|
»
[bindings] Some thoughts
|
View:
New views
8 Messages
—
Rating Filter:
Alert me
|
 |
[bindings] Some thoughts
by Rutger ter Borg-2
::
Rate this Message:
Guys, I think I will not proceed creating a high-level interface in the bindings. Instead, I will put more effort in wrapping up the bindings, making them more user-friendly, and including the necessary facilities that enables integration in high-level linear algebra libraries (like uBLAS or GLAS). I would also like to propose that the stuff contained in the traits subdirectory is made more user-friendly. E.g., moving bindings/traits/ublas_matrix.hpp to bindings/ublas/matrix.hpp and enabling binding to whole ublas by bindings/ublas.hpp. The same for the standard containers, into std/vector.hpp, etc.. Cheers, Rutger _______________________________________________ ublas mailing list ublas@... http://lists.boost.org/mailman/listinfo.cgi/ublas Sent to: lists@... |
|
|
Re: [bindings] Some thoughts
by Jesse Perla
::
Rate this Message:
On Fri, Sep 18, 2009 at 3:06 AM, Rutger ter Borg <rutger@...> wrote: I think I will not proceed creating a high-level interface in the bindings. I think that makes a lot of sense until we can get a high level of coverage of the routines. I would also love to see traits for MTL4 if that is possible since it seems to be a very active library. _______________________________________________ ublas mailing list ublas@... http://lists.boost.org/mailman/listinfo.cgi/ublas Sent to: lists@... |
|
|
Re: [bindings] Some thoughts
by Rutger ter Borg-2
::
Rate this Message:
Jesse Perla wrote: > I think that makes a lot of sense until we can get a high level of > coverage > of the routines. I would also love to see traits for MTL4 if that is > possible since it seems to be a very active library. It's quite easy to provide traits for mtl4. It's even easier if their code base would have been stable :-) What classes would you need traits for? Cheers, Rutger _______________________________________________ ublas mailing list ublas@... http://lists.boost.org/mailman/listinfo.cgi/ublas Sent to: lists@... |
|
|
Re: [bindings] Some thoughts
by Peter Gottschling
::
Rate this Message:
Dear Jesse and Rutger,
I am glad to hear about your serious interest of providing bindings for MTL4. If I can help you with this please let me know. Karl (Meerbergen) and I decided a long time ago to unify the interfaces of MTL4 and GLAS for the sake of seamless interoperability. Unfortunately, we never found the time to realize this plan. Regarding Rutger's concern about the stability, it is true that I occasionally change some internals for better performance or a cuter interface but that did not affect basics like memory layout. If there were some significant changes we can still adapt the traits. This said, it is probably best storing them in a MTL4 subdirectory, so we can provide a consistent versioning. Cheers, Peter Rutger ter Borg schrieb: Jesse Perla wrote: -- Peter Gottschling, Ph.D. mail: pgottsch@... www: www.osl.iu.edu/~pgottsch _______________________________________________ ublas mailing list ublas@... http://lists.boost.org/mailman/listinfo.cgi/ublas Sent to: lists@... |
|
|
Re: [bindings] Some thoughts
by Jesse Perla
::
Rate this Message:
On Fri, Sep 18, 2009 at 1:56 PM, Peter Gottschling <pgottsch@...> wrote:
Thanks guys, I am more interested in exploring MTL as a potential next matrix librarychoice, and without the bindings it would be very difficult for me to make the move. I think that my immediate needs would be fairly simple. Dense matrices with symmetric and triangular traits. However, don't consider my interests a high priority. _______________________________________________ ublas mailing list ublas@... http://lists.boost.org/mailman/listinfo.cgi/ublas Sent to: lists@... |
|
|
[bindings] lapack errors
by Uldarico Muico-2
::
Rate this Message:
Hi. I get the following errors using functions like lapack::potrf and
lapack::sysv under debug mode boost/numeric/bindings/lapack/driver/sysv.hpp:117: error: no matching function for call to 'max(int, ptrdiff_t)' Perhaps you should use std::max<integer_t> explicitly? On a related note, I also get boost/numeric/bindings/lapack/driver.hpp:12:51: error: boost/numeric/ bindings/lapack/solve.hpp: No such file or directory _______________________________________________ ublas mailing list ublas@... http://lists.boost.org/mailman/listinfo.cgi/ublas Sent to: lists@... |
|
|
Re: [bindings] lapack errors
by Rutger ter Borg-2
::
Rate this Message:
Uldarico Muico wrote:
> Perhaps you should use std::max<integer_t> explicitly? I've updated it to std::max< std::ptrdiff_t >(...), that's the return type of all size/count-related traits functions. > On a related note, I also get > > boost/numeric/bindings/lapack/driver.hpp:12:51: error: boost/numeric/ > bindings/lapack/solve.hpp: No such file or directory Should be fixed, too. Please note that the uplo stuff for sysv will most likely change. Cheers, Rutger _______________________________________________ ublas mailing list ublas@... http://lists.boost.org/mailman/listinfo.cgi/ublas Sent to: lists@... |
|
|
Re: [bindings] lapack errors
by Uldarico Muico-2
::
Rate this Message:
They compile fine now. Thanks for the heads-up.
-uldarico On Oct 9, 2009, at 1:48 PM, Rutger ter Borg wrote: > Uldarico Muico wrote: > >> Perhaps you should use std::max<integer_t> explicitly? > > I've updated it to std::max< std::ptrdiff_t >(...), that's the > return type > of all size/count-related traits functions. > >> On a related note, I also get >> >> boost/numeric/bindings/lapack/driver.hpp:12:51: error: boost/numeric/ >> bindings/lapack/solve.hpp: No such file or directory > > Should be fixed, too. Please note that the uplo stuff for sysv will > most > likely change. > > Cheers, > > Rutger > > > > _______________________________________________ > ublas mailing list > ublas@... > http://lists.boost.org/mailman/listinfo.cgi/ublas > Sent to: um@... _______________________________________________ ublas mailing list ublas@... http://lists.boost.org/mailman/listinfo.cgi/ublas Sent to: lists@... |
| Free embeddable forum powered by Nabble | Forum Help |
