You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2020/07/13 17:13:25 UTC

[GitHub] [incubator-tvm] kevinthesun commented on a change in pull request #6040: [TOPI] Improve schedule for injective on x86

kevinthesun commented on a change in pull request #6040:
URL: https://github.com/apache/incubator-tvm/pull/6040#discussion_r453802772



##########
File path: topi/include/topi/x86/injective.h
##########
@@ -44,14 +47,29 @@ namespace x86 {
  * \return The updated schedule.
  */
 inline Schedule schedule_injective_from_existing(Schedule sch, const Tensor& out) {
-  auto axis = sch[out]->op.as<ComputeOpNode>()->axis;
-  if (axis.size() == 4) {
-    auto n = axis[0];
-    auto c = axis[1];
-    auto fused = detail::Fuse(sch[out], {n, c});  // for nhwc layout, fuse n and h
-    sch[out].parallel(fused);
+  const auto& axes = sch[out]->op.as<ComputeOpNode>()->axis;

Review comment:
       Is this schedule func used in the common compilation flow? In op strategy we use python version of injective schedule and it should handle parallelism appropriately. 




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org