未验证 提交 430e5590 编写于 作者: G Guo Sheng 提交者: GitHub

Merge pull request #15611 from guoshengCS/fix-beam-search-size

To make CUDA_LAUNCH_KERNEL_HELPER support large size.
...@@ -53,10 +53,12 @@ inline static int RoundToPowerOfTwo(int dim) { ...@@ -53,10 +53,12 @@ inline static int RoundToPowerOfTwo(int dim) {
__VA_ARGS__; \ __VA_ARGS__; \
} break } break
#define CUDA_LAUNCH_KERNEL_HELPER(...) \ #define CUDA_LAUNCH_KERNEL_HELPER(...) \
CUDA_LAUNCH_KERNEL_BASE(256, ##__VA_ARGS__); \ CUDA_LAUNCH_KERNEL_BASE(1024, ##__VA_ARGS__); \
CUDA_LAUNCH_KERNEL_BASE(128, ##__VA_ARGS__); \ CUDA_LAUNCH_KERNEL_BASE(512, ##__VA_ARGS__); \
CUDA_LAUNCH_KERNEL_BASE(64, ##__VA_ARGS__); \ CUDA_LAUNCH_KERNEL_BASE(256, ##__VA_ARGS__); \
CUDA_LAUNCH_KERNEL_BASE(128, ##__VA_ARGS__); \
CUDA_LAUNCH_KERNEL_BASE(64, ##__VA_ARGS__); \
CUDA_LAUNCH_KERNEL_BASE(32, ##__VA_ARGS__); CUDA_LAUNCH_KERNEL_BASE(32, ##__VA_ARGS__);
template <typename T> template <typename T>
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册