[llvm-bugs] [Bug 26106] New: Loop Vectorizer on ARM - why is the relative speedup so much worse?

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=26106

            Bug ID: 26106
           Summary: Loop Vectorizer on ARM - why is the relative speedup
                    so much worse?
           Product: libraries
           Version: 3.6
          Hardware: Other
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: Loop Optimizer
          Assignee: unassignedbugs at nondot.org
          Reporter: tulipawn at gmail.com
                CC: llvm-bugs at lists.llvm.org
    Classification: Unclassified

Created attachment 15600
  --> https://llvm.org/bugs/attachment.cgi?id=15600&action=edit
armv7 LLVM-IR output

Hi,

It's my first issue here (I only use llvm via rust) so I might be mistaken in
my understanding but here's the problem. The code snippet from this rust issue:

https://github.com/rust-lang/rust/issues/30805

is being successfully optimized on x86_64 and posts a 20x speedup using SSE2
whereas on ARM it's a meagre 3x speedup using NEON.

The relative difference seems a little high - is there a problem somewhere on
ARM?

I'm attaching armv7 LLVM-IR in case there is.

Thanks.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20160110/dd6ba3e3/attachment.html>