On May 22, 3:17*am, "Skybuck Flying" wrote:
"Chris S." wrote in message

...



- Direct Compute is Microsoft's own CUDA and I think it's different
from -
- NVIDIA's CUDA... not sure though. -

I am not at all into any of this, but reading about it I became
convinced that you are mixing up the two concepts: DirectCompute is
not another version of CUDA - CUDA is a Chip technology and
DirectComputing is the API for programming such a chip. They belong
together - they would be completely useless without each other, but
different concepts.

If you know about terms like High and Low level language, you will see
that these are levels of abstraction away from the chip - I am sure we
all remember the days when everybody tried to beat everybody else over
the head with their own 'new' application that was often referred to
as being developed by way of programming some device 'directly' (i.e.)
by subverting the BIOS of that device. It could always be shown to be
fast, but sadly not a very stable way of handling things, because
everybody had to invent the soup-bowl from scratch every time.

So now, it would seem, we are still being beat over the head by a
completely new level of abstraction that might actually turn out to be
useful. High level Programming Language using a layer of abstraction
that is designed to give Low level instruction to a chip that is
designed to specifically execute such instructions (perhaps not using
any BIOS support at all?) Then probably co-operating with a standard
partner-chip that supposedly is relaying on its BIOS.?

Gee, Skybuck, I also don't realy know, but this I am sure of, we are
dealing with a special sort of API (really a CPI) designed to instruct
a special sort of chip. So, different concepts altogether.


Tony. . .