[cello-devel] Speed Week!

[Kenny Tilton]

I got a report of Cello running at just one frame per second* from 
someone running a pretty hot setup on win32 under LW 4.2. At the same 
time I had noticed it would be quite easy to eliminate a ton of GF 
dispatch (one thousand per frame!, mostly on slower 'eql 
specializations), and that there was an excellent chance Cello could 
make extraordinary advantage of opengl display lists, because it turns 
out the cello graphics scheme maps almost perfectly onto the opengl 
scheme (a tree of nodes, each with its own coordinate system relative to 
its parent node), so I am having fantasies of a thousand widgets running 
as fast as the hardware frame rate upper bound.

We will see. Eliminating the GF dispatch is a no-brainer, but the 
display list win sounds too good to be true. Stay tuned.

Oh, and on the light panel with the spinning shape, I could also simply 
put the spinning shape in a subwindow and then it is the only thing 
redrawn, in which case I would not even have to do a Speed Week. But as 
long as I have a good stress test on which to base a Speed Week, I may 
as well take advantage of it. (OK, I enjoy Speed Weeking.)

kt

* I get 10fps on just a 50% faster CPU, and I /thought/ I saw LW4.2 
running at the same rate when I tested there, but with two huge 
performance wins sitting in my lap I would rather just implement those 
and then look at the deets of why the user is running so much slower 
than I. Possibly I am wrong about LW performing as well, and if GF 
dispatch is just a little slower on LW those 1000 dispatches could 
explain a lot. And maybe I have a graphics card that is especially fast 
or just better at some aspect of opengl such as textures.

-- 
Home? http://tilton-technology.com
Cells? http://www.common-lisp.net/project/cells/
Cello? http://www.common-lisp.net/project/cello/
Why Lisp? http://alu.cliki.net/RtL%20Highlight%20Film 
Your Project Here! http://alu.cliki.net/Industry%20Application