[Ecls-list] Ecl version of with-float-traps-masked and vector-sap?
Juan Jose Garcia-Ripoll
juanjose.garciaripoll at googlemail.com
Wed Feb 8 20:38:48 UTC 2012
On Wed, Feb 8, 2012 at 6:05 PM, Raymond Toy <toy.raymond at gmail.com> wrote:
> On Wed, Feb 8, 2012 at 6:01 AM, Juan Jose Garcia-Ripoll <
> juanjose.garciaripoll at googlemail.com> wrote:
>> On Wed, Feb 8, 2012 at 6:59 AM, Raymond Toy <toy.raymond at gmail.com>wrote:
>>> Does ecl have the equivalent of cmucl's (and sbcl's)
>> Not yet. This can be implemented using SI::TRAP-FPE
>> (defmacro with-float-traps-masked (&body body)
>> `(let ((%trap-bits (si::trap-fpe 'cl:last t)))
>> (progn (si::trap-fpe %trap-bits nil) , at body)
>> (si::trap-fpe %trap-bits t))))
> This will probably work. I take it that this will disable all traps?
> This would be the typical expected Fortran mode, so that's perfect.
'LAST tells ECL to return the last set of active traps and then the second
statement deactivates them all.
>>> and vector-sap?
>> Hmm, is this for the CFFI? Because it is going to depend very much on
>> what VECTOR-SAP has to return. For UFFI you have
>> SI::MAKE-FOREIGN-DATA-FROM-ARRAY which takes an array and returns an UFFI
>> pointer. Alternatively, C-INLINE may be easily used to retrieve the pointer
>> in a C format.
> Thanks for the tips. I'm not very familiar with cffi or uffi. Basically,
> vector-sap needs to return something that can be made into a pointer for a
> foreign function to use as the address of an array/vector.
I do not know how MATLISP invokes BLAS or equivalent functions. If you need
a pointer, then you need to know in which format MATLISP needs it.
I had a look at MATLISP. It is probably not unconceivable to translate it
all to ECL's FFI constructs. Seems that all the fortran interface is in one
file (ffi-*.lisp) and generates code automatically. Producing the
associated C code would be easy and more efficient.
Instituto de Física Fundamental, CSIC
c/ Serrano, 113b, Madrid 28006 (Spain)
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the ecl-devel