diff options
Diffstat (limited to 'include')
| -rw-r--r-- | include/clblast.h | 11 | ||||
| -rw-r--r-- | include/internal/routines/xher2k.h | 48 |
2 files changed, 59 insertions, 0 deletions
diff --git a/include/clblast.h b/include/clblast.h index d7c902d9..ef279fe5 100644 --- a/include/clblast.h +++ b/include/clblast.h @@ -161,6 +161,17 @@ StatusCode Syr2k(const Layout layout, const Triangle triangle, const Transpose a cl_mem c_buffer, const size_t c_offset, const size_t c_ld, cl_command_queue* queue, cl_event* event); +// Templated-precision rank-2K update of a hermitian matrix: CHER2K/ZHER2K +template <typename T, typename U> +StatusCode Her2k(const Layout layout, const Triangle triangle, const Transpose ab_transpose, + const size_t n, const size_t k, + const T alpha, + const cl_mem a_buffer, const size_t a_offset, const size_t a_ld, + const cl_mem b_buffer, const size_t b_offset, const size_t b_ld, + const U beta, + cl_mem c_buffer, const size_t c_offset, const size_t c_ld, + cl_command_queue* queue, cl_event* event); + // Templated-precision triangular matrix-matrix multiplication: STRMM/DTRMM/CTRMM/ZTRMM template <typename T> StatusCode Trmm(const Layout layout, const Side side, const Triangle triangle, diff --git a/include/internal/routines/xher2k.h b/include/internal/routines/xher2k.h new file mode 100644 index 00000000..1836a812 --- /dev/null +++ b/include/internal/routines/xher2k.h @@ -0,0 +1,48 @@ + +// ================================================================================================= +// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This +// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max- +// width of 100 characters per line. +// +// Author(s): +// Cedric Nugteren <www.cedricnugteren.nl> +// +// This file implements the Xher2k routine. The precision is implemented using the template argument +// 'T', whereas the alpha/beta arguments are of type 'U'. The implementation is very similar to the +// Xsyr2k routine. +// +// ================================================================================================= + +#ifndef CLBLAST_ROUTINES_XHER2K_H_ +#define CLBLAST_ROUTINES_XHER2K_H_ + +#include "internal/routine.h" + +namespace clblast { +// ================================================================================================= + +// See comment at top of file for a description of the class +template <typename T, typename U> +class Xher2k: public Routine { + public: + Xher2k(CommandQueue &queue, Event &event); + + // Templated-precision implementation of the routine + StatusCode DoHer2k(const Layout layout, const Triangle triangle, const Transpose ab_transpose, + const size_t n, const size_t k, + const T alpha, + const Buffer &a_buffer, const size_t a_offset, const size_t a_ld, + const Buffer &b_buffer, const size_t b_offset, const size_t b_ld, + const U beta, + const Buffer &c_buffer, const size_t c_offset, const size_t c_ld); + + private: + // Static variable to get the precision + const static Precision precision_; +}; + +// ================================================================================================= +} // namespace clblast + +// CLBLAST_ROUTINES_XHER2K_H_ +#endif |