tree 139940f975365439755a40857b7af15e9de8d1c2
parent 02f9d9edf4131337257e115da832667c7a5fec96
author Han-Chung Wang <hanchung@google.com> 1603734437 -0700
committer GitHub <noreply@github.com> 1603734437 -0700
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJflwulCRBK7hj4Ov3rIwAAdHIIADjvoQE0UNCGuP/k1WpG1HD0
 h1IqqJhFmBY9KHoSaNFF2bft8Juy4VOjD+sWnYrwaAmIn6wPW8QtqDz4XtJwUcUo
 IFdYsTbwWl5JxJz7yp2CWpGoG1WwnoNNryoXUE1Qp5qNagk8OsLsX5XTGhg/YiZc
 2oseZUkk3/rSe9QsUrMdEMvCzqP1CFjP2w2w2uKXuC8dVzjq++UnSthkztIybvyY
 n+DIqGkyw0FyVJc+mFlTNf5WNxupaeMdoeLVFP8+ulRWRltYH9MxB1Sex+d2niOO
 9CvdttiGrBSy00K5gXa70zVXlXDu13GAgbhChV0hM1qQUL3denMVbrGSpfC+7bM=
 =kiqr
 -----END PGP SIGNATURE-----
 

Move torch_index_select op lowering to Linalg on tensors. (#3578)

Use extract_element op instead of load op, since it works in tensor's world. It will be lowered to a load op in iree-codegen-hlo-to-linalg-on-buffers pass.