType) INST(typename DTypeTrait<DType>::ctype)
MEGDNN_FOREACH_COMPUTING_DTYPE(cb)
} // namespace fill
} // namespace rocm
} // namespace megdnn
// vim: syntax=cpp.doxygen foldmethod=marker foldmarker=f{{{,f}}}