[PATCH] D62024: [InstCombine] fold shuffles of insert_subvectors

[Eli Friedman via Phabricator via llvm-commits]

efriedma added inline comments.


================
Comment at: llvm/docs/CodeGenerator.rst:932
+* Extract subvector --- The vector is pulled from a longer vector type starting
+at index 0.
+
----------------
This description is useful from the perspective from the optimizer, but it would probably also be a good idea to mention what this means in practice from the perspective of the backend.  For shuffles where operand and result types are all the same, and legal, that's pretty straightforward, but some of these are more complicated. Can a splat change the vector length, and if so, what DAG nodes does that produce?  What DAG nodes are we actually talking about for an "insert subvector"?

Also, which of those categories does the proposed transform fall under?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D62024/new/

https://reviews.llvm.org/D62024