[PATCH] D57148: [X86][Btver2] Improved latency/throughput model for scalar int-to-float conversions.
Andrea Di Biagio via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Tue Jan 29 05:30:32 PST 2019
andreadb updated this revision to Diff 184062.
andreadb marked an inline comment as done.
andreadb added a comment.
Patch updated.
I verified that the int-to-fp convert latency is 4cy (instead of 3cy) if we exclude the extra 6cy of bypass delay.
So, I went ahead and set the latency directly to 4cy.
I've also verified by running some microbenchmarks that the RM variants have a latency of ~9cy (5cy for the load opcode + 4cy of convert). That latency seems consistent with what I found when testing the RR variants. The only difference for the RM variants is the number of opcodes (it is 1 COP instead of 2 COPs).
So, I went ahead and fixed all those writes in this patch (rather than splitting the change in two patches).
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D57148/new/
https://reviews.llvm.org/D57148
Files:
lib/Target/X86/X86InstrSSE.td
lib/Target/X86/X86ScheduleBtVer2.td
test/CodeGen/X86/sse-schedule.ll
test/CodeGen/X86/sse2-schedule.ll
test/tools/llvm-mca/X86/BtVer2/int-to-fpu-forwarding-2.s
test/tools/llvm-mca/X86/BtVer2/resources-avx1.s
test/tools/llvm-mca/X86/BtVer2/resources-sse1.s
test/tools/llvm-mca/X86/BtVer2/resources-sse2.s
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D57148.184062.patch
Type: text/x-patch
Size: 17187 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20190129/d6105496/attachment.bin>
More information about the llvm-commits
mailing list