From: <ba...@ne...> - 2012-08-12 16:47:13
|
Hi Boris, hi Kilo, i'm not a master about code things, but i'd like to test it. So can someone provide a 64 bit binary for Linux instead of patches? Cheers Eckhard Zitat von Boris Brönner <bor...@gm...>: > Hello, > > I started implementing a solution for that bug, not finished yet, > though. Up to now I decoupled object splitting and texture collapsing > (I talked to Kilo to verify that) and exchanged their execution order > (as discussed in the bug report in the last post, solution part b) > http://sourceforge.net/apps/trac/speed-dreams/ticket/691). Now however > I have to extend splitOb() for the additional texture channels. > > That's not the problem, though. I want to improve the readability of > accc a little bit. Not of the whole accc, but at least at the parts I > have to touch for that bug, so that at least there I'm the last one > that has to depuzzle those parts. I know there are more important > problems to solve, but well, it's fun and my own freetime so I figured > I just go ahead. I don't know, though, if you guys are interested in > that at all, so attached you find the improvements I've done so far, > mainly just stylish improvements and very small refactorings. > > > Please tell me, if that's fine! If not, I just undo all the changes, > fix the bug directly and do some more important problems. > > > Also, I stumbled over those lines in splitOb() and they're odd. > > vatmp = (tcoord_t *) calloc(4 * num, sizeof(tcoord_t)); > pttmp = (point_t *) calloc(4 * num, sizeof(point_t)); > oldva = (int *) calloc(4 * num, sizeof(int)); > snorm = (point_t *) calloc(4 * num, sizeof(point_t)); > text = (double *) calloc(3 * num, sizeof(double)); > > I went over the whole function and looked at the accesses to those > arrays and imo all but the last line should have 3 * num fields and > the last line 2 * num fields. > > Also I ran valgrind over accc and his has 28 MB of memory leaks in the > debug version, but I guess to solve that, we'd have to rewrite all of > accc, which is probably too much work. Anyway, attached you find the > memcheck output of valgrind. I suggest to at least report a bug on > this issue, being it with very low priority. The statistic is done by > executing the following command on the debug version of 4809. > > sd2-accc -g allondaz-gen.acc -l0 allondaz-src.ac -l1 > allondaz-shade-src.ac -d3 1000 -d2 500 -d1 300 -S 300 -es > > So, that's it, tell me what you think :) > > (@Kilo: I removed the attachment and copied it to dropbox in case you wonder) > > Here's the attachment, that was too large for SD's mail server: > http://dl.dropbox.com/u/11989467/mail_speed-dreams.tar.gz > > Regards, > Boris > > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ > Speed-dreams-devel mailing list > Spe...@li... > https://lists.sourceforge.net/lists/listinfo/speed-dreams-devel |