[LLVMbugs] [Bug 21231] use fsel to avoid branch and compare (PPC, fast-math)

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=21231

Hal Finkel <hfinkel at anl.gov> changed:

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

--- Comment #2 from Hal Finkel <hfinkel at anl.gov> ---
The problem here is that using fsel for the select_cc nodes here is safe only
if we assume no-NaNs and no-Infs. If you run the test with
-enable-no-infs-fp-math -enable-no-nans-fp-math also, then you'll see:

.L.goo3:
# BB#0:
        frsqrtes 0, 1
        addis 3, 2, .LCPI0_0 at toc@ha
        fneg 13, 1
        lfs 2, .LCPI0_0 at toc@l(3)
        addis 3, 2, .LCPI0_1 at toc@ha
        lfs 12, .LCPI0_1 at toc@l(3)
        fnmsubs 3, 1, 2, 1
        fmuls 4, 0, 0
        fmadds 2, 3, 4, 2
        fmuls 0, 0, 2
        fmuls 0, 1, 0
        fsel 2, 1, 12, 0
        fsel 1, 13, 2, 0
        blr

-- 
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/20141010/86eac090/attachment.html>