[llvm-bugs] [Bug 26859] [x86, SSE] only use phaddw / phaddd when optimizing for minsize?

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=26859

Sanjay Patel <spatel+llvm at rotateright.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED

--- Comment #14 from Sanjay Patel <spatel+llvm at rotateright.com> ---
(In reply to Simon Pilgrim from comment #13)
> https://reviews.llvm.org/D53095

Committed here:
https://reviews.llvm.org/rL344361

There's a stunning amount of vector duplication + unrolling in these tests
currently:
$ clang -O2 accum.c -S -o - -mavx | grep padd | wc -l
     155

...but that's not this bug.

This is the current behavior:

$ clang -Os accum.c -S -o - -mavx | grep phadd | grep mm
        vphaddd %xmm0, %xmm0, %xmm0
        vphaddw %xmm0, %xmm0, %xmm0
$ clang -O2 accum.c -S -o - -mavx | grep phadd | grep mm
$ clang -O2 accum.c -S -o - -mavx -march=btver2 | grep phadd | grep mm
        vphaddd %xmm0, %xmm0, %xmm0
        vphaddw %xmm0, %xmm0, %xmm0

Ie, if we are optimizing for size or Jaguar, we'll use horizontal ops,
otherwise, we use regular ops and shuffles. 

It's possible that our combiner predicate will need adjustments to optimize
that decision depending on code pattern and uarch, but we now have that
ability. Ideally, we can refine the choice using CPU instruction
latency/throughput models rather than with the DAG heuristic in the patch, but
that's also another bug.

-- 
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/20181012/9b1eb487/attachment-0001.html>