[PATCH] D51639: [LV] Fix PR38786 - consider first order recurrence phis non-uniform

[Anna Thomas via Phabricator via llvm-commits]

anna added inline comments.


================
Comment at: test/Transforms/LoopVectorize/X86/uniform-phi.ll:90
+  %i32next = add i32 %phi32, 1
+  %i64next = zext i32 %i32next to i64
+  %xip = getelementptr inbounds double, double* %x, i64 %i64next
----------------
orivej wrote:
> Note: if I replace `%i64next = zext i32 %i32next to i64` with `%i64next = add i64 %phi64, 1` it ceases to be a first order recurrence, and the loop vectorizer considers `%phi32` and `%phi64` uniform and correctly optimizes this.
yes, that is expected :)

iv64next is a uniform IV user, so both the phis and their increments are considered uniforms.  See the processing of uniforms for IV phis in the source code, just after the above fix. 


Repository:
  rL LLVM

https://reviews.llvm.org/D51639