30 Reputation

One Badge

1 years, 97 days

MaplePrimes Activity

These are questions asked by johnksellers

I discovered the LinearAlgebra package can use CUDA, 

CUDA, the Compute Unified Device Architecture, is an extension of C created by nVidia that allows the programmer to take advantage of the massive parallel computing power of an nVidia graphics card to do general purpose computation.

I also discovered that my new GTX 1660 Ti graphics card has 1,536 CUDA Cores, and the Maple CUDA package allows Maple to use the GPU hardward to accelerate certain LinearAlgebra routines.

The Maple command calling sequence is

CUDA[command]( arguments )

command( arguements )

But that is all I know.

Do any of you know just how much Maple can do with this sort of thing?  I've only had my individual license a couple of weeks so I don't have any perspective on what this means for Maple.

Can a lot of Maple make use of this sort of thing, or just a little bit?

Are there any other facilities in Maple that can make effective use of CUDA or any multi-processing like fully engaging all the coomputers CPUs?

I am interested regardless where the capabilities are implemented since enhancement can come from Maple, extensions, the Windows OS, the CPUs, or grahics card.... or any combination of these.

I rescently downloaded and installed my new Maple 2020 on my Windows 10 64x computer.

I noticed this in the install log:

   Log started 06/13/2020 at 03:25:42
   Preferred installation mode : win32
   Trying to init installer in mode win32
   Mode win32 successfully initialized
   Preparing Check for Update Tool
   Checking for Maple 2020 Updates

My question is this:  I know that under windows 10,  32 bit progams will generally install on even if a computer is 64x.

So should have Maple installed under mode win32?  Is there a 64 bit mode that will take better advanage of the system?

For that matter, is there any configuring that can install it or configure it to take better advanatage of the multi-threaded hardward capabilities?

Thanks, JS

( edited after the fact

BTW I tried to give your comment, kernelopts(wordsize) as the best answer, but it wouldn't let me because it is a comment.  So I gave your previous answer the credit.

kernelopts(wordsize) really hits it on the head for me.)

From a newbie:

Is there any way to get the Simpify from the Context Panel to clean up a little more?

I'm impressed with the job it does because it really did the job on some very hairy 5x5 matrices I had.

But I was not able to get it to completely collection all the terms of the following:

Clearly w,x, and y could have been collected like z was.  but I tried a number of approaches and could not get any of them to collect the w,x, and y terms..

Any thoughts about it?



I am having problems with VectorCalculus[Norm] with argument passing.

In certain cases I get the error message:

     "Error, (in VectorCalculus:-GetCoordinates) the first argument, when present, must be a Vector"

In all cases, the first argument IS a vector.

Here is the case I need to succeed:

VectorCalculus[Norm](DeleteRow(Vector[column](5, [x, y, z, w, 1]), RowDimension(Vector[column](5, [x, y, z, w, 1]))))

or more completely:

here is the test Document with several cases of the problem explored:

Background.  Calculus Vector Norm doesn't know about Homogenous systems so it inculdes an unwanted term in its norm.  Trimming off the last row of the vector fixes this problem.

(Note: as I am new, so if there is an existing cononical notation that will take care of that without further ado, then I could leave CalculusVector[Norm] to its own devices and move on.

Also, if I knew how to peek at the underside of such gotchas, I might have been able to sort it out myself.  Any leads in this direction would be taken to heart.  Thanks)

Page 1 of 1