Uploaded image for project: 'SystemDS'
  1. SystemDS
  2. SYSTEMDS-446

Phase 1: Exploit GPU BLAS libraries (integration)

    XMLWordPrintableJSON

Details

    • Epic
    • Status: Reopened
    • Major
    • Resolution: Unresolved
    • SystemML 0.13, SystemML 0.14
    • SystemML 1.1
    • Compiler, Runtime, Test
    • None

    Attachments

      1.
      		 Implement functionality to transfer CP sparse matrixblock to GPU (and back) Sub-task Closed Nakul Jindal
      2.
      		 Implement GPU sparse matrix multiplication Sub-task Closed Nakul Jindal
      3.
      		 Add bufferpool integration logic to CUDA backend Sub-task Closed Niketan Pansare
      4.
      		 Implement GPU dense matrix multiplication Sub-task Closed Niketan Pansare
      5.
      		 Add GPU instructions that utilizes CuDNN v4's conv2d and pooling related functions Sub-task Closed Unassigned
      6.
      		 Implement MMTSJGPUInstruction instruction for GPU backend along with corresponding Hops/Lops Sub-task Closed Tanuj Kr Aasawat
      7.
      		 Error while allocating CSRPointer Sub-task Resolved Nakul Jindal
      8.
      		 Add support for cusparse geam Sub-task Closed Nakul Jindal
      9.
      		 Add support for cusparse axpy Sub-task Closed Nakul Jindal
      10.
      		 Improve the performance of sparse TSMM either by using/implement sparse dsyrk Sub-task Open Nakul Jindal
      11.
      		 LibMatrixCUDA's vectorScalarMultiply() produces incorrect results. Sub-task Open Unassigned
      12.
      		 Make sparse memory estimation robust by handling unknown nnz. Sub-task Open Nakul Jindal
      13.
      		 Conduct initial performance experiments for mat mult Sub-task Closed Nakul Jindal
      14.
      		 Enable setting GPU from MLContext (and related APIs) Sub-task Closed Nakul Jindal
      15.
      		 Create documentation explaining setup/usage for the GPU backend Sub-task Closed Niketan Pansare
      16.
      		 Add LU and QR functionality to GPU backend Sub-task Open Nakul Jindal
      17.
      		 Implement solve builtin function using cublas kernels Sub-task Closed Nakul Jindal
      18.
      		 Add support for aggregate unary operations on GPU Sub-task Closed Unassigned
      19.
      		 Implement relu_maxpooling instruction for GPU Sub-task Closed Niketan Pansare
      20.
      		 Implement conv2d_bias_add instruction for GPU Sub-task Closed Niketan Pansare
      21.
      		 Implement conv2d_bias_add instruction for GPU Sub-task Closed Niketan Pansare
      22.
      		 Implement Mathematical and Trigonometric Built-In Functions on GPU Sub-task Closed Nakul Jindal
      23.
      		 Add support for matrix-vector GPU axpy operation Sub-task Resolved Niketan Pansare
      24.
      		 Support GPU via Python APIs Sub-task Resolved Niketan Pansare
      25.
      		 Support alternative algorithms for CuDNN operators such as convolution Sub-task Open Niketan Pansare
      26.
      		 Support fused weight update operators (similar to codegen on CP) Sub-task Open Unassigned
      27.
      		 Add (Unit) Tests for GPU functions Sub-task Closed Nakul Jindal
      28.
      		 Fix the need to add force to -gpu always Sub-task Closed Nakul Jindal
      29.
      		 Add additional binary element wise operations Sub-task Closed Nakul Jindal
      30.
      		 Add relational operators for GPU Sub-task Closed Nakul Jindal
      31.
      		 Add cbind (and rbind) GPU ops Sub-task Closed Nakul Jindal
      32.
      		 Support left and right indexing on GPU Sub-task Open Niketan Pansare

      Activity

        People

          niketanpansare Niketan Pansare
          mboehm7 Matthias Boehm
          Votes:
          0 Vote for this issue
          Watchers:
          5 Start watching this issue

          Dates

            Created:
            Updated: