juliusb | an example of why I really want to get rid of the OR1K 256-byte exception vector default sizes, is that you have to waste at least 4k on them, and when you are limited to 32KB it's very annoying | 18:33 |
---|---|---|
jeremybennett | juliusb: If you really want to pack stuff in, you could interleave it in the spaces between the vectors. | 18:40 |
juliusb | jeremybennett: that sounds horrible! :-o | 18:52 |
stekern | I kind of agree, but it could of course be done | 18:52 |
juliusb | the architecture just needs to be fixed, and I'm sure it will be :) | 18:53 |
stekern | but if it'd be smaller I'd say it could be as small as 4 byte (without delay slot) | 18:54 |
juliusb | still, you need contiguous space for the linker to use it don't you? | 18:54 |
stekern | perhaps even should | 18:54 |
juliusb | I just think we need registers with the addresses | 18:55 |
juliusb | or | 18:55 |
juliusb | a table | 18:55 |
juliusb | just a pointer to a table | 18:55 |
juliusb | which is an address | 18:55 |
juliusb | err, which just has an address in it, offset for each exception | 18:55 |
stekern | well.. that is what a 4 byte exception ector would be | 18:55 |
juliusb | yes but at the moment one each has to live at 0x100, 0x200, 0x300 etc | 18:56 |
stekern | mmm | 18:57 |
* juliusb doesn't like that at all | 18:57 | |
juliusb | BTW, olofk and I will be going to FSCONS this year to present | 18:57 |
juliusb | it's in less than a month in Gothenburg | 18:57 |
jeremybennett | juliusb: I think this stuff is too deeply ingrained to change for OR1K. But you could make a case for having an exception vector table for OR2K. | 18:57 |
juliusb | jeremybennett: I agree, it's hard to change | 18:58 |
jeremybennett | Makes interrupts slower (one more indirection), but more compact. | 18:58 |
juliusb | well, it'd be easy to change in the spec, it'd be a pain to have two majorly different exception and reset handling models around the place | 18:58 |
juliusb | jeremybennett: you make the machine handle the redirection so it's not really slower | 18:58 |
jeremybennett | Maybe it's for OR3K then? | 18:58 |
juliusb | hah, no it's _definitely_ a system which'll be used in or2k | 18:58 |
stekern | well, interrupts will typically not fit in the 100 byte space anyways, so I don't think it'd be that big difference | 19:01 |
juliusb | yes, unless you write a particularly compact timer interrupt implementation for baremetal to make benchmark software run a bit faster *ahem* | 19:02 |
stekern | bah, you don't even want that timer interrupt enabled =) | 19:07 |
juliusb | where are people hosting their presentations? | 19:22 |
juliusb | i saw olof's are on dropbox | 19:24 |
juliusb | but you need to download some software for that?! | 19:24 |
stekern | yeah, I was there looking too, but didn't feel like installing it on my work computer | 19:43 |
stekern | can't you upload files to the wiki? | 19:44 |
stekern | there's slideshare, but to download the pdf's you need to register | 19:44 |
stekern | you can look at them without registering though | 19:45 |
jeremybennett | juliusb: stekern: You can use dropbox just through its web interface. | 20:06 |
jeremybennett | I actually have their software on my Linux machine, because it's convenient. Just move a file into the dropbox directory and it gets mirrored. | 20:06 |
juliusb | How is it done via the web interface? | 20:11 |
juliusb | I see | 20:13 |
juliusb | the "Dropbox" link in teh top left hand corner after you register and sign in | 20:13 |
stekern | jeremybennett: ah, ok, I didn't see the register button up there when I looked at it at work | 20:34 |
stekern | just the huge "Download dropbox" button ;) | 20:34 |
juliusb | yeah scared me off | 20:36 |
juliusb | so I actually updated my slides with a couple of pages on some topics I discussed during the presentation but didn't have on the slides | 20:43 |
juliusb | licensing, noted in the discussion page that the dhrystone and coremark on the board favoured mor1kx | 20:44 |
juliusb | and added URLs to the repos | 20:44 |
simoncook | I've added all the videos I've uploaded (all but the last two) to the wiki page along side audio versions of all videos. Last two videos will upload overnight and I'll update the two missing links in the morning | 20:44 |
stekern | heh, I noticed, got a conflict when I tried to add the slides ;) | 20:45 |
olofk | jeremybennett: Can you dig up some info about that open source core managment thing from broadcom and whoever else it was. I can't find a word about this on the internet | 23:40 |
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!