| --- Log opened Sat Jul 20 00:00:34 2013 | ||
| poke53281 | stekern: Made a small benchmark on qemu and or1ksim. | 00:39 |
|---|---|---|
| poke53281 | QEMU is around 3.7 times faster in computation speed than or1ksim. | 00:39 |
| poke53281 | Actually I expected more. | 00:45 |
| poke53281 | Did the test with some check scripts for the bc calculator | 00:46 |
| poke53281 | But I have here a reproducible effect. bc crashes on QEMU when I try to start it. On or1ksim no problem so far. I use exactly the same image for both. If I start it with "bc -l" no problem in qemu. | 01:00 |
| poke53281 | If I link it statically it works without problems in both emulators. | 01:00 |
| poke53281 | Maybe it is worth looking after it. | 01:01 |
| poke53281 | I mean to take a closer look | 01:03 |
| poke53281 | Damn, it's a Heisenbug | 01:14 |
| stekern | those are the worst :( | 04:44 |
| stekern | and this time the gcc suite hasn't crashed | 04:51 |
| poke53281 | perfect, also finally the torture test for the varargs was 100% succesfull | 06:52 |
| stekern | now it did crash | 06:56 |
| stekern | I need to find a better way to reproduce that | 06:56 |
| poke53281 | qemu or or1ksim? | 06:58 |
| stekern | qemu | 06:58 |
| stekern | I'm not sure how I should move forward with the hw tlb reload neither | 06:59 |
| poke53281 | well, I have here a repoducible effect with QEMU. Unfortunately it is gone when I write a lot of printf's in the source code. | 06:59 |
| stekern | to do it according to the spec, I would need to use the permission index register instead | 06:59 |
| stekern | of just using the PTE from the kernel | 07:00 |
| stekern | the pte from the kernel set the exec bit in the lowest PPN bit in the pte defined by the spec | 07:02 |
| poke53281 | Difficult to say for me. Have not analyzed that part from the spec. I found it confusing when I read it the first time before I saw that we use software-refill. | 07:04 |
| poke53281 | So I would ignore the spec. :) | 07:05 |
| stekern | yeah, the mmu part of the spec is probably one of the worst parts | 07:05 |
| stekern | I'll at least do the 'proof-of-concept' with the linux pte first in mor1kx | 07:06 |
| stekern | (the problem with ignoring the spec is that then it's on your shoulders to update the spec and push the changes through ;)) | 07:08 |
| -!- Netsplit *.net <-> *.split quits: blueCmd | 08:30 | |
| -!- Netsplit *.net <-> *.split quits: blueCmd_ | 08:54 | |
| -!- Netsplit *.net <-> *.split quits: Amadiro, heroux, blueCmd, LoneTech, olofk, jakob_, poke53281, trevorman, forkG, stekern, (+1 more, use /NETSPLIT to show all of them) | 09:06 | |
| -!- Netsplit over, joins: blueCmd, trevorman, heroux, poke53281, LoneTech, Amadiro, jakob_, stekern, forkG, enghong (+1 more) | 09:07 | |
| -!- Netsplit *.net <-> *.split quits: simoncook | 09:08 | |
| -!- Netsplit over, joins: simoncook | 09:10 | |
| -!- Netsplit *.net <-> *.split quits: hle__ | 09:24 | |
| poke53281 | What is the result of the testsuite run? | 18:28 |
| stekern | poke53281: hmm, which of the runs? | 19:29 |
| poke53281 | the last gcc run | 19:49 |
| stekern | with qemu? or the last with or1ksim? with 4.8.1 or 4.9.0? | 20:01 |
| poke53281 | 4.8.1 would be nice. if possible with or1ksim. I don't trust qemu | 20:29 |
| poke53281 | Just want to see what is failing right now. At the moment I am working on the crash of X. But it is like looking for a needle in a haystack. At least I am making progress in understanding X | 20:31 |
| stekern | poke53281: http://lists.openrisc.net/pipermail/openrisc/2013-July/001788.html | 20:48 |
| poke53281 | damn, it is in my mail folder | 21:27 |
| poke53281 | and now I would like to have the list which files are failing. | 21:28 |
| poke53281 | "grep FAIL gcc.sum" or something like that | 21:30 |
| poke53281 | Just want to take a closer look :) | 21:37 |
| stekern | poke53281: http://oompa.chokladfabriken.org/tmp/gcc.log | 21:46 |
| stekern | http://oompa.chokladfabriken.org/tmp/g++.log | 21:46 |
| poke53281 | thanks | 21:48 |
| stekern | I put the .sum there too, the .log are pretty big | 21:48 |
| poke53281 | Ok, the sum is the only one I need | 21:48 |
| --- Log closed Sun Jul 21 00:00:36 2013 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!