#11084: Singular 3-1-1-4.p4 fails to build with gcc 4.6.0.
----------------------------------------------------------------------------------------+
Reporter: drkirkby
| Owner: GeorgSWeber
Type: defect
| Status: positive_review
Priority: critical
| Milestone: sage-4.7
Component: build
| Keywords:
Author: David Kirkby
| Upstream: None of the above - read trac for reasoning.
Reviewer: AlexanderDreyer
| Merged:
Work_issues: Testing on Linux, OS X, Solaris 10 (both x86 and SPARC) with gcc
4.6.0. |
----------------------------------------------------------------------------------------+
Comment(by drkirkby):
Replying to [comment:19 jdemeyer]:
> Replying to [comment:17 AlexanderDreyer]:
> > This seems to be the right solution
>
> I completely disagree. Dropping the optimization is at best a
workaround.
Agreed.
> IMHO it can ''never'' be "the right solution". It means that either
the compiler or the code has a bug. Maybe dropping the optimization to
`-O2` covers up any issues, but I think one should really investigate what
the true problem is.
Well, a lot of the code in Sage is very difficult to navigate. Sage
developers did not write it. Much is incomprehensible. Most gives a lot of
compiler warnings. Sorting out if a gcc bug or a bug in the code is
difficult. Then if it is a bug in the code, finding where is often very
very difficult. One could spend ages on it.
> I am ''very strongly'' against saying to never increase the optimization
level to `-O3`. If some later upstream version has fixed this problem, we
''should'' increase the optimization level back to `-O3`.
Given this code has caused many issues on many platforms with optimization
issues, then I think not increasing it is worthwhile. It should also be
realised that -O3 does not necessarily run faster than -O2. The higher
optimization level invokes more speculative optimisations, which may cause
the code to run slower. See this point from a GCC developer.
http://gcc.gnu.org/ml/gcc-help/2010-07/msg00190.html
--------
''The -O3 optoin should be safe for correct code. An important difference
between -O2/-O3 and -O1 is that -O2 and -O3 enable strict aliasing and
strict overflow. Those options provide better optimization for correct
code, but are far more likely to cause unexpected code generation for
incorrect code. See the -fstrict-aliasing and -fstrict-overflow options.''
''The main difference between -O3 and -O2 is that -O3 enables more
speculative optimizations. These should not miscompile your code, but they
may cause your program to run more slowly. ''
------
So in summary
* The upstream developers use -O2
* Both -O2 and -O3 can cause problems with bad code.
* -O3 can be slower than -O2
* Previous attempts to use -O3 have caused problems on numerous platforms
* Fedora
* OpenSUSE on Itanium
* gcc on any platform
* Just for good measure, the code can't be compiled with the Sun
compiler, as it has hard-coded uses gcc-specific flags, so we can't even
get a second opinion about doggy lines of code.
* Finding whether it's a gcc bug or a bug in Singular would take quite a
bit of effort, and IMHO would be better spent elsewhere.
There seems to be little appetite among the Sage community for
investigating (and possibly fixing) the '''thousands''' of compiler
warnings - some of which look potentially serious to me. To me at least,
permitting -O3 on code which has caused numerous issues in the past is
unwise.
Dave
--
Ticket URL: <http://trac.sagemath.org/sage_trac/ticket/11084#comment:20>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/sage-trac?hl=en.
