Well, let me try. Sometimes it's neccessary to make dot product of big
numerical arrays. So, maybe it's possible to implement fast version of
it with NativeBoost.
This is simple to do: multiply pairs in loop and collect the sum :)
Do you mean it's easy with NativeBoost? My doubt is accessing of smalltalk's
collections from provided low-level code. Another problem is how to prepare
this low-level code, how to put it inside the method and so on. Something like
"Hello World" for NativeBoost.
I don't remember an algorithm for multiplying matrices of arbitrary size.
I was taught it in university but it was long time ago :)
Matrix algoritm is not that hard to find...wikipedia, you know. Generally speaking,
if dot product solution will be demonstrated, I think, matrix multiplication will be
easy.
... and it converts it to utf8 encoding before pushing it to function.
What function do you mean? One of smalltalk object's methods or something
other?
> > Well, let me try. Sometimes it's neccessary to make dot product of big
> > numerical arrays. So, maybe it's possible to implement fast version of
> > it with NativeBoost.
> >
> This is simple to do: multiply pairs in loop and collect the sum :)
Do you mean it's easy with NativeBoost? My doubt is accessing of smalltalk's
collections from provided low-level code. Another problem is how to prepare
this low-level code, how to put it inside the method and so on. Something like
"Hello World" for NativeBoost.
> I don't remember an algorithm for multiplying matrices of arbitrary size.
> I was taught it in university but it was long time ago :)
Matrix algoritm is not that hard to find...wikipedia, you know. Generally speaking,
if dot product solution will be demonstrated, I think, matrix multiplication will be
easy.
> ... and it converts it to utf8 encoding before pushing it to function.
What function do you mean? One of smalltalk object's methods or something
other?
