I experience a strange behaviour with ARB_fragment_program on a Radeon
9800...
My program is around 30 instructions long. It took twice the time to
execute this program if there is no texture instruction inside.
If I add a neutral texture instruction (no indirection, and the result
of the instruction is not use by the subsequent ALU instructions), the
program is appreciably faster...
Is anybody else experience such kind of problem with fp ?
Anybody from ATI around want to comment on this ?
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Mac-opengl mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/mac-opengl/email@hidden
