Merge pull request #32 from dan-zheng/cudnn-conv2d

Implement cuDNN conv2D grad.
feiwang3311 · Oct 19, 2018 · 663632d · 663632d
2 parents ceab5ee + 535c1bb
commit 663632d
Show file tree

Hide file tree

Showing 3 changed files with 397 additions and 127 deletions.
diff --git a/src/main/scala/lantern/DslAPI.scala b/src/main/scala/lantern/DslAPI.scala
@@ -497,6 +497,10 @@ trait DslGenCublas extends DslGenBase {
       |  data[index] = value;
       |}
       |
+      |__global__ void arrayFill(float *data, float value) {
+      |  int tid = threadIdx.x + blockIdx.x * blockDim.x;
+      |  data[tid] = value;
+      |}
       |
       |// From: https://github.com/pytorch/pytorch/blob/master/aten/src/THC/THCIntegerDivider.cuh
       |// Result of div/mod operation stored together.