[LLVMbugs] [Bug 5990] [x86, x86-64] MMX vectors trample on FP stack, breaking 80-bit floats

[bugzilla-daemon at cs.uiuc.edu]

http://llvm.org/bugs/show_bug.cgi?id=5990


Dan Gohman <gohman at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |gohman at apple.com
             Status|REOPENED                    |RESOLVED
         Resolution|                            |DUPLICATE




--- Comment #6 from Dan Gohman <gohman at apple.com>  2010-01-11 13:59:06 ---
It's a serious bug in LLVM's MMX support, and it's also reported as PR1838.
It's been open for a long time, and to date no one has stepped forward to fix
it.

If you have a front-end which wants to emit vector code and not care about the
capabilities of the target, the -disable-mmx option is highly recommended. With
this option, <2 x i32> vectors are handled in SSE registers, with the compiler
doing all the bookkeeping necessary.

I agree that this state of affairs is not ideal, but LLVM is attempting to meet
the needs of several different classes of users, with drastically different
needs.

*** This bug has been marked as a duplicate of bug 1838 ***


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.