[Ecls-list] ERROR caused by FPU control settings...

Dietrich Bollmann dietrich at formgames.org
Sun Apr 21 07:11:41 UTC 2013

Hi Juanjo,

Thanks for your detailed answer!

On Fri, Apr 19, 2013 at 11:47 PM, Juan Jose Garcia-Ripoll <
juanjose.garciaripoll at gmail.com> wrote:

> On Fri, Apr 19, 2013 at 5:07 AM, Dietrich Bollmann <dietrich at formgames.org
> > wrote:
>> Yes, I am:  ECL relies on the underlying operating system > The operating
>> system is Windows > And here is the problem :)
> ECL needs to touch the FPU because Common Lisp mandates that arithmetic
> errors give rise to exceptions (which are then turned into Common Lisp
> conditions). However, ECL contains a flag to build without touching the FPU
> but rather checking the outcome of the computation and the FPU flags (See
> src/c/number.d and the headers it includes)
> Note that not using the FPU is fragile: to be conformant, in the compiled
> code we would have to add such checks _everywhere_, but we currently avoid
> that by setting the exception flags in the FPU.

Now I understand the motivation for manipulating the FPU flags better.  I
will try to avoid using the build flag then.

For the time being I get along with wrapping the code into the Rhino
macros ON_FPU_BeforeSloppyCall()
and ON_FPU_AfterSloppyCall().  I add your comment to the code to explain
the motivation for using them.

By the way, let me use the opportunity to express my gratitude for your
diligent work on ECL: thank you very much for this very interesting and
useful project!

With my best wishes,


Dietrich Bollmann, Ph.D.
Computational Architecture
The University of Tokyo
Department of Architecture
4-6-1 Komaba, Meguro-ku, Tokyo 153-8505, JAPAN

> Juanjo
> --
> Instituto de Física Fundamental, CSIC
> c/ Serrano, 113b, Madrid 28006 (Spain)
> http://juanjose.garciaripoll.googlepages.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.common-lisp.net/pipermail/ecl-devel/attachments/20130421/a594186a/attachment.html>

More information about the ecl-devel mailing list