#milkymist IRC log for Sunday, 2011-10-02

cdehai09:43
lekernelhi09:44
sb0hm, why is struct sched_ctx so big...20:24
wpwraksb0: it has a bunch of pointers and integers for each instruction21:05
wpwraksb0: on 64 bit, it's even almost half a megabyte :)21:06
wpwrakon 64 bit, each instruction has a 216 bytes data structure. and there are up to 2048 instructions ...21:09
wpwrak(FPVM_MAXCODELEN)21:09
wpwrakthe 216 bytes structure is  struct insn21:10
sb0ok21:21
sb0well, currently FN is only using about 25MB out of the 128MB of the M1, so we can bloat it a little :-)21:22
wpwrakhehe :)21:22
wpwrakah yes, a bigger stack helps :)21:23
wpwrakso the things in malloc.c don't work in the demo firmware ?21:24
sb0they work21:32
sb0you can use malloc() if you want... but try to avoid the non-reentrant global variable for the state21:33
wpwrakah, so you expect to use this from threads21:35
sb0using the stack is fine... if you have working code with the stack atm, don't change it21:39
wpwrakyeah, right now i have it with the stack. need to get rid of the global "sc", though21:40
sb0struct vm_reg regs[frag->nbindings-frag->next_sur]; ? :)22:49
wpwrakmmh ? dynamic arrays. available since C99 :)23:09
wpwrakbtw, i only tested with FN, not with the demo. still need to work on the prerequisites of it.23:12
wpwrakand now, off to the next barbecue :)23:12
--- Mon Oct 3 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!