Repository: incubator-systemml Updated Branches: refs/heads/master ee6bc8ce2 -> af93ca8a4
[MINOR] Cleanup of some comments Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo Commit: http://git-wip-us.apache.org/repos/asf/incubator-systemml/commit/af93ca8a Tree: http://git-wip-us.apache.org/repos/asf/incubator-systemml/tree/af93ca8a Diff: http://git-wip-us.apache.org/repos/asf/incubator-systemml/diff/af93ca8a Branch: refs/heads/master Commit: af93ca8a40befaaba79b0f96b4dbf8b8db85be13 Parents: ee6bc8c Author: Nakul Jindal <[email protected]> Authored: Mon Mar 27 14:39:31 2017 -0700 Committer: Nakul Jindal <[email protected]> Committed: Mon Mar 27 14:39:31 2017 -0700 ---------------------------------------------------------------------- .../runtime/matrix/data/LibMatrixCUDA.java | 25 ++++---------------- 1 file changed, 5 insertions(+), 20 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/af93ca8a/src/main/java/org/apache/sysml/runtime/matrix/data/LibMatrixCUDA.java ---------------------------------------------------------------------- diff --git a/src/main/java/org/apache/sysml/runtime/matrix/data/LibMatrixCUDA.java b/src/main/java/org/apache/sysml/runtime/matrix/data/LibMatrixCUDA.java index 8074e3a..23790c4 100644 --- a/src/main/java/org/apache/sysml/runtime/matrix/data/LibMatrixCUDA.java +++ b/src/main/java/org/apache/sysml/runtime/matrix/data/LibMatrixCUDA.java @@ -26,15 +26,16 @@ import jcuda.jcublas.cublasFillMode; import jcuda.jcublas.cublasHandle; import jcuda.jcublas.cublasOperation; import jcuda.jcudnn.cudnnActivationDescriptor; +import jcuda.jcudnn.cudnnBatchNormMode; import jcuda.jcudnn.cudnnConvolutionDescriptor; import jcuda.jcudnn.cudnnConvolutionFwdPreference; import jcuda.jcudnn.cudnnFilterDescriptor; import jcuda.jcudnn.cudnnHandle; import jcuda.jcudnn.cudnnPoolingDescriptor; +import jcuda.jcudnn.cudnnStatus; import jcuda.jcudnn.cudnnTensorDescriptor; import jcuda.jcusparse.JCusparse; import jcuda.jcusparse.cusparseHandle; - import org.apache.commons.logging.Log; import org.apache.commons.logging.LogFactory; import org.apache.sysml.api.DMLScript; @@ -88,6 +89,9 @@ import org.apache.sysml.utils.Statistics; import static jcuda.jcublas.cublasOperation.CUBLAS_OP_N; import static jcuda.jcublas.cublasOperation.CUBLAS_OP_T; import static jcuda.jcudnn.JCudnn.cudnnActivationForward; +import static jcuda.jcudnn.JCudnn.cudnnBatchNormalizationBackward; +import static jcuda.jcudnn.JCudnn.cudnnBatchNormalizationForwardInference; +import static jcuda.jcudnn.JCudnn.cudnnBatchNormalizationForwardTraining; import static jcuda.jcudnn.JCudnn.cudnnConvolutionBackwardData; import static jcuda.jcudnn.JCudnn.cudnnConvolutionBackwardFilter; import static jcuda.jcudnn.JCudnn.cudnnConvolutionForward; @@ -126,11 +130,6 @@ import static jcuda.runtime.cudaMemcpyKind.cudaMemcpyDeviceToHost; import static jcuda.runtime.cudaMemcpyKind.cudaMemcpyHostToDevice; import static org.apache.sysml.runtime.instructions.gpu.context.JCudaObject.allocate; import static org.apache.sysml.runtime.instructions.gpu.context.JCudaObject.cudaFreeHelper; -import jcuda.jcudnn.cudnnBatchNormMode; -import jcuda.jcudnn.cudnnStatus; -import static jcuda.jcudnn.JCudnn.cudnnBatchNormalizationForwardInference; -import static jcuda.jcudnn.JCudnn.cudnnBatchNormalizationForwardTraining; -import static jcuda.jcudnn.JCudnn.cudnnBatchNormalizationBackward; //FIXME move could to respective instructions, this is not a block library public class LibMatrixCUDA { @@ -2518,22 +2517,11 @@ public class LibMatrixCUDA { Pointer A = getDensePointer(out, instName); int rlen = (int) out.getNumRows(); int clen = (int) out.getNumColumns(); -// if(constant == 0) { -// out.getMatrixCharacteristics().setNonZeros(0); -// } -// else { -// out.getMatrixCharacteristics().setNonZeros(rlen*clen); -// } - // dense_matrix_set(double* A, double scalar, int rlen, int clen) - long t0=0; if (GPUStatistics.DISPLAY_STATISTICS) t0 = System.nanoTime(); int size = rlen * clen; kernels.launchKernel("fill", ExecutionConfig.getConfigForSimpleVectorOperations(size), A, constant, size); - // kernels.launchKernel("dense_matrix_set", - // ExecutionConfig.getConfigForSimpleMatrixOperations(rlen, clen), - // A, constant, rlen, clen); if (GPUStatistics.DISPLAY_STATISTICS) GPUStatistics.maintainCPMiscTimes(instName, GPUInstruction.MISC_TIMER_FILL_KERNEL, System.nanoTime() - t0); } @@ -2549,9 +2537,6 @@ public class LibMatrixCUDA { private static void deviceCopy(String instName, Pointer src, Pointer dest, int rlen, int clen) throws DMLRuntimeException { long t0=0; if (GPUStatistics.DISPLAY_STATISTICS) t0 = System.nanoTime(); - //kernels.launchKernel("dense_matrix_copy", - // ExecutionConfig.getConfigForSimpleMatrixOperations(rlen, clen), - // src, dest, rlen, clen); int size = rlen * clen * Sizeof.DOUBLE; cudaMemcpy(dest, src, size, cudaMemcpyDeviceToDevice); if (GPUStatistics.DISPLAY_STATISTICS) GPUStatistics.maintainCPMiscTimes(instName, GPUInstruction.MISC_TIMER_DEVICE_TO_DEVICE, System.nanoTime() - t0);
