| andresjk | hi | 02:56 |
|---|---|---|
| stekern | another bug detected | 09:29 |
| stekern | if an exception hits while reading SR, the wrong SR will be written into the register | 09:30 |
| LoneTech | I just realized I might as well use the delay slot to store the number directly instead of the jal method | 12:29 |
| * LoneTech comes up with overcomplicated solutions sometimes | 12:29 | |
| -!- X-Scale` is now known as X-Scale | 13:33 | |
| blueCmd | is it valid to do: | 15:08 |
| blueCmd | l.bf | 15:08 |
| blueCmd | "instr that sets flag" | 15:08 |
| blueCmd | in order to use the delay slot? | 15:09 |
| blueCmd | I seem to recall that some archs require the flag to be set before any delay slots | 15:09 |
| LoneTech | I believe the spec forbids it unclearly. in theory it should mean the delay slot does not affect the branch decision | 15:30 |
| blueCmd | LoneTech: right, i will skip it and just do l.nop | 15:31 |
| stekern | I'd say it's valid | 16:02 |
| stekern | and I think gcc emit's that kind of sequence | 16:03 |
| stekern | but, the flag that will be used is of course the one that is before the l.bf | 16:04 |
| LoneTech | yep, that's what I think happens | 16:05 |
| LoneTech | the related actual conflict is that one mustn't do l.jalr r9, as jumps may be reexecuted when an exception interrupts the delay slot instruction | 16:06 |
| stekern | yeah, that's nasty ;) | 16:08 |
| stekern | but the delay-slot biting works the other way around, I found myself interpreting the instructions after branches in some avr asm as delay-slots a while ago... | 16:13 |
| blueCmd | stekern: heh :) | 16:21 |
| LoneTech | yep, it has many ways to bite | 16:21 |
| blueCmd | i wonder how "fast" openssh would be running in or1ksim? | 16:22 |
| LoneTech | no worse than on my sparcstation elc, I am guessing | 16:22 |
| blueCmd | anybody want a nice static zsh for or1k? ;) it's much better than busybox! | 16:28 |
| blueCmd | much slower as well, but hey - can't have everything :P | 16:28 |
| blueCmd | anyway, two syscalls left in eglibc to implement afaik. time to start working on dynamic linking. | 16:35 |
| stekern | mfspr bug fixed => three more linux console lines before it oopses ;) | 17:29 |
| stekern | progress nevertheless! | 17:30 |
| LoneTech | well done | 17:31 |
| stekern | http://pastie.org/5909564 | 17:32 |
| stekern | seems our linux port doesn't detect 2 way caches properly ;) | 17:35 |
| stekern | (completely unrelated to the oops) | 17:35 |
| stekern | http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=blob;f=arch/openrisc/kernel/setup.c;h=f4d5bedc3b4f72bc92a939d0cb84b71215fbd929;hb=949db153b6466c6f7cad5a427ecea94985927311#l131 | 17:36 |
| LoneTech | not surprising, or1200 did not have them iirc | 17:36 |
| stekern | hardcoded to 1 | 17:37 |
| stekern | it's actually just in the printout, the size is correctly calculated with the number of ways included | 17:37 |
| blueCmd | stekern: what are you working on? | 22:10 |
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!