[quake3-bugzilla] [Bug 3533] Some calculations have different results in shared libraries/QVMs

bugzilla-daemon at icculus.org bugzilla-daemon at icculus.org
Thu Oct 22 05:42:38 EDT 2009


http://bugzilla.icculus.org/show_bug.cgi?id=3533

M. Kristall <mkpdev at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mkpdev at gmail.com

Thilo Schulz <arny at ats.s.bawue.de> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |arny at ats.s.bawue.de

--- Comment #5 from M. Kristall <mkpdev at gmail.com> 2009-10-21 20:59:55 EDT ---
I think the specific optimization responsible is -fno-rounding-math. Try
-frounding-math instead of removing -ffast-math, which would disable a number
of other useful optimizations.

I've been informed that this should no longer occur in the specific cases
mentioned and I'm not sure anyone else noticed it, so is this really necessary?

--- Comment #6 from Thilo Schulz <arny at ats.s.bawue.de> 2009-10-22 05:42:34 EDT ---
I suggest using separate C-Flags for building the QVM files, that have the
-ffast-math removed. Maybe just strip the -ffast-math flag via some regex in
the calls for compiling the QVMs in the Makefile? Could this be done portably?

-- 
Configure bugmail: http://bugzilla.icculus.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.


More information about the quake3-bugzilla mailing list