[PATCH] D70392: [Attributor] Track a GEP Instruction in align deduction

[Roman Lebedev via Phabricator via llvm-commits]

lebedev.ri added inline comments.


================
Comment at: llvm/test/Transforms/FunctionAttrs/align.ll:352-359
+; FXIME: %p should have nonnull
+; ATTRIBUTOR: define i64 @test12-2(i32* nocapture nofree readonly align 16 %p)
+define i64 @test12-2(i32* align 4 %p) {
+  %p-cast = bitcast i32* %p to i64*
+  %arrayidx0 = getelementptr i64, i64* %p-cast, i64 0
+  %ret = load i64, i64* %arrayidx0, align 16
+  ret i64 %ret
----------------
Is there test coverage where this is a `store` instead of `load`,
and maybe there is a call to a function that expects certain aliment of it's pointer argument?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D70392/new/

https://reviews.llvm.org/D70392