Commits · b0607df187193e45ca2e7bfd9f4b21c2bb69f11a · Kaushik Kulkarni / viennacl-dev · GitLab

Apr 18, 2015
- 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
Apr 10, 2015
- SpGEMM: Fixed incorrect address scaling and aligned vs. nonaligned load. · f719164a
  Karl Rupp authored Apr 11, 2015
  
  f719164a
- SpGEMM: Added AVX2-version for merging 8 rows simultaneously. · 016b1781
  Karl Rupp authored Apr 10, 2015
  016b1781
- SpGEMM: Disabling use of hash_map. · 5b2ee22f
  Karl Rupp authored Apr 10, 2015
  5b2ee22f
- SpGEMM: Added implementation for multi-row merging using vectors. · 41688809
  Karl Rupp authored Apr 10, 2015
  41688809
Apr 09, 2015
- SpGEMM: Fixed bug in hash-based implementation. · d97b6bbd
  Karl Rupp authored Apr 09, 2015
  d97b6bbd
- SpGEMM: Added hash-based implementation for large rows. · 7c635ae6
  Karl Rupp authored Apr 09, 2015
  7c635ae6
- SpGEMM: Fixed a bug for the case that the row in A has only one entry. · d5618962
  Karl Rupp authored Apr 09, 2015
  d5618962
- SpGEMM: Replacing NDEBUG with BOOST_UBLAS_NDEBUG in test. · a088afea
  Karl Rupp authored Apr 09, 2015
  a088afea
Apr 08, 2015
- SpGEMM: Added CUDA first implementation. Needs further tweaks. · 3246d0d6
  Karl Rupp authored Apr 08, 2015
  3246d0d6
Apr 07, 2015
- SpGEMM: Rearranged code in host backend for better navigation. · d7846a11
  Karl Rupp authored Apr 07, 2015
  d7846a11
- SpGEMM: Now using an adaptive algorithm (vector vs. list) · a94082b9
  Karl Rupp authored Apr 07, 2015
  a94082b9
Apr 06, 2015
- compressed_matrix: Merging of two rows in a list completed. · eacd3d98
  Karl Rupp authored Apr 06, 2015
  eacd3d98
Apr 02, 2015
- Optimization of SpGEMM to load and merge multiple rows from B. · 34d878d6
  Karl Rupp authored Apr 02, 2015
  34d878d6
- SpGEMM: Attempt to introduce a list for faster row merging. · 90c8ee9b
  Karl Rupp authored Apr 02, 2015
  90c8ee9b
- compressed_matrix: Removed use of an auxiliary array in SpGEMM. · 09476768
  Karl Rupp authored Apr 02, 2015
  09476768
Apr 01, 2015
- compressed_matrix: Optimizing host-based sparse matrix-matrix product. · 87edf140
  Karl Rupp authored Apr 01, 2015
  87edf140