<p dir="ltr"><br>
On Aug 9, 2015 20:15, "Matan Ziv-Av" <<a href="mailto:matan@svgalib.org">matan@svgalib.org</a>> wrote:<br>
><br>
> On Sun, 9 Aug 2015, Shachar Shemesh wrote:<br>
><br>
>> I should point out that the simplified matrix I was using to prove to myself that the idea has merit is 20x20 (inversed using state of the art in the vi text editing front). The code<br>
>> I'll actually have to run will be 273x273. The Matrix itself is sparsish, so it's really not that bad. Still, to prove things are working I'll have to (probably pre-calculate) several<br>
>> thousand such matrices.<br>
>><br>
>> In fact, one of my problems in using tools such as Mathematica and friends is that I don't want to hand code the Matrix to be inversed.<br>
><br>
><br>
> GAP (and similarly SAGE, maxima) can be scripted.<br>
><br>
> You do not give your exact setup, but if you can write the matrices to a text file in any format, you can read it in GAP. If you can output to the same format as I posted, you do not even need to write any program in GAP to read it.<br>
><br>
Not being connected to the field of applied mathematics can only recommend solutions observed.<br>
There is vast amount of code already written for LinAlg.<br>
in python you can use numpy.<br>
In C(++) any BLAS library have all you need.<br>
OpenBLAS seems more effective.<br>
If yo need to manipulate matrices "en masse" strongly suggested to use GPU.<br>
Both OpenCL and CUDA are good.</p>
<p dir="ltr">Best Regards, Evgeniy.</p>