[PATCH] D88380: [VPlan] Extend VPValue to also model sub- & 'virtual' values.

Sun Nov 15 11:59:16 PST 2020

fhahn abandoned this revision.
fhahn added a comment.

Thanks for all the feedback. I'd propose to continue the discussion at the VPDef proposal (D90558 <https://reviews.llvm.org/D90558>). I'll abandon this one now, to avoid any confusion.

In D88380#2395190 <https://reviews.llvm.org/D88380#2395190>, @dcaballe wrote:

> Just looking around! Glad to see VPlan moving forward! :)

Thanks for taking a look! We moved on to a slightly different implementation of the same ideas in D90558 <https://reviews.llvm.org/D90558>.

================
Comment at: llvm/lib/Transforms/Vectorize/VPlanValue.h:196
+
+  VPValue *getDefiningValue() {
+    assert(isSubValue() && "can only get defining value of sub-value");
----------------
dcaballe wrote:
> This, indeed, looks pretty much like MLIR API :)
Glad things align there :) We moved on to a slightly different implementation which represents the defined values of an operation through a dedicated VPDef class, without the need to complicate the VPValue hierarchy.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88380/new/

https://reviews.llvm.org/D88380