On Sat, 25 Jun 2011 01:11:41 +1000, Bogdan Opanchuk <[email protected]> wrote:
> Hello Andreas,
> 
> On Sat, Jun 25, 2011 at 12:34 AM, Andreas Kloeckner
> <[email protected]> wrote:
> >> Are you planning on adding
> >> similar feature to PyCuda, just for the sake of symmetry?
> >
> > Sure--if you can think of some, let me know.
> > Depending on how obscure they are, I might leave them undocumented.
> 
> Well, by 'symmetry' I meant implementing OpenCL-style kernel call in
> PyCuda, where the product of grid dimensions is equal to the total
> number of threads (like CL's global work size) and block size is
> optional (and if it is not set, PyCuda will try to find the suitable
> one, I am guessing with the help of occupancy calculator and number of
> registers available). It can be convenient if all threads in a kernel
> are independent.

Oh, sorry. I misunderstood. I'm actually not planning on doing what you
describe. What I'd like is to get towards a shared subset of
functionality that makes it less annoying than currently to write code
that talks to both packages.

Andreas

Attachment: pgpol15yNehXc.pgp
Description: PGP signature

_______________________________________________
PyOpenCL mailing list
[email protected]
http://lists.tiker.net/listinfo/pyopencl

Reply via email to