Commit graph

407 commits

Author SHA1 Message Date
Cedric Nugteren bf84463ab2 Separated the GEMM kernel in two parts to reduce string length for MSVC 2016-02-08 20:06:02 +01:00
Cedric Nugteren 38c56bbde2 Split-up the XGEMV kernel in two parts 2016-02-08 19:43:34 +01:00
Cedric Nugteren 6f4b34f813 Added tuning parameters for various devices using the new database script 2016-02-07 16:41:09 +01:00
Cedric Nugteren 165a94c200 Various fixes to the database script 2016-02-07 16:39:37 +01:00
Cedric Nugteren 00be6f7530 Added dictionary with short and long OpenCL vendor names to fix issues with Intel having multiple names 2016-02-07 11:59:30 +01:00
Cedric Nugteren c76f1d9dbb Made the tuning database an optional external download 2016-02-07 10:59:51 +01:00
CNugteren 704a729f5c Made the database script compatible with Python 3 2016-02-06 13:11:36 +01:00
CNugteren b7900652b2 Reduced the maximum workgroup-size for GEMV kernels further 2016-02-06 13:07:19 +01:00
Cedric Nugteren bb985f010b Changed the order of tuners in the alltuners target 2016-02-06 12:48:42 +01:00
CNugteren 40346bb3a5 Reduced unrolling factor in xgemv kernel to reduce compilation times 2016-02-06 12:09:21 +01:00
CNugteren fbf071ba62 Fixed a linker error in the performance client under GCC 2016-02-06 10:53:44 +01:00
CNugteren 9622d3be22 Fixes for compilation under Visual Studio 2016-01-30 14:57:49 +01:00
Cedric Nugteren 44fb40e5c4 Prepared for MSVC support 2016-01-30 11:54:29 +01:00
Cedric Nugteren f573fe6bb3 Fixed a bug in the graph scripts (thanks to Victor Pakhomov) 2016-01-30 11:53:54 +01:00
Cedric Nugteren 310d05d187 Updated to version 4.0 of the CLCudaAPI header 2016-01-30 11:52:21 +01:00
Cedric Nugteren deb58709c8 Merge branch 'tuning_database' into development
is merge is necessary,
2016-01-30 11:46:45 +01:00
Cedric Nugteren 276e772a2c Added first auto-generated database headers from the Python database; only K40 and Iris supported now 2016-01-30 11:43:21 +01:00
Cedric Nugteren 76c9148030 Minor improvements to the database script, including proper file paths 2016-01-24 17:56:27 +01:00
Cedric Nugteren f0b3091cdb Added Python function to compute defaults for a particular device/vendor combination 2016-01-24 17:35:31 +01:00
Cedric Nugteren e4e3663e61 Updated FindOpenCL for Intel Linux OpenCL paths 2016-01-23 16:09:07 +01:00
CNugteren 09c94b17cf Added tuning data for Tesla K40 2015-10-28 21:20:42 +01:00
CNugteren c0d469718a Now sets local memory size in xgemv tuner properly 2015-10-28 21:19:59 +01:00
CNugteren bb4e78f737 Added initial tuning database with Intel Iris data 2015-10-25 16:49:59 +01:00
CNugteren ccd1a5c7cc Updated tuning database script according to the new JSON format 2015-10-25 16:49:29 +01:00
CNugteren 179ad0666d Fixed an arguments-related bug in the GEMV tuner 2015-10-25 16:48:26 +01:00
CNugteren a2d5d7770e Moved the tuner database script to a separate folder 2015-10-25 16:27:14 +01:00
CNugteren 9bf6be8426 Added alpha and beta to tuner meta-data 2015-10-23 11:01:44 +02:00
CNugteren 3f616366bd Prepared the changelog for the next release 2015-10-17 15:57:04 +02:00
CNugteren 92404035e8 Updated to version 0.5.0 2015-10-17 15:48:13 +02:00
CNugteren afb3e64fd3 Travis now also build the development branch 2015-10-17 15:42:45 +02:00
Cedric Nugteren 653feca564 Merge pull request #28 from CNugteren/kernels_reorganization
Kernels re-organization level-3
2015-10-17 15:30:06 +02:00
CNugteren 0d4091fdfb Added guards for routine-specific level-3 pad kernels 2015-10-13 08:29:45 +02:00
CNugteren f74c9a5640 Routine names are now all default arguments defined in the header 2015-10-12 08:35:58 +02:00
CNugteren 54a8723f8c Moved level3 kernel files to a subfolder 2015-10-12 08:28:40 +02:00
Cedric Nugteren 92b4b0d1fe Merge pull request #27 from CNugteren/level2_matrix_vector
Added many level-2 matrix-vector routines
2015-09-26 17:02:34 +02:00
CNugteren 2b56c2c603 Added TRMV/TBMV/TPMV routines 2015-09-26 16:58:03 +02:00
CNugteren 04d28b0420 Made buffer copying a const-method for the source 2015-09-26 16:48:11 +02:00
CNugteren de6547a92b Added SBMV and SPMV routines 2015-09-19 18:01:19 +02:00
CNugteren 80da67d28b Added the HPMV routine 2015-09-19 17:40:38 +02:00
CNugteren c32c4a9739 Added infrastructure for packed matrices 2015-09-19 17:37:42 +02:00
CNugteren aebd156869 Added the HBMV routine 2015-09-19 11:11:34 +02:00
CNugteren 93dddda63e Improved the organization and performance of level 2 routines 2015-09-18 17:46:41 +02:00
CNugteren 4507ba4997 Added first version of banded matrix-vector multiplication 2015-09-18 15:25:20 +02:00
Cedric Nugteren 42db8ea968 Merge pull request #26 from CNugteren/routine_definitions
Generated API interface and implementations
2015-09-18 10:23:16 +02:00
CNugteren 4796c9bcbd Added generated main functions for correctness/performance tests for level 2 routines 2015-09-18 10:19:03 +02:00
CNugteren 6105ad6f5b Added interface of all level 2 routines 2015-09-17 17:05:45 +02:00
CNugteren 6307d2e5db Added script to generate API interface and implementation automatically 2015-09-17 10:14:33 +02:00
CNugteren 1c24210026 Made Travis always build pushes to the master branch 2015-09-14 17:16:31 +02:00
Cedric Nugteren a2b773573d Merge pull request #25 from CNugteren/level1_routines
Added several level 1 routines
2015-09-14 17:12:23 +02:00
CNugteren 224c967584 Removed routines from the table which are not supported by clBLAS 2015-09-14 17:02:33 +02:00