Commits · 61238f405dfe8b8e2ade8fe5b8f541bb8075a1e9 · Kaushik Kulkarni / viennacl-dev · GitLab

Aug 20, 2013
- Generator : added basic OpenCL to CUDA kernel translation. Can now dump valid CUDA kernel code · 61238f40
  Philippe Tillet authored Aug 20, 2013
  
  61238f40
- Matrix elementwise operations no longer taken from auxiliary/, but part of the source tree. · 0326e73e
  Karl Rupp authored Aug 20, 2013
  
  0326e73e
- Generator : Fixed type safety when passing some arguments · 637cec0d
  Philippe Tillet authored Aug 20, 2013
  
  637cec0d
Aug 19, 2013
- OpenCL kernel enqueue no longer tries smaller work group sizes, but instead errors out directly. · 43299944
  Karl Rupp authored Aug 19, 2013
  43299944
- Added support for integer matrices for the host and CUDA. · be21f86c
  Karl Rupp authored Aug 19, 2013
  be21f86c
- matrix_expression: Unified code base with vector_expression. Also removed a spurious debug output. · 582ac014
  Karl Rupp authored Aug 19, 2013
  582ac014
- Generator : explicit cast of strides, starts to cl_uint · 0155922a
  Philippe Tillet authored Aug 20, 2013
  0155922a
- Autotuner : added light_tuning parameter for gemm · d0a0f280
  Philippe Tillet authored Aug 20, 2013
  
  d0a0f280
- Removed remaining traces of matrix_kernel_class_deducer.hpp · 4ad86951
  Karl Rupp authored Aug 19, 2013
  4ad86951
Aug 17, 2013
- OpenCL: Matrix kernels now embedded in source tree (except element-wise kernels). · 47becc33
  Karl Rupp authored Aug 16, 2013
  47becc33
Aug 16, 2013
- Scheduler: Added sparse matrix includes. · 25877124
  Karl Rupp authored Aug 16, 2013
  25877124
- Generator : Column-Major+no_trans now grabs the profile of Row-Major + trans · 04c75f32
  Philippe Tillet authored Aug 16, 2013
  
  04c75f32
- Scheduler: Added support for sparse matrices. · acbfe3d7
  Karl Rupp authored Aug 16, 2013
  acbfe3d7
- Tests: Fixed check for failed file read in sparse-test · 78430384
  Karl Rupp authored Aug 16, 2013
  
  78430384
- Scheduler: Now handling += and -= correctly for matrix-vector product. · 9c3f8810
  Karl Rupp authored Aug 15, 2013
  
  9c3f8810
- Generator : Improvement in the BLAS3 kernel template · 9283fddd
  Philippe Tillet authored Aug 16, 2013
  9283fddd
- Scheduler: Added support for matrix-matrix products. · af0b5370
  Karl Rupp authored Aug 15, 2013
  af0b5370
Aug 15, 2013
- Generator: Bugfixed default vendor profiles and replaced raw pointers by shared_ptr · 472c06e8
  Philippe Tillet authored Aug 16, 2013
  
  472c06e8
- Scheduler: Added support for matrix-vector multiplications. · d1aa35b8
  Karl Rupp authored Aug 15, 2013
  d1aa35b8
- Unified operator<< for matrices. · 8c20fce4
  Karl Rupp authored Aug 15, 2013
  8c20fce4
- Fixed a bug in host-based matrix-vector multiplications. · be9cdc1b
  Karl Rupp authored Aug 15, 2013
  be9cdc1b
- Unlocked LHS Storage and RHS Storage in the autotuner... · 197a707e
  Philippe Tillet authored Aug 15, 2013
  197a707e
- Updated AMD profiles · cdf7ea64
  Philippe Tillet authored Aug 15, 2013
  
  cdf7ea64
- Fixed typos in device_utils.hpp · b48e8037
  Philippe Tillet authored Aug 15, 2013
  
  b48e8037
Aug 14, 2013
- Fixed warnings when using -Wall -pedantic -Wextra on GCC and Clang · e03dae39
  Karl Rupp authored Aug 14, 2013
  
  e03dae39
- Autotuner : Made the tuning names more explicit. · 16c473f4
  Philippe Tillet authored Aug 14, 2013
  
  16c473f4
- Scheduler: Now supporting unary element-wise operations on dense matrices. · 8ecb79b9
  Karl Rupp authored Aug 14, 2013
  
  8ecb79b9
- Autotuner : device name converted to lower case in the output file name. · b074cfa0
  Philippe Tillet authored Aug 14, 2013
  
  b074cfa0
- Autotuner : fixed output filename of the BLAS3 tuning procedure · 26b41252
  Philippe Tillet authored Aug 14, 2013
  
  26b41252
- Polished the autotuner · cf229fd4
  Philippe Tillet authored Aug 14, 2013
  
  cf229fd4
- Scheduler: Added test for element_prod() and element_div() on matrices · a18a874c
  Karl Rupp authored Aug 14, 2013
  
  a18a874c
- Scheduler: Support for unary elementwise operations on vectors. · 988855ae
  Karl Rupp authored Aug 14, 2013
  988855ae
- Polished autotuner's output names · 5a8085ab
  Philippe Tillet authored Aug 14, 2013
  
  5a8085ab
- Removed the "#include CL/cl.hpp" from the .cpp's header. · 057b7aa4
  Philippe Tillet authored Aug 14, 2013
  057b7aa4
- Fixed conditional jump depending on an uninitialized member in ocl::device · ceb8ae7b
  Karl Rupp authored Aug 13, 2013
  
  ceb8ae7b
- Improved generated kernel name. · 0030c00c
  Philippe Tillet authored Aug 14, 2013
  0030c00c
- Polished and bugfixed the autotuner · 72114602
  Philippe Tillet authored Aug 14, 2013
  
  72114602
- Added flexibility on indentation in ocl::device::info() and ocl::device::full_info() · 264b0261
  Philippe Tillet authored Aug 14, 2013
  
  264b0261
- Autotuner now skips non-unrolled kernel on NVidia and unrolled-kernel on AMD. · 5c083388
  Philippe Tillet authored Aug 14, 2013
  
  5c083388
- Scheduler: Added support for elementwise vector operations. · 156bc9d3
  Karl Rupp authored Aug 13, 2013
  
  156bc9d3