[LLVMbugs] [Bug 21291] New: reassociate pass needs to check for function-level fast-ness

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=21291

            Bug ID: 21291
           Summary: reassociate pass needs to check for function-level
                    fast-ness
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: Scalar Optimizations
          Assignee: unassignedbugs at nondot.org
          Reporter: spatel+llvm at rotateright.com
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified

Filing this bug as requested in http://reviews.llvm.org/D5222 and discussed in
http://reviews.llvm.org/D5787.

Currently, the reassociate pass checks for IR-level fast-math-flags (FMF) when
performing optimizations. It should also check function-level attributes to see
if reassociation is allowed.

LTO (and possibly just regular inlining?) creates a scenario where non-fast
instructions could end up in a function that has some level of fast-ness. In
that case, the parent function's attributes apply to whatever was inlined.

The other case where IR level FMF is not currently available is with FP
intrinsics. This is tracked by bug 21290.

-- 
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/20141015/d47bb79a/attachment.html>