[PATCH] D64662: [FPEnv] [PowerPC] Lower ppc_fp128 StrictFP Nodes to libcalls

[Jinsong Ji via Phabricator via llvm-commits]

jsji added inline comments.


================
Comment at: test/CodeGen/PowerPC/ppcf128-constrained-fp-instrinsics.ll:12
+; PC64LE-NEXT:    stdu 1, -32(1)
+; PC64LE-NEXT:    .cfi_def_cfa_offset 32
+; PC64LE-NEXT:    .cfi_offset lr, 16
----------------
ajwock wrote:
> kbarton wrote:
> > Others may disagree with me, but I would recommend getting rid of the checks for cfi here.
> > I don't think the location of the cfi has any bearing on the correctness here. I also think the compiler should be free to move them with respect to the  std and stdu instructions. If it does so, this test case will fail. 
> Alright- loosened -NEXT restrictions around std/u instructions.
You should be able to avoid checking those by adding `nounwind` to all these functions, you can see examples in vector-constrained-fp-intrinsics.ll in https://reviews.llvm.org/D67016.
This way, you can still use the script to generate the check automatically.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D64662/new/

https://reviews.llvm.org/D64662