RE: a small issue in berkeley upc

• Previous message: jcduell_at_lbl_dot_gov: "Re: ranlib requirement in upcc installation"

At 01:35 PM 5/6/2004, you wrote:
>Dear Dan,
>
>I was wondering if you still have the number of the bug report somewhere. I
>searched but it looks like I can't locate it.

The bug is reported here:
https://mantis.lbl.gov/bugzilla/show_bug.cgi?id=548
and is already fixed in our development head as you suggest:
#define _F8RECIP(v) (1.0/(v))

I've just confirmed your report that this fix is not in our current snapshot 
download - we'll be updating that shortly, but in the meantime you can 
manually apply the macro fix above to whirl2c.h

Sorry for the confusion...
Dan


>Meanwhile I tried the stable snapshot but it is still there. In the 
>translated
>file, the operation
>
>c = 1.0e0/(a*b)
>
>is translated into
>
>c = _F8RECIP(a * b)
>
>but this macro, defined in <upcc prefix>/include/upcr-postinclude/whirl2c.h 
>does
>not handle as it was #defined to be 1.0/v and not 1.0/(v) .
>
>I solved the problem by modifying the macro. It could be solved the other way 
>by
>generating c = _F8RECIP((a*b)) .
>
>I won't post this a a new bug as you say it is already there and fixed. 
>However,
>I'd appreciate if you still remember the bug report number that reported the
>thing.
>
>Sincerely,
>Mehmet F.
>
>PS: I wrote the path out of my head, so it may not be correct.
>
> > ----- Message from bonachea_at_cs_dot_berkeley_dot_edu ---------
> >     Date: Tue, 04 May 2004 10:06:10 -0700
> >     From: Dan Bonachea <bonachea_at_cs_dot_berkeley_dot_edu>
> > Reply-To: Dan Bonachea <bonachea_at_cs_dot_berkeley_dot_edu>
> >  Subject: RE: a small issue in berkeley upc
> >       To: Katherine Yelick <yelick_at_eecs_dot_berkeley_dot_edu>, "Mehmet F. Su"
><mfatihsu_at_eece_dot_unm_dot_edu>
> >
> > Yup - this bug was already found and fixed. You need the latest runtime to
> > fix
> > the problem (I believe it's in our current stable snapshot, but it may not
> > have made it into there yet).
> > Please report any future bugs (or suspected bugs) in our bug database 
> here:
> > http://mantis.lbl.gov
> >
> > Thanks for the report.
> > Dan
> >
> > At 11:06 PM 5/3/2004, Katherine Yelick wrote:
> > >Mehmet,
> > >
> > >Thanks for the note.  I think this is a know bug, or at least there
> > >were some bugs in 1.1 that fall into this general category.  I'm 
> forwarding
> > >your note to the developers list.  Note that there is a bug database
> > >available from the web page as well, so you can check whether this 
> problem
> > >is already known.  The rest of the group may be able to tell you whether
> > >a bug report is needed and if there is a version that fixes this 
> particular
> > >problem.
> > >                         Kathy
> > >
> > > > -----Original Message-----
> > > > From: Mehmet F. Su [mailto:mfatihsu_at_eece_dot_unm_dot_edu]
> > > > Sent: Monday, May 03, 2004 10:48 PM
> > > > To: yelick_at_cs_dot_berkeley_dot_edu
> > > > Cc: dbader_at_ece_dot_unm_dot_edu
> > > > Subject: a small issue in berkeley upc
> > > >
> > > >
> > > > Dear Prof Yelick,
> > > >
> > > > I am a graduate student of Dr David Bader at University of New Mexico 
> ECE
> > > > department. I am writing to you per his suggestions to inform you of a
> > > > small problem we have discovered recently.
> > > >
> > > > It appears that an expression of type
> > > >
> > > > c = 1.0e0/(a*b)
> > > >
> > > > is translated to
> > > >
> > > > c = 1.0e0/a*b
> > > >
> > > > as should be visible when the attached snippet is compiled and run. I
> > have
> > > > been using the 1.1.0 release of Berkeley UPC with the default (CGI)
> > > > translator enabled.
> > > >
> > > > Best regards,
> > > > Mehmet F. Su
> >
> >
> >
> > ----- End message from bonachea_at_cs_dot_berkeley_dot_edu -----

• Previous message: jcduell_at_lbl_dot_gov: "Re: ranlib requirement in upcc installation"