| --- Log opened Sun Feb 01 00:00:22 2015 | ||
| -!- Netsplit *.net <-> *.split quits: LoneTech | 01:43 | |
| -!- Netsplit over, joins: LoneTech | 01:48 | |
| -!- Netsplit *.net <-> *.split quits: sfa_, arokux1, julzmb, heroux | 08:45 | |
| -!- heroux_ is now known as heroux | 08:45 | |
| -!- Netsplit over, joins: julzmb | 08:46 | |
| -!- ysangkok_ is now known as ysangkok | 10:29 | |
| olofk | Anyone at FOSDEM btw? | 10:41 |
|---|---|---|
| olofk | Anyone here handy with mmap? | 13:37 |
| knz | olofk: perhaps I am | 15:04 |
| rschmidlin | Hey guys, I am not able to synthesize mor1kx with ISE. Is it because some files are only System Verilog compliant? | 18:15 |
| olofk | knz: Well, let's give it a try :) I have a peripheral mapped to the address space that wants to read and write to another part of the address space (DMA). I'm thinking that I could mmap both the address space of the peripheral to read and write control/status registers, and the memory area just as would have done if I was running it without an OS. Will that work? | 19:30 |
| olofk | rschmidlin: I haven't had any problems with that. I built the Atlys port just a few days ago without any problems | 19:31 |
| olofk | rschmidlin: Can you paste the errors? | 19:31 |
| rschmidlin | There were errors of localparams inside generates. Ise didn't like it. | 19:41 |
| rschmidlin | I could paste in a moment. The automatic system had some problems too. It was not able to set the project to my device. Well i have to go over my files to check of everything is alright. I just installed version 14.7. | 19:43 |
| rschmidlin_ | The other problem with mor1kx-generic was due to r0 not being initialized at all. The implementation seems to rely on the fact that FPGA rams are always initialized to zero. | 19:46 |
| rschmidlin_ | olofk, http://pastebin.com/EfHQzvkC | 19:49 |
| olofk | rschmidlin_: Ah right. True. I fixed the localparams inside generates locally here. Forgot about that | 19:51 |
| olofk | And I have local fixes for the register init problems as well. Never done a proper patch though :/ | 19:51 |
| rschmidlin_ | And I have this odd behavior with fusesoc and ise: http://pastebin.com/04N2FXsn | 19:52 |
| rschmidlin_ | olofk, I fixed that one too. I think the proper fix would be to fix assign r0 to zero. | 19:52 |
| rschmidlin_ | I will be back in some minutes… | 19:53 |
| olofk | Yeah, that error was weird | 19:53 |
| rschmidlin_ | olofk, especially odd if you consider that I have exactly that configuration in the prj file. | 19:53 |
| olofk | And regarding r0, there has been a lot of discussion about that, and the conclusion is that it's software's responsibility to ensure that it is 0 | 19:53 |
| rschmidlin_ | aha | 19:54 |
| olofk | rschmidlin_: Can you open the project file that fusesoc creates in ISE? | 19:54 |
| rschmidlin_ | olofk, well, then at least r0 has to be initialized to zero | 19:54 |
| rschmidlin_ | olofk, I didnt try it | 19:54 |
| rschmidlin_ | could you provide a patch for those localparam things? | 19:54 |
| olofk | Sure. Just have to fire up another computer. Give me 15 minutes or so | 19:55 |
| rschmidlin_ | and I had that odd clog2 missing in arbiter.v. When I include verilog_utils.vh, ISE seems to go completely crazy then. | 19:55 |
| rschmidlin_ | olofk, yup, I’m back in a couple of minuts too | 19:55 |
| olofk | rschmidlin: Sorry, but you will get them later tonight | 20:14 |
| rschmidlin | Back | 20:24 |
| knz | olofk: of course I assume you have no iommu, so you really need the physical mappings not the virtual ones | 22:33 |
| knz | olofk: my reference in this situation is http://www.xml.com/ldd/chapter/book/ch13.html | 22:35 |
| knz | in short it is likely you need to make a kernel module and play with __va/__pa and/or kiobufs | 22:38 |
| knz | also possibly check http://stackoverflow.com/questions/3333959/mapping-dma-buffers-to-userspace | 22:40 |
| --- Log closed Mon Feb 02 00:00:23 2015 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!