Commits · 54bf0c38beffc93732600d7870edfad58e74ed72 · Kaushik Kulkarni / viennacl-dev · GitLab

May 23, 2015
- Scan: Fixed incomplete migration of OpenCL kernels from SVD module. · 54bf0c38
  Karl Rupp authored May 23, 2015
  
  54bf0c38
May 22, 2015
- Scan: Refurbished CUDA and OpenCL implementations. · aadb5b72
  Karl Rupp authored May 22, 2015
  aadb5b72
May 21, 2015
- SpGEMM: Fixed missing barrier in OpenCL kernels. · 525fc3ae
  Karl Rupp authored May 21, 2015
  525fc3ae
May 20, 2015
- Scan: Added fast OpenMP implementation, fixed OpenCL bug, shifted to vectors. · 54474134
  Karl Rupp authored May 20, 2015
  54474134
- SpGEMM: Minor improvements to OpenCL kernel. · 6347fe4d
  Karl Rupp authored May 20, 2015
  6347fe4d
May 10, 2015
- Merge branch 'karlrupp/sparse-matrix-matrix-product' · 66a8949c
  Karl Rupp authored May 10, 2015
  66a8949c
- SpGEMM: Switched back to dynamic scheduling with OpenMP. · b3e5daa0
  Karl Rupp authored May 10, 2015
  b3e5daa0
May 07, 2015
- SpGEMM: Fixed bug in OpenCL kernel. · f0e57ab7
  Karl Rupp authored May 07, 2015
  f0e57ab7
- compressed_matrix: Improved handling of dimensions for STL matrix. · 0cfd0263
  Karl Rupp authored May 07, 2015
  0cfd0263
- SpGEMM: Fixed warnings and added asserts() to check for proper dimensions. · 5bf86cc7
  Karl Rupp authored May 07, 2015
  
  5bf86cc7
- SpGEMM: Added OpenCL implementation of RMerge. · d1685c1e
  Karl Rupp authored May 07, 2015
  d1685c1e
Apr 27, 2015
- custom-cuda: Fixed incorrect passing of arguments to CUDA kernel. · f29e01e0
  Karl Rupp authored Apr 27, 2015
  f29e01e0
Apr 18, 2015
- SpGEMM: Added AVX2 to numerical stage. · fa7ba2fe
  Karl Rupp authored Apr 18, 2015
  
  fa7ba2fe
- SpGEMM: Broader use of AVX2 in symbolic stage. · c78438cf
  Karl Rupp authored Apr 18, 2015
  
  c78438cf
- SpGEMM: Now merging two rows from B before merging into scratchpad. · 278e37b3
  Karl Rupp authored Apr 18, 2015
  278e37b3
- SpGEMM: Using smaller scratchpad buffers to increase cache hit rates. · 8e4805b2
  Karl Rupp authored Apr 18, 2015
  8e4805b2
- AVX2: Allocating memory via aligned_alloc with 32-byte alignment. · b0607df1
  Karl Rupp authored Apr 18, 2015
  b0607df1
- SpGEMM: Added RMerge-like factorization if nonzeros in A are above 1024. · 764230c7
  Karl Rupp authored Apr 18, 2015
  764230c7
- SpGEMM: Added missing optimizations for single row merge in RMerge. · 616f1e1c
  Karl Rupp authored Apr 18, 2015
  616f1e1c
- SpGEMM: Fixed index bug in CUDA kernel for RMerge. · cb3a8546
  Karl Rupp authored Apr 18, 2015
  cb3a8546
Apr 17, 2015
- SpGEMM: Completed RMerge implementation. · a8ebd2f1
  Karl Rupp authored Apr 18, 2015
  
  a8ebd2f1
- SpGEMM: Added RMerge implementation. · 3943a3b6
  Karl Rupp authored Apr 17, 2015
  3943a3b6
- SpGEMM: Reduced memory footprint of CUDA implementation. · efcfd827
  Karl Rupp authored Apr 17, 2015
  efcfd827
- IO: Fixed warning in Matrix Market reader. · b6f77e61
  Karl Rupp authored Apr 17, 2015
  b6f77e61
- SpGEMM: Added profiling information to CUDA implementation, using __ldg() · 50668951
  Karl Rupp authored Apr 17, 2015
  50668951
- SpGEMM: Changed atomic write section to critical section. · 3b9df078
  Karl Rupp authored Apr 17, 2015
  3b9df078
Apr 16, 2015
- MatrixMarket: Enabled reading of complex matrices. · b41a6b97
  Karl Rupp authored Apr 16, 2015
  b41a6b97
Apr 15, 2015
- SpGEMM: Finished first version of refined RMerge kernel. · 095d73bc
  Karl Rupp authored Apr 15, 2015
  095d73bc
- SpGEMM: Fixed warning for host-based SpGEMM if not using OpenMP. · 89f67fc4
  Karl Rupp authored Apr 15, 2015
  89f67fc4
- SpGEMM: Replaced use of Boost.uBLAS in tests with STL. · 6f5e8f40
  Karl Rupp authored Apr 15, 2015
  6f5e8f40
Apr 14, 2015
- SpGEMM: Changed tasking flag from bool to int to address complaints by ICC. · ce99b213
  Karl Rupp authored Apr 15, 2015
  ce99b213
- SpGEMM: Implemented OpenMP task generation per thread. · 4fd6e1e1
  Karl Rupp authored Apr 15, 2015
  4fd6e1e1
- SpGEMM: Removed unnecessary writes to temporary buffers in vector merge. · 17e09bb9
  Karl Rupp authored Apr 14, 2015
  17e09bb9
- Memory: Added OpenMP-parallelization to memory transfer operations. · 52ecbc09
  Karl Rupp authored Apr 14, 2015
  52ecbc09
Apr 13, 2015
- MatrixMarket: Added support for pattern matrices. · 3b8a3d56
  Karl Rupp authored Apr 13, 2015
  3b8a3d56
Apr 12, 2015
- SpGEMM: Reduced serial components by moving some work to parallel section. · 1c66ff1f
  Karl Rupp authored Apr 12, 2015
  1c66ff1f
Apr 11, 2015
- SpGEMM: Improved OpenMP work estimation and reduction for aggregating overall work. · e257d83b
  Karl Rupp authored Apr 11, 2015
  
  e257d83b
- SpGEMM: Improved OpenMP distribution of work by using work estimates. · 128f17d1
  Karl Rupp authored Apr 11, 2015
  128f17d1
- SpGEMM: Added option for printing profiling information by setting VIENNACL_WITH_SPGEMM_TIMINGS. · 1de47f39
  Karl Rupp authored Apr 11, 2015
  1de47f39
- SpGEMM: Simplified and (hopefully) optimized row-by-row merging, disabled vector-merge · 1544f1d0
  Karl Rupp authored Apr 11, 2015
  1544f1d0