[PATCH] D78768: [VE] Update floating-point arithmetic instructions

[Kazushi Marukawa via Phabricator via llvm-commits]

kaz7 marked an inline comment as done.
kaz7 added inline comments.


================
Comment at: llvm/lib/Target/VE/VEInstrInfo.td:1168
 // Cast to f32
-def : Pat<(f32 (sint_to_fp i64:$sy)), (CVSr (FLTXr i64:$sy))>;
+def : Pat<(f32 (sint_to_fp i64:$sy)), (CVTSDr (CVTDLr i64:$sy))>;
 
----------------
simoll wrote:
> Doesn't the CVTLDr require an explicit rounding mode as in l1164f above?
Yes, CVTLD (double to long) requires rounding mode.  For example, CVTDL (long to double) doesn't require rounding mode, though.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D78768/new/

https://reviews.llvm.org/D78768