From a2e726d3bd4294f1eae1735f6ba23105dccc6b10 Mon Sep 17 00:00:00 2001 From: CNugteren Date: Mon, 14 Sep 2015 16:57:00 +0200 Subject: Added xDOT/xDOTU/xDOTC dot-product routines --- include/clblast.h | 26 ++++++++++++++++++++++++++ 1 file changed, 26 insertions(+) (limited to 'include/clblast.h') diff --git a/include/clblast.h b/include/clblast.h index 326c9ec2..72825e0b 100644 --- a/include/clblast.h +++ b/include/clblast.h @@ -68,6 +68,8 @@ enum class StatusCode { kInvalidLocalMemUsage = -2046, // Not enough local memory available on this device kNoHalfPrecision = -2045, // Half precision (16-bits) not supported by the device kNoDoublePrecision = -2044, // Double precision (64-bits) not supported by the device + kInvalidVectorDot = -2043, // Vector dot is not a valid OpenCL buffer + kInsufficientMemoryDot = -2042, // Vector dot's OpenCL buffer is too small }; // Matrix layout and transpose types @@ -114,6 +116,30 @@ StatusCode Axpy(const size_t n, cl_mem y_buffer, const size_t y_offset, const size_t y_inc, cl_command_queue* queue, cl_event* event); +// Dot product of two vectors: SDOT/DDOT +template +StatusCode Dot(const size_t n, + cl_mem dot_buffer, const size_t dot_offset, + const cl_mem x_buffer, const size_t x_offset, const size_t x_inc, + const cl_mem y_buffer, const size_t y_offset, const size_t y_inc, + cl_command_queue* queue, cl_event* event); + +// Dot product of two complex vectors: CDOTU/ZDOTU +template +StatusCode Dotu(const size_t n, + cl_mem dot_buffer, const size_t dot_offset, + const cl_mem x_buffer, const size_t x_offset, const size_t x_inc, + const cl_mem y_buffer, const size_t y_offset, const size_t y_inc, + cl_command_queue* queue, cl_event* event); + +// Dot product of two complex vectors, one conjugated: CDOTC/ZDOTC +template +StatusCode Dotc(const size_t n, + cl_mem dot_buffer, const size_t dot_offset, + const cl_mem x_buffer, const size_t x_offset, const size_t x_inc, + const cl_mem y_buffer, const size_t y_offset, const size_t y_inc, + cl_command_queue* queue, cl_event* event); + // ================================================================================================= // BLAS level-2 (matrix-vector) routines // ================================================================================================= -- cgit v1.2.3