|
From: Dave N. <dc...@us...> - 2006-04-11 18:41:58
|
I submitted my rounding mode test case with regression test support
files a couple of weeks ago ("FP rounding mode test case for
ppc32/ppc64") but never heard anything further about it.
Was there a problem with the test or did it just fall through the cracks?
Julian Seward wrote:
> Hi. This seems like a good thing to do and I think you should
> continue with it. You'll see that in the past couple of days I
> added auxprogs/ppcround.c for doing similar testing, but that
> produces ~ 30MB of output and so isn't a viable regression test
> (and so is not part of the standard build). So it would be good
> to develop your program further.
>
>
>>I would like to submit it as a regression test but wanted to make sure I
>>am following whatever conventions are required for regression tests and
>>that it isn't redundant. I haven't yet figured out what is needed in
>>terms of the *.std{err,out}.exp but if it is simply a matter of
>>capturing the stdout/stderr and filtering out things like addresses then
>>that step should be pretty trivial.
>
>
> Really the requirements are
>
> - that the .stderr.exp/.stderr.out are not too large, say < 100k
>
> - that it produces identical output on different CPU implementations,
> eg not only on ppc970 but also on power5 and perhaps on low-end
> 32-bit only implementations
>
> - that it works in both 32- and 64-bit mode (doesn't have to produce
> the same output in both modes)
>
> How I tests to the framework is (this is a bit of a drag, but still):
>
> - edit Makefile.am in the relevant directory. Add entries to
> check_PROGRAMS and EXTRA_DIST (see eg none/tests/Makefile.am
> for an example)
>
> - put the .c file in the directory
>
> - create a .vgtest file - very easy
>
> - go up to the top of the tree and rebuild from ./configure onwards.
> This is needed so that the changes in Makefile.am work.
>
> - make install
>
> - make check. This builds the regression tests
>
> - run your tests individually:
>
> perl tests/vg_regtest none/tests/ppc32/mytest (or whatever)
>
> - when it complains about missing .std{err,out}.exp, copy the
> corresponding .out files (which will have been generated)
>
> Note, you may find the --enable-only32bit and --enable-only64bit
> configure flags to be helpful while hacking. They disable the biarch
> builds which is sometimes useful.
>
> J
>
|
|
From: Julian S. <js...@ac...> - 2006-04-16 00:15:30
|
Committed (r5853). Thanks for the patch. Apologies for delay
on it.
J
On Tuesday 11 April 2006 19:42, Dave Nomura wrote:
> I submitted my rounding mode test case with regression test support
> files a couple of weeks ago ("FP rounding mode test case for
> ppc32/ppc64") but never heard anything further about it.
>
> Was there a problem with the test or did it just fall through the cracks?
>
> Julian Seward wrote:
> > Hi. This seems like a good thing to do and I think you should
> > continue with it. You'll see that in the past couple of days I
> > added auxprogs/ppcround.c for doing similar testing, but that
> > produces ~ 30MB of output and so isn't a viable regression test
> > (and so is not part of the standard build). So it would be good
> > to develop your program further.
> >
> >>I would like to submit it as a regression test but wanted to make sure I
> >>am following whatever conventions are required for regression tests and
> >>that it isn't redundant. I haven't yet figured out what is needed in
> >>terms of the *.std{err,out}.exp but if it is simply a matter of
> >>capturing the stdout/stderr and filtering out things like addresses then
> >>that step should be pretty trivial.
> >
> > Really the requirements are
> >
> > - that the .stderr.exp/.stderr.out are not too large, say < 100k
> >
> > - that it produces identical output on different CPU implementations,
> > eg not only on ppc970 but also on power5 and perhaps on low-end
> > 32-bit only implementations
> >
> > - that it works in both 32- and 64-bit mode (doesn't have to produce
> > the same output in both modes)
> >
> > How I tests to the framework is (this is a bit of a drag, but still):
> >
> > - edit Makefile.am in the relevant directory. Add entries to
> > check_PROGRAMS and EXTRA_DIST (see eg none/tests/Makefile.am
> > for an example)
> >
> > - put the .c file in the directory
> >
> > - create a .vgtest file - very easy
> >
> > - go up to the top of the tree and rebuild from ./configure onwards.
> > This is needed so that the changes in Makefile.am work.
> >
> > - make install
> >
> > - make check. This builds the regression tests
> >
> > - run your tests individually:
> >
> > perl tests/vg_regtest none/tests/ppc32/mytest (or whatever)
> >
> > - when it complains about missing .std{err,out}.exp, copy the
> > corresponding .out files (which will have been generated)
> >
> > Note, you may find the --enable-only32bit and --enable-only64bit
> > configure flags to be helpful while hacking. They disable the biarch
> > builds which is sometimes useful.
> >
> > J
>
> -------------------------------------------------------
> This SF.Net email is sponsored by xPML, a groundbreaking scripting language
> that extends applications into web and mobile media. Attend the live
> webcast and join the prime developer group breaking into this new coding
> territory!
> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642
> _______________________________________________
> Valgrind-developers mailing list
> Val...@li...
> https://lists.sourceforge.net/lists/listinfo/valgrind-developers
|