http://llvm.org/bugs/show_bug.cgi?id=21435

            Bug ID: 21435
           Summary: LLVM no longer honours -stack-alignment=4
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: Backend: X86
          Assignee: [email protected]
          Reporter: [email protected]
                CC: [email protected]
    Classification: Unclassified

Created attachment 13268
  --> http://llvm.org/bugs/attachment.cgi?id=13268&action=edit
bug1347703.ll

LLVM stoped honouring 

This can easily be reproduced with the attached bug1347703.ll doing

  $ llc -o - -mtriple=i686-pc-mingw32 -mattr=+avx -stack-alignment=4
bug1347703.ll | grep '\<and.*[er]sp'
  $

which produces nothing showing that the stack pointer is never aligned.

However if one disables frame-pointer elimination then it works again:

  $
/home/jfonseca/work/vmware/llvm/llvm/build/linux-x86_64/Debug+Asserts/bin/llc
-disable-fp-elim -o - -mtriple=i686-pc-mingw32 -mattr=+avx -stack-alignment=4
bug1347703.ll | grep '\<and.*[er]sp'
    andl    $-32, %esp
    andl    $-32, %esp

  $

Which doesn't seem the right behavior -- using the frame-pointer to align the
stack seems an implementation detail -- there are probably other ways, and if
frame-pointer is indeed necessary it should be automatically employed as
necessary.


This regression happened between LLVM release 3.2 and release 3.3.  I strongly
suspect r168627 --
https://github.com/llvm-mirror/llvm/commit/1243922fc1a1e3d2681ed9e78503eeabd875ba93

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
LLVMbugs mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/llvmbugs

Reply via email to