Skip to content

Adding missing __shfl overloads. #557

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

Closed
Jorghi12 wants to merge 2 commits into from
Closed

Adding missing __shfl overloads. #557

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f52b52b
Adding missing overloads.
Jorghi12 Jul 6, 2018
e062d9e
Update device_util.cpp
Jul 6, 2018
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
8 changes: 8 additions & 0 deletions include/hip/hcc_detail/hip_runtime.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -287,6 +287,10 @@ __device__ float __shfl(float input, int lane, int width = warpSize);
__device__ float __shfl_up(float input, unsigned int lane_delta, int width = warpSize);
__device__ float __shfl_down(float input, unsigned int lane_delta, int width = warpSize);
__device__ float __shfl_xor(float input, int lane_mask, int width = warpSize);
__device__ double __shfl(double input, int lane, int width = warpSize);
__device__ double __shfl_up(double input, unsigned int lane_delta, int width = warpSize);
__device__ double __shfl_down(double input, unsigned int lane_delta, int width = warpSize);
__device__ double __shfl_xor(double input, int lane_mask, int width = warpSize);
#else
__device__ int __shfl(int input, int lane, int width);
__device__ int __shfl_up(int input, unsigned int lane_delta, int width);
Expand All @@ -296,6 +300,10 @@ __device__ float __shfl(float input, int lane, int width);
__device__ float __shfl_up(float input, unsigned int lane_delta, int width);
__device__ float __shfl_down(float input, unsigned int lane_delta, int width);
__device__ float __shfl_xor(float input, int lane_mask, int width);
__device__ double __shfl(double input, int lane, int width);
__device__ double __shfl_up(double input, unsigned int lane_delta, int width);
__device__ double __shfl_down(double input, unsigned int lane_delta, int width);
__device__ double __shfl_xor(double input, int lane_mask, int width);
#endif //__cplusplus

__device__ unsigned __hip_ds_bpermute(int index, unsigned src);
Expand Down
16 changes: 15 additions & 1 deletion src/device_util.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -319,6 +319,20 @@ __device__ float __shfl_xor(float input, int lane_mask, int width) {
return hc::__shfl_xor(input, lane_mask, width);
}

__device__ double __shfl(double input, int lane, int width) { return hc::__shfl(input, lane, width); }

__device__ double __shfl_up(double input, unsigned int lane_delta, int width) {
return hc::__shfl_up(input, lane_delta, width);
}

__device__ double __shfl_down(double input, unsigned int lane_delta, int width) {
return hc::__shfl_down(input, lane_delta, width);
}

__device__ double __shfl_xor(double input, int lane_mask, int width) {
return hc::__shfl_xor(input, lane_mask, width);
}

__host__ __device__ int min(int arg1, int arg2) {
return (int)(hc::precise_math::fmin((float)arg1, (float)arg2));
}
Expand All @@ -333,4 +347,4 @@ __device__ void* __get_dynamicgroupbaseptr() {
__host__ void* __get_dynamicgroupbaseptr() { return nullptr; }


__device__ void __threadfence_system(void) { std::atomic_thread_fence(std::memory_order_seq_cst); }
__device__ void __threadfence_system(void) { std::atomic_thread_fence(std::memory_order_seq_cst); }