Skip to content

【训练营】BF16 Elementwise Kernel 优化 #126

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
WjhisTy wants to merge 1 commit into
base: master
Choose a base branch
from
Open

【训练营】BF16 Elementwise Kernel 优化 #126

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 12 additions & 2 deletions infini_train/src/kernels/cuda/elementwise.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,17 +4,27 @@

#include "infini_train/include/common/cuda/common_cuda.h"
#include "infini_train/include/common/cuda/kernel_helper.cuh"
#include "infini_train/include/core/runtime/device_guard.h"
#include "infini_train/include/core/device_guard.h"
#include "infini_train/include/dispatcher.h"
#include "infini_train/include/tensor.h"

#include "infini_train/src/core/runtime/cuda/cuda_runtime_common.h"
#include "infini_train/src/core/cuda/cuda_stream.h"

namespace infini_train::kernels::cuda {
namespace {
using namespace infini_train::common::cuda;
constexpr int kWarpSize = 32;

template <typename T, int VecSize>
struct alignas(sizeof(T) * VecSize) AlignedVector {
T val[VecSize];
};

template <typename T>
__device__ __host__ inline bool IsAligned(const T* ptr, int vec_size) {
return reinterpret_cast<uintptr_t>(ptr) % (sizeof(T) * vec_size) == 0;
}

template <typename T, typename Func>
__global__ void UnaryForwardKernel(T *output, Func fn, size_t num_elements, size_t offset, const T *input) {
size_t idx = blockIdx.x * blockDim.x + threadIdx.x + offset;
Expand Down