Added C API for getting GEMM temp buffer size

1 files changed, 86 insertions, 0 deletions

diff --git a/include/clblast_c.h b/include/clblast_c.h
index a00aca45..051871ce 100644
--- a/include/clblast_c.h
+++ b/include/clblast_c.h
@@ -96,6 +96,7 @@ typedef enum CLBlastStatusCode_ {
   CLBlastInsufficientMemoryY       = -1007, // Vector Y's OpenCL buffer is too small
 
   // Custom additional status codes for CLBlast
+  CLBlastInsufficientMemoryTemp    = -2050, // Temporary buffer provided to GEMM routine is too small
   CLBlastInvalidBatchCount         = -2049, // The batch count needs to be positive
   CLBlastInvalidOverrideKernel     = -2048, // Trying to override parameters for an invalid kernel
   CLBlastMissingOverrideParameter  = -2047, // Missing override parameter(s) for the target kernel
@@ -1536,6 +1537,91 @@ CLBlastStatusCode PUBLIC_API CLBlastHgemmStridedBatched(const CLBlastLayout layo
                                                         cl_command_queue* queue, cl_event* event);
 
 // =================================================================================================
+// General matrix-matrix multiplication with temporary buffer from user (optional, for advanced users): SGEMM/DGEMM/CGEMM/ZGEMM/HGEMM
+CLBlastStatusCode PUBLIC_API CLBlastSgemmWithTempBuffer(const CLBlastLayout layout, const CLBlastTranspose a_transpose, const CLBlastTranspose b_transpose,
+                                          const size_t m, const size_t n, const size_t k,
+                                          const float 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 float beta,
+                                          cl_mem c_buffer, const size_t c_offset, const size_t c_ld,
+                                          cl_command_queue* queue, cl_event* event, cl_mem temp_buffer);
+CLBlastStatusCode PUBLIC_API CLBlastDgemmWithTempBuffer(const CLBlastLayout layout, const CLBlastTranspose a_transpose, const CLBlastTranspose b_transpose,
+                                          const size_t m, const size_t n, const size_t k,
+                                          const double 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 double beta,
+                                          cl_mem c_buffer, const size_t c_offset, const size_t c_ld,
+                                          cl_command_queue* queue, cl_event* event, cl_mem temp_buffer);
+CLBlastStatusCode PUBLIC_API CLBlastCgemmWithTempBuffer(const CLBlastLayout layout, const CLBlastTranspose a_transpose, const CLBlastTranspose b_transpose,
+                                          const size_t m, const size_t n, const size_t k,
+                                          const cl_float2 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 cl_float2 beta,
+                                          cl_mem c_buffer, const size_t c_offset, const size_t c_ld,
+                                          cl_command_queue* queue, cl_event* event, cl_mem temp_buffer);
+CLBlastStatusCode PUBLIC_API CLBlastZgemmWithTempBuffer(const CLBlastLayout layout, const CLBlastTranspose a_transpose, const CLBlastTranspose b_transpose,
+                                          const size_t m, const size_t n, const size_t k,
+                                          const cl_double2 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 cl_double2 beta,
+                                          cl_mem c_buffer, const size_t c_offset, const size_t c_ld,
+                                          cl_command_queue* queue, cl_event* event, cl_mem temp_buffer);
+CLBlastStatusCode PUBLIC_API CLBlastHgemmWithTempBuffer(const CLBlastLayout layout, const CLBlastTranspose a_transpose, const CLBlastTranspose b_transpose,
+                                          const size_t m, const size_t n, const size_t k,
+                                          const cl_half 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 cl_half beta,
+                                          cl_mem c_buffer, const size_t c_offset, const size_t c_ld,
+                                          cl_command_queue* queue, cl_event* event, cl_mem temp_buffer);
+
+// =================================================================================================
+// Retrieves the required size of the temporary buffer for the GEMM kernel: SGEMM/DGEMM/CGEMM/ZGEMM/HGEMM (optional)
+CLBlastStatusCode PUBLIC_API CLBlastSGemmTempBufferSize(const CLBlastLayout layout, const CLBlastTranspose a_transpose, const CLBlastTranspose b_transpose,
+                               const size_t m, const size_t n, const size_t k,
+                               const size_t a_offset, const size_t a_ld,
+                               const size_t b_offset, const size_t b_ld,
+                               const size_t c_offset, const size_t c_ld,
+                               cl_command_queue* queue,
+							   size_t* temp_buffer_size);
+
+CLBlastStatusCode PUBLIC_API CLBlastDGemmTempBufferSize(const CLBlastLayout layout, const CLBlastTranspose a_transpose, const CLBlastTranspose b_transpose,
+                               const size_t m, const size_t n, const size_t k,
+                               const size_t a_offset, const size_t a_ld,
+                               const size_t b_offset, const size_t b_ld,
+                               const size_t c_offset, const size_t c_ld,
+                               cl_command_queue* queue,
+							   size_t* temp_buffer_size);
+
+CLBlastStatusCode PUBLIC_API CLBlastCGemmTempBufferSize(const CLBlastLayout layout, const CLBlastTranspose a_transpose, const CLBlastTranspose b_transpose,
+                               const size_t m, const size_t n, const size_t k,
+                               const size_t a_offset, const size_t a_ld,
+                               const size_t b_offset, const size_t b_ld,
+                               const size_t c_offset, const size_t c_ld,
+                               cl_command_queue* queue,
+							   size_t* temp_buffer_size);
+
+CLBlastStatusCode PUBLIC_API CLBlastZGemmTempBufferSize(const CLBlastLayout layout, const CLBlastTranspose a_transpose, const CLBlastTranspose b_transpose,
+                               const size_t m, const size_t n, const size_t k,
+                               const size_t a_offset, const size_t a_ld,
+                               const size_t b_offset, const size_t b_ld,
+                               const size_t c_offset, const size_t c_ld,
+                               cl_command_queue* queue,
+							   size_t* temp_buffer_size);
+
+CLBlastStatusCode PUBLIC_API CLBlastHGemmTempBufferSize(const CLBlastLayout layout, const CLBlastTranspose a_transpose, const CLBlastTranspose b_transpose,
+                               const size_t m, const size_t n, const size_t k,
+                               const size_t a_offset, const size_t a_ld,
+                               const size_t b_offset, const size_t b_ld,
+                               const size_t c_offset, const size_t c_ld,
+                               cl_command_queue* queue,
+							   size_t* temp_buffer_size);
+
+// =================================================================================================
 
 // CLBlast stores binaries of compiled kernels into a cache in case the same kernel is used later on
 // for the same device. This cache can be cleared to free up system memory or in case of debugging.