**** BEGIN LOGGING AT Sat Oct 15 02:59:56 2011 Oct 15 07:24:20 hey Oct 15 07:24:28 how to convert doc to fb2? Oct 15 07:24:35 or pdf to fb2... Oct 15 09:28:40 does anyone know how to calculate glamo's clocks from its pll settings? Oct 15 14:06:42 Weiss: any idea why the glamo pll1 and pll2 are badly aligned? i can't help but think that matching 48Mhz to 90Mhz is asking for trouble. Oct 15 17:43:28 well, the basic way a pll works allows for virtually arbitrary divisors Oct 15 17:45:26 divide the 48MHz pll osc clock by 48, the 90MHZ by 90, then feed both 1MHz signals to the pll comparator that assures to adjust one of both voltage-controlled oscillators to be exactly 48/90 of the other Oct 15 17:45:46 if that's what you were concerned about Oct 15 17:46:22 obviously you can go for other ratios as well, like 481/900 Oct 15 17:47:41 afaik the glamo bus is only loosely related to RAM bus, I think it is more like "async" and just suspends the CPU until the transaction got ACK Oct 15 17:49:58 (which btw is one of the most annoying details of glamo: it just suspends CPU basically, so no benefit whatsoever by delegating tasks to glamo, CPU can't do anything reasonable concurrently anyway as it can't access RAM. That's what I seem to recall about this whole glamo nightmare) Oct 15 17:50:52 quitte: Weiss: ^^^ Oct 15 17:53:56 yes. i didn't quite get which clocks are generated from which pll,yet. and the glamo making the cpu stop is something I#m concerned about. Oct 15 17:54:41 but i don't know how to quantify changes. e.g. i set the pll2 to 100mhz and i have no idea wether effects are imaginary or not. Oct 15 18:42:30 without looking into any details, I'd say changing a glamo clock from 90 -> 100 will just cause glamo to run unstable but not result in any noticeable increase of any speed of rendering/whatever Oct 15 18:47:22 also watch out for uSD access which is handled by glamo Oct 15 18:48:10 yeah i know ;) i don't use the usd and have no important data on it. Oct 15 18:50:03 after what i've seen so far I think that all the clock setup of glamo should be moved from the kernel to the bootloader. but i guess that's too late now. Oct 15 18:58:59 I guess bootloader needs to do it anyway, unless it'S a bootloader that doesn't need any display (aka qi) Oct 15 18:59:43 qi still does part of it. for the mmc access. Oct 15 18:59:53 aaah, true :-) Oct 15 19:00:11 anyway kernel will redo it unconditionally Oct 15 19:00:17 but i guess if cpu frequency scaling were made to work ... Oct 15 19:01:05 DocScrutinizer: the thing is that I think that the kernel can't redo it if the memory clock is set to 133 without preparing glamo first Oct 15 19:02:16 why? Oct 15 19:05:08 too short a delay from address setup to write enable or some other timing issue Oct 15 19:06:54 hmm, would need close inspection/evaluation. Quite possibly you're right, then access to glamo would have to take care about making sure that memory clock is set to 100 before doing actual chat to glamo Oct 15 19:08:28 maybe the bootloader could pass some kind of glamo initialized flag? Oct 15 19:17:34 would signing an nda enable me to discuss the datasheets here - if that even is still possible? Oct 15 19:25:39 you may discuss the datasheets any time. Signing NDA is basically possible but might get a bit cumbersome as you'd need to negotiate that with Sean MossPultz. I heard he's about to let "OM lawyers check the details" or sth. I'd suggest to /join #openmoko-cdevel and ask there, there are others interested in glamo datasheet as well, and I promise *I* at least wouldn't mind if somebody of those who have access to the datasheet will grant Oct 15 19:25:40 you access to excerpts from it Oct 15 19:42:47 DocScrutinizer: ok thanks. I'll just assume you have some authority to say so. Oct 15 19:43:06 not really Oct 15 19:43:38 I'm just the ex-OM-EE from western hemisphere Oct 15 19:43:53 emphasis on EX Oct 15 19:44:10 good enough. seeing as smedia doesn't even exist anymore. Oct 15 19:44:21 that'S the point ;-D Oct 15 19:44:40 http://admin-trac.openmoko.org/trac/ticket/1671 Oct 15 19:49:36 are the memory controller registers in the register specs? Oct 15 20:10:41 yes Oct 15 20:31:18 quitte: still here? Oct 15 20:31:27 hey Weiss :-) Oct 15 20:33:06 Weiss: which of the files as listed in that trac-ticket above are actually relevant? Oct 15 20:34:16 only the full datasheet has the clock stuff: Glamo 3362 Datasheet V1.0-Full-FIC.pdf Oct 15 20:46:20 Weiss: thanks **** ENDING LOGGING AT Sun Oct 16 02:59:56 2011