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 2021/03/08 08:47:05 UTC

[GitHub] [tvm] masahi opened a new pull request #7607: [SPIRV] Minor update to TIR sort to make it work on VK/SPIR-V

masahi opened a new pull request #7607:
URL: https://github.com/apache/tvm/pull/7607


   Didn't look into why but this change seems to make it possible to run TIR sort with VK/SPIR-V on **static** shape. Dynamic shapes on SPIR-V seem to have some issues.  


----------------------------------------------------------------
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



[GitHub] [tvm] mbrookhart merged pull request #7607: [SPIRV] Minor update to TIR sort to make it work on VK/SPIR-V

Posted by GitBox <gi...@apache.org>.
mbrookhart merged pull request #7607:
URL: https://github.com/apache/tvm/pull/7607


   


----------------------------------------------------------------
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



[GitHub] [tvm] tmoreau89 commented on a change in pull request #7607: [SPIRV] Minor update to TIR sort to make it work on VK/SPIR-V

Posted by GitBox <gi...@apache.org>.
tmoreau89 commented on a change in pull request #7607:
URL: https://github.com/apache/tvm/pull/7607#discussion_r589977760



##########
File path: tests/python/topi/python/test_topi_sort.py
##########
@@ -75,7 +75,7 @@ def check_device(device):
         f(tvm_data, tvm_out)
         tvm.testing.assert_allclose(tvm_out.asnumpy(), np_sort, rtol=1e0)
 
-    for device in ["llvm", "cuda", "opencl"]:
+    for device in ["llvm", "cuda", "opencl", "vulkan", "nvptx"]:

Review comment:
       Thanks @masahi  - in the case of vulkan here the tests won't be run until we add vulkan to our CI docker image, right? These are just tested locally?




----------------------------------------------------------------
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



[GitHub] [tvm] mbrookhart commented on pull request #7607: [SPIRV] Minor update to TIR sort to make it work on VK/SPIR-V

Posted by GitBox <gi...@apache.org>.
mbrookhart commented on pull request #7607:
URL: https://github.com/apache/tvm/pull/7607#issuecomment-792835543


   When I was developing this, I noticed that implicitly allocating the start/middle/end variables, i.e. removing this line: `start = ib.allocate("int64", (1,), name="start", scope="local")` broke things on nvptx, that's why I included them. NVPTX doesn't seem to be enabled for this test, could we turn that backend on and make sure we didn't break something?


----------------------------------------------------------------
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



[GitHub] [tvm] masahi commented on a change in pull request #7607: [SPIRV] Minor update to TIR sort to make it work on VK/SPIR-V

Posted by GitBox <gi...@apache.org>.
masahi commented on a change in pull request #7607:
URL: https://github.com/apache/tvm/pull/7607#discussion_r590019693



##########
File path: tests/python/topi/python/test_topi_sort.py
##########
@@ -75,7 +75,7 @@ def check_device(device):
         f(tvm_data, tvm_out)
         tvm.testing.assert_allclose(tvm_out.asnumpy(), np_sort, rtol=1e0)
 
-    for device in ["llvm", "cuda", "opencl"]:
+    for device in ["llvm", "cuda", "opencl", "vulkan", "nvptx"]:

Review comment:
       yes vulkan was tested locally, but not on CI.




----------------------------------------------------------------
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



[GitHub] [tvm] masahi commented on pull request #7607: [SPIRV] Minor update to TIR sort to make it work on VK/SPIR-V

Posted by GitBox <gi...@apache.org>.
masahi commented on pull request #7607:
URL: https://github.com/apache/tvm/pull/7607#issuecomment-793377898


   seems everything is ok, ready to merge? @mbrookhart 


----------------------------------------------------------------
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



[GitHub] [tvm] masahi commented on pull request #7607: [SPIRV] Minor update to TIR sort to make it work on VK/SPIR-V

Posted by GitBox <gi...@apache.org>.
masahi commented on pull request #7607:
URL: https://github.com/apache/tvm/pull/7607#issuecomment-793840005


   @mbrookhart I found the reason why dynamic shape is not working https://github.com/apache/tvm/pull/7620. We can merge this PR first and I can add a dynamic sort test in https://github.com/apache/tvm/pull/7620.


----------------------------------------------------------------
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