floating point control in extensions

[Robin Becker]

Is there a preferred way to intercept floating point exceptions in
python extensions? I assume one should be careful to restore any
existing error handler. Does python have a standard mechanism for
setting up fpu control words etc?
--
Robin Becker

[Robin Becker]

Re: floating point control in extensions

In article <mailman.105811 9520.21914.pyth on-list@python.org >, Tim Peters
<tim.one@comcas t.net> writes
.......[color=blue]
>Python itself assumes that no-stop mode is in effect (the IEEE-754 mandated
>default: all FPU traps are disabled).
>[/color]
.....
That at least makes resetting easy when one can do it.



I guessed all Python floating point ops are tested either pre or post
for the correct outcome, but then I tried
[color=blue][color=green][color=darkred]
>>> a=1e290
>>> a[/color][/color][/color]
1.0000000000000 001e+290[color=blue][color=green][color=darkred]
>>> a*a1.#INF
>>>[/color][/color][/color]

so I suppose only the obvious things are tested.
-it's hard being an fpu'ly yrs-
Robin Becker