| --- Log opened Wed Jan 28 00:00:16 2015 | ||
| -!- radii_ is now known as radii | 02:18 | |
| olofk_ | I see that openrisc.net is back online | 06:43 |
|---|---|---|
| -!- olofk_ is now known as olofk | 06:43 | |
| sb0_ | hi | 09:09 |
| olofk | hi | 09:09 |
| sb0_ | trying to build or1k binutils | 09:09 |
| sb0_ | gcc -g -O2 -o sysinfo sysinfo.o syslex_wrap.o | 09:09 |
| sb0_ | syslex_wrap.o: In function `main': | 09:09 |
| sb0_ | sysinfo.o:/home/sb/or1k-src/build/binutils/sysinfo.c:1: first defined here | 09:09 |
| sb0_ | is that something you are familiar with? | 09:10 |
| olofk | sb0_: Which build instructions do you use? | 09:10 |
| sb0_ | http://m-labs.hk/artiq/manual/installing.html | 09:10 |
| olofk | Don't know about the error unfortunately | 09:10 |
| sb0_ | it works fine on arch linux, not on linux mint apparently | 09:10 |
| olofk | Don't know if it helps, but you could try the build instructions here: http://opencores.org/or1k/OpenRISC_GNU_tool_chain | 09:11 |
| olofk | I updated the instructions for the newlib and musl toolchains to use the upstream binutils instead of or1k-src | 09:12 |
| sb0_ | looks like this http://virtuallyfun.superglobalmegacorp.com/?p=4911 | 09:14 |
| sb0_ | stupid error message indeed | 09:14 |
| sb0_ | recompiling ... | 09:14 |
| olofk | Fantastic domain name :) | 09:14 |
| olofk | ah ok. Yes. Missing bison or flex has given me weird errors before | 09:15 |
| sb0_ | works now | 09:16 |
| olofk | great | 09:16 |
| wallento | olofk and sb0_: I updated also those http://openrisc.github.io/newlib/ | 10:59 |
| wallento | upstream continuous buils still fail due to "connection reset by peer" | 10:59 |
| olofk | wallento: You really killed sourceware :( | 12:38 |
| wallento | I feel guilty | 12:41 |
| olofk | I got a little tricky Linux question. I have a board with two completely DDR2 memories. Can I somehow use one of them as a normal memory and the other one exclusively as a DMA buffer? | 15:01 |
| ysionneau | I guess it depends on how you layout the A[] connections | 15:02 |
| ysionneau | if you can have something where the low addresses of the SoC are all mapped to the 1st chip , and then higher addresses on the second chip | 15:03 |
| ysionneau | then you can declare those memory regions in Linux as DMAable / high memory | 15:03 |
| olofk | That sounds like what I want to do | 15:04 |
| ysionneau | and then you increase the burst size to compensate the fact that you have less bits out per cycle since you now only talk to one DDR2 chip at a time | 15:05 |
| olofk | They already have separate control pins | 15:06 |
| olofk | So I just need to find out how to do it in Linux | 15:06 |
| ysionneau | s/burst size/burst length/ | 15:06 |
| ysionneau | https://www.kernel.org/doc/gorman/html/understand/understand005.html <= | 15:07 |
| ysionneau | something like that (zones) | 15:07 |
| ysionneau | but I guess this page is not up to date anymore | 15:07 |
| Me1234 | Why does linux compiled with or1k-linux-gnu works only on qemu? (Does not work on or1ksim). Is it because of atomic | 15:11 |
| poke53282 | Which QEMU version? The official one? | 15:30 |
| poke53282 | or the one from blueCmd? Or the static one from the or1k Debian website? | 15:32 |
| poke53282 | the patches for QEMU for the atomic instructions were accepted today. | 15:33 |
| poke53282 | or1ksim is the reference implementation and should always work. | 15:55 |
| -!- clopez_ is now known as clopez | 16:34 | |
| stekern | Me1234: if you use or1ksim from current head, atomic instructions are supported | 17:02 |
| stekern | olofk: which peripheral do you want to DMA with? | 17:04 |
| Me1234 | Ok, it was just linux configured for 1GB ram in the dts. | 17:44 |
| --- Log closed Thu Jan 29 00:00:17 2015 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!