diff options
author | Cedric Nugteren <web@cedricnugteren.nl> | 2017-03-08 20:10:20 +0100 |
---|---|---|
committer | Cedric Nugteren <web@cedricnugteren.nl> | 2017-03-08 20:10:20 +0100 |
commit | fa0a9c689fc21a2a24aeadf82ae0acdf6d8bf831 (patch) | |
tree | 404e85900a4c9038d407addb38798d06bb48868c /include | |
parent | 6aba0bbae71702c4eebd88d0fe17739b509185c1 (diff) |
Make batched routines based on offsets instead of a vector of cl_mem objects - undoing many earlier changes
Diffstat (limited to 'include')
-rw-r--r-- | include/clblast.h | 4 | ||||
-rw-r--r-- | include/clblast_c.h | 20 |
2 files changed, 12 insertions, 12 deletions
diff --git a/include/clblast.h b/include/clblast.h index f3f73893..a1f14471 100644 --- a/include/clblast.h +++ b/include/clblast.h @@ -614,8 +614,8 @@ StatusCode Omatcopy(const Layout layout, const Transpose a_transpose, template <typename T> StatusCode AxpyBatched(const size_t n, const T *alphas, - const cl_mem *x_buffers, const size_t x_inc, - cl_mem *y_buffers, const size_t y_inc, + const cl_mem x_buffer, const size_t *x_offsets, const size_t x_inc, + cl_mem y_buffer, const size_t *y_offsets, const size_t y_inc, const size_t batch_count, cl_command_queue* queue, cl_event* event = nullptr); diff --git a/include/clblast_c.h b/include/clblast_c.h index 5c84b5d7..4f21ba17 100644 --- a/include/clblast_c.h +++ b/include/clblast_c.h @@ -1331,32 +1331,32 @@ CLBlastStatusCode PUBLIC_API CLBlastHomatcopy(const CLBlastLayout layout, const // Batched version of AXPY: SAXPYBATCHED/DAXPYBATCHED/CAXPYBATCHED/ZAXPYBATCHED/HAXPYBATCHED CLBlastStatusCode PUBLIC_API CLBlastSaxpyBatched(const size_t n, const float *alphas, - const cl_mem *x_buffers, const size_t x_inc, - cl_mem *y_buffers, const size_t y_inc, + const cl_mem x_buffer, const size_t *x_offsets, const size_t x_inc, + cl_mem y_buffer, const size_t *y_offsets, const size_t y_inc, const size_t batch_count, cl_command_queue* queue, cl_event* event); CLBlastStatusCode PUBLIC_API CLBlastDaxpyBatched(const size_t n, const double *alphas, - const cl_mem *x_buffers, const size_t x_inc, - cl_mem *y_buffers, const size_t y_inc, + const cl_mem x_buffer, const size_t *x_offsets, const size_t x_inc, + cl_mem y_buffer, const size_t *y_offsets, const size_t y_inc, const size_t batch_count, cl_command_queue* queue, cl_event* event); CLBlastStatusCode PUBLIC_API CLBlastCaxpyBatched(const size_t n, const cl_float2 *alphas, - const cl_mem *x_buffers, const size_t x_inc, - cl_mem *y_buffers, const size_t y_inc, + const cl_mem x_buffer, const size_t *x_offsets, const size_t x_inc, + cl_mem y_buffer, const size_t *y_offsets, const size_t y_inc, const size_t batch_count, cl_command_queue* queue, cl_event* event); CLBlastStatusCode PUBLIC_API CLBlastZaxpyBatched(const size_t n, const cl_double2 *alphas, - const cl_mem *x_buffers, const size_t x_inc, - cl_mem *y_buffers, const size_t y_inc, + const cl_mem x_buffer, const size_t *x_offsets, const size_t x_inc, + cl_mem y_buffer, const size_t *y_offsets, const size_t y_inc, const size_t batch_count, cl_command_queue* queue, cl_event* event); CLBlastStatusCode PUBLIC_API CLBlastHaxpyBatched(const size_t n, const cl_half *alphas, - const cl_mem *x_buffers, const size_t x_inc, - cl_mem *y_buffers, const size_t y_inc, + const cl_mem x_buffer, const size_t *x_offsets, const size_t x_inc, + cl_mem y_buffer, const size_t *y_offsets, const size_t y_inc, const size_t batch_count, cl_command_queue* queue, cl_event* event); |