[SYSTEMML-445] Added sparse scalar-matrix arithmetic/relational operators
Project: http://git-wip-us.apache.org/repos/asf/systemml/repo Commit: http://git-wip-us.apache.org/repos/asf/systemml/commit/61139e40 Tree: http://git-wip-us.apache.org/repos/asf/systemml/tree/61139e40 Diff: http://git-wip-us.apache.org/repos/asf/systemml/diff/61139e40 Branch: refs/heads/master Commit: 61139e40052ea6591840a6925258a0dcbfdf0f8e Parents: 0e323ec Author: Niketan Pansare <npan...@us.ibm.com> Authored: Tue Sep 18 16:20:14 2018 -0700 Committer: Niketan Pansare <npan...@us.ibm.com> Committed: Tue Sep 18 16:22:06 2018 -0700 ---------------------------------------------------------------------- src/main/cpp/kernels/SystemML.cu | 41 + src/main/cpp/kernels/SystemML.ptx | 4053 +++++++++++++----- .../context/ExecutionContext.java | 2 +- .../instructions/gpu/context/CSRPointer.java | 20 + .../runtime/matrix/data/LibMatrixCUDA.java | 99 +- 5 files changed, 3026 insertions(+), 1189 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/systemml/blob/61139e40/src/main/cpp/kernels/SystemML.cu ---------------------------------------------------------------------- diff --git a/src/main/cpp/kernels/SystemML.cu b/src/main/cpp/kernels/SystemML.cu index b874cdd..a53d07a 100644 --- a/src/main/cpp/kernels/SystemML.cu +++ b/src/main/cpp/kernels/SystemML.cu @@ -772,6 +772,47 @@ extern "C" __global__ void matrix_scalar_op_f(float *A, double scalar, float *C, matrix_scalar_op(A, (float)scalar, C, size, op, isLeftScalar); } + +/** + * Performs sparse-dense arithmetic operation between a matrix and a scalar. + * C = s op A or C = A op s (where A is the matrix, s is the scalar and op is + * the operation) + * @param cooRowPtrA row pointers for input matrix allocated on GPU in coo format + * @param colPtrA col index pointers for input matrix allocated on GPU + * @param valA val array for input matrix allocated on GPU + * @param scalar scalar input + * @param C output matrix allocated on GPU + * @param nnz number of non-zero elements in matrix A + * @param colsA number of columns in matrix A + * @param op number code of the arithmetic operation to perform + * @param isLeftScalar whether the scalar is on the left side + */ +template <typename T> +__device__ void sparse_dense_matrix_scalar_op(int* cooRowPtrA, int* colPtrA, T *valA, T scalar, T *C, int nnz, int colsA, int op, + int isLeftScalar) { + int index = blockIdx.x * blockDim.x + threadIdx.x; + if (index < nnz) { + T inputVal = valA[index]; + int outIndex = cooRowPtrA[index]*colsA + colPtrA[index]; + if (isLeftScalar) { + C[outIndex] = binaryOp(scalar, inputVal, op); + } else { + C[outIndex] = binaryOp(inputVal, scalar, op); + } + } + __syncthreads(); +} + +extern "C" __global__ void sparse_dense_matrix_scalar_op_d(int* cooRowPtrA, int* colPtrA, double *valA, double scalar, double *C, + int nnz, int colsA, int op, int isLeftScalar) { + sparse_dense_matrix_scalar_op(cooRowPtrA, colPtrA, valA, scalar, C, nnz, colsA, op, isLeftScalar); +} + +extern "C" __global__ void sparse_dense_matrix_scalar_op_f(int* cooRowPtrA, int* colPtrA, float *valA, double scalar, float *C, + int nnz, int colsA, int op, int isLeftScalar) { + sparse_dense_matrix_scalar_op(cooRowPtrA, colPtrA, valA, (float) scalar, C, nnz, colsA, op, isLeftScalar); +} + /** * Sets all elements (fills) of a double array of given length with a given * scalar value
