From: Frank K. <fbk...@ne...> - 2000-08-25 02:11:44
|
h-peter recktenwald wrote: > Pse, take my thaughts just as some suggestion. I don't have any idea > whether it would be worth the programming effort or, if I would simply > be > expecting too much. No harm discussing what a "perfect Nasm" would be like, just to have a clear "compass" to what direction we're heading. I really like some of your ideas. Since optimization involves multiple passes, doing it on a "per-section-of-code" might be a problem, but it *might* be feasable to devise a directive to instruct Nasm to skip actually making any changes (to a section) even though it's doing the pass (?). But, first things first... Last patch I've seen (0.98.06b, but nasm.c 1.7 - I assume that's what's referred to in the subject) still generates incorrect code for "objtest.asm" - the redefine messages are gone, but the result "hangs" before printing either "hello" message - similar to what 0.98.05 was doing (IIRC). Wish I could help more with this, but I'm still finding John's changes "challenging" - I'll keep studying it. Latest changes to the web-site have caused the "mailing-list archives" to vanish. This is somewhat bothersome to me right now, as my "mailbox" is being flukey. Not a big problem. The greatest improvement to the web-site I can imagine will be the "download the new version here" button - I don't give a damn if it's "nuked" or "noped". ( if this is supposed to be a "chat forum"... there's some code at http://www.agner.org that might be worth looking at - I don't know how it works, but I imagine it's the *fastest* chat forum going!) More later, Frank |