[PATCH] D152963: [RISCV] Don't assume tail undefined if there's no policy op

[Philip Reames via Phabricator via llvm-commits]

reames added a comment.

In D152963#4424968 <https://reviews.llvm.org/D152963#4424968>, @reames wrote:

> Can I ask you to hold this patch for a day or so?  I've got an idea on correctly inferring tail undefined here that I want to try out.  If it works, this becomes much closer to NFC.

The more complicated idea turned out not to be needed, and instead I posted the mini series ending in https://reviews.llvm.org/D153070.

I think if you rebase on that, some of your diffs should disappear.  I am curious as to the case which requires the _TU pseudo for VPatBinaryM.  I didn't run into that with my change, and I suspect your refactor here is creating more TU nodes.



================
Comment at: llvm/test/CodeGen/RISCV/rvv/roundtozero-vp.ll:698
 ; CHECK-NEXT:    fld fa5, %lo(.LCPI29_0)(a1)
+; CHECK-NEXT:    vmset.m v0
 ; CHECK-NEXT:    vsetvli zero, a0, e64, m8, ta, ma
----------------
This delta is very odd as v0 doesn't appear to be used.  Why is this not getting DCE?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D152963/new/

https://reviews.llvm.org/D152963