> The thing is, taken as a few cycles, it doesn't matter much. But taken > over the sum total of everything that the machine does, these cycles > ultimately consume time and power. Some of this stuff is in critical > paths. We can do better.
I think it is precisely for the reason that the machine does so many things that the overall effect of optimizing single-address swapping functions will be negligible. And if you need to byte-swap massive amounts of data, you'd probably want to write the entire "while ddi_swap32" loop in assembly, and not just the loop body. OTOH, in terms of ROI, optimizing ddi_swap* is a very small "I", so we might as well just do it. -Artem _______________________________________________ opensolaris-code mailing list opensolaris-code@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
