[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Make sparse automatic mutation to dense matrices optional
From: |
David Bateman |
Subject: |
Re: Make sparse automatic mutation to dense matrices optional |
Date: |
Mon, 26 Nov 2007 23:22:02 +0100 |
User-agent: |
Thunderbird 1.5.0.7 (X11/20060921) |
John W. Eaton wrote:
> On 26-Nov-2007, David Bateman wrote:
>
> | John W. Eaton wrote:
> | > On 16-Nov-2007, David Bateman wrote:
> | >
> | > | While looking at SeDuMi one issue I had was that much of the code
> | > | assumes that once a matrix is sparse it will stay that way and it so
> | > | SeDuMi just assumes that certain matrices are sparse in some of its
> | > | internal mex functions. Octave's habit of converting sparse matrices to
> | > | full ones if that saves space, plays havoc with this.
> | > |
> | > | I therefore propose to include an internal variable Vsparse_auto_mutate
> | > | that can be used to enable or disable this functionality and that the
> | > | "-traditional" option turns this off as that is what matlab does.
> | >
> | > I think this patch is OK, but maybe we should set sparse_auto_mutate
> | > to false by default for compatibility.
> | >
> | > jwe
> |
> |
> | It is for matlab compatibility mode, but not in general. Do you want it
> | to be off by default always?
>
> It doesn't matter much to me, but looking at the rest of the options
> implied by --traditional, the only one that is not cosmetic seems to
> be
>
> bind_internal_variable ("default_save_options", "-mat-binary");
>
> so maybe it would be best to have sparse_auto_mutate false by
> default as it can change the behavior of valid Matlab code, so I think
> having the default be compatible would eliminate some confusion for
> Matlab users who don't expect sparse matrices to suddenly become
> full.
>
Yeah, there were already some complaints, though I prefer it to be true.
Ok, I checked it in as off by default.
D.