On 2/7/13 10:54 PM, Bruce Sherwood wrote:
To repeat, Windows for my 3D graphics development purposes has been
far more stable than either Mac or Ubuntu Linux.
Windows is the biggest market for gamers. 3D innovation has
historically always been first on WIndows.
If all you want a computer to do is a fixed set of 2d and 3d graphics
APIs, then, sure, use Windows. But performance and stability are only
two dimensions.
I care much more about flexibility than stability or graphics
performance. For example, I want to use GPUs for accelerated
computation. It is inappropriate in my situation to code using
unportable (CUDA) or crudely simple APIs like OpenCL. That's no way to
write complex, long-lived, maintainable software. It could be a way
to write simple, static, scientific codes that perform on particular
cards, if that's all you need to do. I want the possibility of *some*
acceleration over generations of cards, not peak performance for one
generation.
AMD GPUs on Linux now have the driver bits (in Mesa, a free OpenGL) and
compiler bits in LLVM (a free compiler). Together there's now the
possibility of integrating real compilers with accelerator technology.
On Windows, this kind of integration and experimentation is not possible.
Now fast forward to the day this all just works. Someone writes a code
using these compiler tools, but, oops there's a strange anomaly in a
particular calculation. How do you fix it? Get your favorite
bloggers to complain in a public setting? No thanks, I want direct
control. That means source code.
Marcus
============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
