[GitHub] [incubator-tvm] roastduck commented on pull request #5382: [TE] Fix MakeLoopNest for warp memory
roastduck commented on pull request #5382: URL: https://github.com/apache/incubator-tvm/pull/5382#issuecomment-625574365 @yongfeng-nv Fixed. 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: us...@infra.apache.org
[GitHub] [incubator-tvm] roastduck commented on pull request #5382: [TE] Fix MakeLoopNest for warp memory
roastduck commented on pull request #5382: URL: https://github.com/apache/incubator-tvm/pull/5382#issuecomment-625044388 @yongfeng-nv Thanks. Your improved test helps a lot. I call `tvm.driver.build_module.form_irmodule` directly, in order not to run the transformation passes in `lower`, otherwise the result will not make a difference. Then I assert `threadIdx.y` is not in the indices accessing `A.warp`. This test should be enough for the 2nd situation. We can ignore the 3rd situation for now, since it is not supported by `lower_warp_memory` yet. 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: us...@infra.apache.org
[GitHub] [incubator-tvm] roastduck commented on pull request #5382: [TE] Fix MakeLoopNest for warp memory
roastduck commented on pull request #5382: URL: https://github.com/apache/incubator-tvm/pull/5382#issuecomment-623041791 It will be great if someone can make a review. 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: us...@infra.apache.org
[GitHub] [incubator-tvm] roastduck commented on pull request #5382: [TE] Fix MakeLoopNest for warp memory
roastduck commented on pull request #5382: URL: https://github.com/apache/incubator-tvm/pull/5382#issuecomment-620456420 @wpan11nv Could you explain how subgroups in OpenCL works? Up to now, we always assumed `threadIdx.x` equals to warp in the `lower_warp_memory` pass. Does it mean there are more problems with `lower_warp_memory` for OpenCL? 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: us...@infra.apache.org
[GitHub] [incubator-tvm] roastduck commented on pull request #5382: [TE] Fix MakeLoopNest for warp memory
roastduck commented on pull request #5382: URL: https://github.com/apache/incubator-tvm/pull/5382#issuecomment-619291622 Do you mean requiring the users to tag the iter_var or we do it in InferBound? If the former, maybe we can merge this PR first and then start an RFC for the API change. If the latter, can you point out where to tag the iter_var? I am not familiar with InferBound. 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: us...@infra.apache.org
