| --- Log opened Mon May 27 00:00:16 2013 | ||
| stekern | thanks juliusb | 10:34 |
|---|---|---|
| juliusb | np :) | 10:39 |
| juliusb | i've finally fixed the flop cache thing | 10:39 |
| juliusb | just bringing everythign in line | 10:39 |
| juliusb | will do a push later | 10:39 |
| juliusb | I actually had a question about the new write-back signal from the decode stage | 10:39 |
| juliusb | ahh, it's just a renaming of the signal | 10:41 |
| juliusb | cool | 10:41 |
| juliusb | from rf_wb_o to decode_rf_wb_o | 10:41 |
| stekern | ah, yes | 10:44 |
| mor1kx | [mor1kx] juliusbaxter pushed 1 new commit to master: https://github.com/openrisc/mor1kx/commit/3e1de595ea8adcdd75edfb7cc748eadddafee168 | 10:52 |
| mor1kx | mor1kx/master 3e1de59 Julius Baxter: pronto: add cache made of flip-flops... | 10:52 |
| juliusb | nice, it passes the GCC tests, too, with teh flop-cache one | 12:52 |
| stekern | cool | 12:54 |
| juliusb | stekern: I'm now thinking of something for mor1kx-dev-env which we can /win 9 | 13:59 |
| juliusb | whoops | 13:59 |
| juliusb | we can already /win 9 in irssi, and that's nothing to do with mor1kx-dev-env :P | 14:00 |
| juliusb | I meant to say I'm thinking of a better system for regression testing in mor1kx-dev-env | 14:01 |
| juliusb | I've only given it a little bit of thought, but the immediate shortcoming we have at the moment is the ability to specify the parameters of the mor1kx-core | 14:02 |
| juliusb | s/ability/inability/ | 14:02 |
| juliusb | so I'm wondering if maybe we could do something like `ifdef MOR1KX_INSTANCE_FILE around the parameters on the core, and if so, `include a file we generate? | 14:03 |
| juliusb | I'm also thinking I will try to make something which allows us to do the regression testing in parallel, so build one model, set it running, then build the next, set it running | 14:04 |
| juliusb | same for RTL and verilator | 14:04 |
| juliusb | also, do better results collation, so have something which will timeout if a test hangs, and not stop if a test fails | 14:04 |
| juliusb | I think doing something like this in Python would be easiest | 14:05 |
| stekern | that sounds good, and is in line with thoughts I've had as well | 14:05 |
| juliusb | (regarding parameters on the core) I'm fairly certain there's no way to pass parameters at compile time, do you agree? | 14:06 |
| juliusb | cool, yeah I'd like a system where we can run perhaps a subset or the full thing for all mor1kx builds, get a nice little report out indicating various metrics | 14:07 |
| juliusb | instead of having the complexity of running a regression in Make files, have it in Python and just use Make for what it's best at - building stuff | 14:08 |
| -!- X-Scale`` is now known as X-Scale | 14:18 | |
| stekern | we could even use the python test framework ;) | 14:51 |
| juliusb | unittest — Unit testing framework?? | 14:53 |
| stekern | I think that's the one I'm thinking about, yes | 14:59 |
| olofk | Hey, I started writing a blog. olofkindgren.blogspot.com | 15:21 |
| olofk | Wait a minute. I see you're talking about python based regression testing of mor1kx. That's one of the things orpsocv3 intends to do | 15:22 |
| olofk | juliusb: Using plusargs, you can even change things at runtime. I'm using that to provide different test cases without recompiling the model. Works fine in both icarus and modelsim | 15:24 |
| juliusb | olofk: hiya!! | 15:25 |
| olofk | Looks like I need to catch up a little :) | 15:25 |
| juliusb | Indeed, python-based regression system | 15:25 |
| olofk | My goal is set on ripping mor1kx-dev-env to shreds :) | 15:25 |
| juliusb | olofk: ah right, plusargs, I thought about that, but I'm not sure of Verilator's support, and can we change parameters to the module with plusargs? | 15:25 |
| juliusb | olofk: bring it on :) | 15:25 |
| olofk | No, you can't change parameters with plusargs, but in some cases you actually don't need a parameter | 15:26 |
| olofk | And I'm fairly sure that you can change parameters at compile time at least | 15:26 |
| juliusb | with a switch on the command line? For verilator I'm fairly certain this can't be done... | 15:27 |
| juliusb | I don't mind generating a parameters file, though | 15:28 |
| juliusb | at least that way it's the same across all the models | 15:28 |
| olofk | You could be right that Verilator might be a bit different here. | 15:32 |
| juliusb | hmm, do we have the OR1K coremark code around anywhere? | 16:36 |
| stekern | I'm not sure, you sent it to me by e-mail I think | 17:07 |
| stekern | I can put it somewhere | 17:07 |
| stekern | http://oompa.chokladfabriken.org/openrisc/coremark_v1.0.tar.gz | 17:10 |
| stekern | olofk: nice blog | 17:21 |
| hno | stekern, any luck in hacking the AR100? | 18:21 |
| olofk | stekern: Thanks | 20:06 |
| olofk | stekern: What was your impression of the workshop btw? | 20:10 |
| --- Log closed Tue May 28 00:00:18 2013 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!