[PATCH] D21251: [TTI] The cost model should not assume illegal vector casts get completely scalarized
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 21 08:17:30 PDT 2016
RKSimon added a comment.
The SSE2 costs look very high which I guess is due to missing cost entries - that should be fixable in a pre/post patch if you prefer to keep them separate. The changed x86 costs all look reasonable.
================
Comment at: test/Analysis/CostModel/X86/sitofp.ll:72
@@ -71,3 +71,3 @@
; SSE2-LABEL: sitofpv32i8v32double
; SSE2: cost of 320 {{.*}} sitofp
;
----------------
This looks very high, codegen looks like it takes 26 ops - I haven't checked the throughput.
http://reviews.llvm.org/D21251
More information about the llvm-commits
mailing list