IRC logs for #openrisc Monday, 2015-01-05

--- Log opened Mon Jan 05 00:00:24 2015
_franck__olofk: do you have a fix for that: ?10:33
-!- Netsplit *.net <-> *.split quits: blueCmd11:36
maxpalnolofk: in case you're there - I am making good progress on the addition of true Classic cycles and Constant Address burst cycles. Interestingly it uncovered (yet more) corner case bugs in my mem controller.12:36
maxpalnI am thinking of adding some sort of weighting system for the cycle types - in reality Classic and Constant Address cycles are pretty uninteresting for most of the debugging - I think a system that allows you to select the rough percentage of each type of cycle that the BFM generates would be better. Otherwise 1/3 of the transactions will be classics and another 1/3 constant address burst12:37
maxpalnolofk: I had to reimplement the init and next functions in bfm_master. The way they handled classic cycles seemed to be incompatible with zero wait states - something that only became visible once I started to exercise classic cycles. I have left the originals as-is (I think) and implemented a new function. That way any backwards compatability won't be broken....14:28
-!- blueCmd_ is now known as blueCmd15:08
aneesHi or1k! Trying to run or1k Linux image on qemu, the kernel panics with message:Kernel panic - not syncing: compression method gzip not configured15:43
anees---[ end Kernel panic - not syncing: compression method gzip not configured15:43
aneesAny solution to fix it.15:43
ysangkokanees: did you build your own kernel?15:49
ysangkokanees: where did you get your kernel configuration?15:50
aneesThe default one in the kernel source tree15:51
ysangkokanees: is CONFIG_KERNEL_GZIP=y?16:00
aneesNo in default configuration it is not set. Also in Kbuild file the config option  for gzip is no set to include in the menu selection.16:14
aneesIt seems that problem is with vmlinux, even including the option the kernel panics. how can I build with zImage. Make complains on specifying the zImage target.16:19
olofkmaxpaln: Great! I've also been considering a weighting system to distribute cycle types17:36
olofkBut there's one thing that I didn't like with the BFM improvements that I looked at earlier17:38
olofkThe ability to instantiate a master in a testbench to read and write config registers is lost, since the read function does a comparasion instead of just returning the value to the test bench17:40
olofk_franck_: Hmmm... I thought I had fixed that. Maybe forgot to push17:42
olofkmaxpaln: Also, I have been dropping the burst_type enum here in functions calls to just use cti and bte directly17:47
olofkI thought burst_type was a nice abstraction when I wrote it, but I'm not too sure about that anymore17:48
mor1kx[mor1kx] bandvig pushed 1 new commit to withfpu:
mor1kxmor1kx/withfpu 6e653f7 Andrey Bacherov: Replace separate multiplier and divisor by combined MUL/DIV module with shared multiplier.19:41
poke53282stekern: Do you use the 3.19 kernel already?21:46
poke53282Want to try the patched virtio in the new kernel.21:53
--- Log closed Tue Jan 06 00:00:25 2015

Generated by 2.15.2 by Marius Gedminas - find it at!