[PATCH] D152276: [DAGCombine] Fold (store (insert_elt (load p)) x p) -> (store x)

[Fraser Cormack via Phabricator via llvm-commits]

frasercrmck added inline comments.


================
Comment at: llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert.ll:8
 
 define void @insertelt_v4i64(ptr %x, i64 %y) {
 ; RV32-LABEL: insertelt_v4i64:
----------------
RKSimon wrote:
> luke wrote:
> > This might be obscuring the behaviour of the actual insert_vector_elt codegen. Let me know if I should mark these stores as volatile
> @craig.topper Any thoughts on this?
Maybe we can just write a separate test function that inserts into vectors passed by value? I personally consider the load/store idiom in these tests to be from before we added proper vector calling convention support.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D152276/new/

https://reviews.llvm.org/D152276