Hi all, Today is the "pencils down" date for Googles Summer of Code. I've cleaned up the loop analysis, LCSSA, and range propagation work I've done, and pushed them to my github repo at [1]. They are located in the branches gsoc-final-XXXX
I'll leave them as is for a while, so that people have a chance to review them if that is of interest. It's still not 100%, time sorta ran away from me. But it should be at a state where the chosen approach is settled. (At least in my mind, as long as it is not completely dumb, that is). The range propagation pass "Works as expected (TM)" and so does the loop analysis pass. The LCSSA conversion pass is hitting some issues described in a lengthy comment in the top of the file. tl;dr; There's some assumptions we make about the successors and predecessors that are not valid when you insert phi-nodes like is done in the LCSSA pass. Now I'll focus on getting the last formalities done with, and then I'll probably be "off the chart" for a week or two. There's the whole "getting the new semester on track" thing happening atm, with selecting projects, job as a student counselor, changing to a different gym, getting the master thesis settled, catching up time with everyone, cleaning the freakin' apartment, etc. It's been a challenging but fun experience. Thanks to everyone for being patient with me and helping me out when needed. A special shout-out goes to Matt, Ilia, Jason, Kenneth and Connor for the irc-support, code-review, and the good discussions. [1] https://github.com/thohel/mesa Best regards, Thomas Helland _______________________________________________ mesa-dev mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/mesa-dev
