Simon -- > To be honest, I'm a bit unhappy about all this: > > runops_t0p0b1_core, > runops_t0p1b0_core, > runops_t0p1b1_core, > runops_t1p0b0_core, > runops_t1p0b1_core, > runops_t1p1b0_core, > runops_t1p1b1_core
I don't blame you. Its an interim solution with a bigger idea behind it that hasn't materialized yet. See below. > My feeling is that you get a fast runops core for normal use and > a slow runops core which does everything else. If you're in > "debugging mode", then an extra check to see whether or not you're > profiling doesn't make much difference. As such, unless I hear a > *very* good argument for duplicating and maintaining essentially > the same code 7 times over, I'm going to rip out the slow cores and > replace them with one single one. As the guilty party who wrote those, I don't have a problem with them being replaced, but the intent was that we wouln't be maintaining them individually, but rather we'd use a template system to derive them at build time. Its likely that for the current set a division into one fast and one slow would be sufficient, but I wanted to change this into a generic mechanism that could support a lot of dimensions of variability. For example, suppose someone comes up with an "event-handling" core that checks for events. This would still need to be as fast as possible. I'm not sure its fair to charge that core also for the conditionals and code size of all other options. So, I'm not arguing for duplication (which is what we have today), but I guess I am arguing for the mechanism that would allow a smart choice among a set of cores that might be larger than two, combined with the ability to use a template mechanism to derive most cores from the basic loop and a small amount of info about how the core differs from the unadulterated one. Whether or not that constitutes a "*very* good argument" for maintaining the intent but putting in a real implementation vs. ripping it out and going to two cores is open for debate, I suppose. Regards, -- Gregor _____________________________________________________________________ / Inspiration >> Innovation >> Excellence (TM) \ Gregor N. Purdy [EMAIL PROTECTED] Focus Research, Inc. http://www.focusresearch.com/ 8080 Beckett Center Drive #203 513-860-3570 vox West Chester, OH 45069 513-860-3579 fax \_____________________________________________________________________/
