[PATCH] Model sqrtss as a binary operation with one source operand tied to the destination (PR14221)

Sanjay Patel spatel at rotateright.com
Mon Dec 15 08:01:59 PST 2014


In http://reviews.llvm.org/D6330#101289, @RKSimon wrote:

> Is there anything stopping the sse1_fp_unop_* and sse2_fp_unop_* defs being merged into sse12_fp_unop_* like the binops already have been? SQRTSD is the only reference to sse2_fp_unop_s and generally there is a high level of code duplication between the 2 sets of defs that the sse12_ types reduce a lot.


I was planning to fix sqrtsd in a follow-on patch. I agree that there's more consolidation possible, and I don't think there's anything that would prevent handling the unop cases similarly to the binop cases.


http://reviews.llvm.org/D6330

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/






More information about the llvm-commits mailing list