Strange behavior of mpf_inp_str
bruck at math.usc.edu
Fri Feb 19 17:46:30 CET 2010
Roger Golliver wrote:
> Date: Thu, 18 Feb 2010 09:28:47 -0800
> From: "Golliver, Roger A" <roger.a.golliver at intel.com>
> Subject: Re: Strange behavior of mpf_inp_str
> To: "gmp-bugs at gmplib.org" <gmp-bugs at gmplib.org>
> <ADDEBCD9C145E542A8688C29F1FBDCB95DD13DCB at orsmsx502.amr.corp.intel.com
> Content-Type: text/plain; charset="us-ascii"
> You might want to try gcc and if it works file a bug against icc.
I have tried that, but the behavior continues. I also recompiled
gmp-5.0.1 (using defaults, including gcc, except I configured with --
enable-cxx). On the i7, the behavior continues, despite passing
Now I also have a laptop, using a core2 processor and the same
release of SuSE Linux. When I installed gcc-5.0.1, the bug did NOT
manifest itself. I'm incredulous that it could be the i7. I'm not
sure what to try now.
I can work around the bug. Most of my programs are running under the
direction of scripts anyway; all I have to do is insert a "sed 's/e+/
e/g'" right after the file with the e+ is created. Or I can figure
out which instruction is writing the e+ in the first place (I think
it's gmp_fprintf) and fix that (IIRC there's a format to suppress the
+). But the feeling I'm having is similar to the one I had when I
went deaf in my left ear: from onset of ringing to total deafness
took half-an-hour, and my worry ever since has been, What ELSE might
suddenly fall off? ;-) What if something happens that I DON'T
notice, and my programs suddenly change and begin producing incorrect
output? We reassure our children there's no monster under the bed,
but is that always true?
If someone else is using an i7, I'd appreciate hearing verification
that mpf_inp_str works on YOUR configuration. To repeat the
problem: MY mpf_inp_str doesn't recognize an explicit + sign in the
exponent, e.g. 1.234e+2. I'm puzzled why the installation passes the
My i7 is mildly overclocked, but I have the best MB and memory money
could buy, and when overclocking causes problems it's not usually
something this specific (just general flakiness). Still, I'll turn
that off. Next I'll try starting in failsafe mode. Ultimately I
could reinstall the system, but that's draconian.
More information about the gmp-bugs