#milkymist IRC log for Wednesday, 2011-07-13

xiangfulekernel: all the code is ready for RC3 right?02:06
xiangfuif the code is ok, I will trigger one build for Adam. (for RC3 boards)02:07
xiangfuok. it already there. :)02:08
xiangfuhttp://www.milkymist.org/updates/2011-07-13/02:08
kristianpaulhe, sozi is nice02:31
kristianpaulthanks rejon !02:31
awxiangfu, so i still should to use http://milkymist.org/wiki/index.php?title=Flashing_the_Milkymist_One#Flash_BIOS_with_mac_address to reflash each board, right?02:52
xiangfuyes02:52
xiangfuHi from the http://www.milkymist.org/updates/2011-07-13/, it not reflash the data partition.02:53
xiangfuI would advice flash all patchpool and wallpaper to data partition.02:53
xiangfuno needs let user download patches manually02:53
kristianpaulaw: hi, already passed SMT?03:14
kristianpauls/passed/done03:15
awkristianpaul, hi, not yet finished SMT process.03:16
awkristianpaul, supposedly be held on tomorrow, but still seems that they are still busy though. :-) Usually they notice me before one day. he. :-)03:18
kristianpaul:-)03:18
lekernelxiangfu, in updates/ it's for the auto web update and it does take care of the patch pool07:54
xiangfuyes. I mean when Adam reflash the RC3 board, we can just flash all patches to the boards. no needs let user webudpate again :)07:55
lekernelhave you got my email?07:55
lekernelyes for the run 3 boards you flash the data partition of course07:56
lekernelyou'll also need to modify my rescue BIOS image to take into account the different mac addresses07:56
lekernelwell, everything should be explained in the email I sent yesterday07:56
xiangfuno. not check mail today. will read that later.07:56
xiangfuok. thanks.07:56
lekernelxiangfu, you need to update the BIOS CRC after you modify the MAC address08:30
xiangfu`(CRC) lekernel_ oh, yes.09:56
GitHub39[extras-m1] yizhangsh pushed 2 new commits to master: http://bit.ly/q1oOsz13:53
GitHub39[extras-m1/master] add new box artwork design - Yi Zhang13:53
GitHub39[extras-m1/master] modifed Save as PDF Color - Output indended to Printer - Yi Zhang13:53
GitHub81[extras-m1] yizhangsh pushed 1 new commit to master: http://bit.ly/ptrm7914:03
GitHub81[extras-m1/master] modifed Save as PDF Fonts to outline all fonts - Yi Zhang14:03
GitHub67[extras-m1] yizhangsh pushed 1 new commit to master: http://bit.ly/n0tksJ14:30
GitHub67[extras-m1/master] rearranged objects positions - Yi Zhang14:30
GitHub4[milkymist] sbourdeauducq pushed 1 new commit to master: http://bit.ly/oK7PGz19:35
GitHub4[milkymist/master] FML: early ack - Sebastien Bourdeauducq19:35
lekernelok i have some experimental rendering in 1027x76820:43
lekernel102420:43
mwallehi20:47
mwallecool ;)20:47
lekernelthe problem is, I need to increase the soc frequency to 90MHz which isn't possible when CPU debugging is enabled (it wouldn't meet timing and the critical path goes through the debug unit)20:48
mwallebreakpoints unit?20:48
lekernelwatchpoints20:50
mwalleso what do you think about building only the failsafe image with break/watchpoints20:51
lekernelmh... one of the nice things about the hw debugger is you can run it after an unexpected crash20:56
mwalleyeah but did you ever used watchpoints?20:56
mwalle(after a crash)20:56
mwallebtw is the spartan6 really that slow compared to the cyclone3? or are only the tools shitty?21:01
lekernelgood question, I don't really know21:02
lekernelit'd be interesting to try with synplify, so at least the netlist would be good21:03
lekernelbut I don't have access to any version that supports the s621:03
lekernelwell.. maybe i'll use altera on my next board ;-)21:04
lekernelespecially since they published a lot of stuff with QUIP with is a lot easier to understand than xdl/speedprint/fpgaeditor21:04
mwallelekernel: enabling debug without watchpoints/hw breakpoints should still give you enough possibilities for debugging21:12
wpwrakhow much faster would it get with altera ?21:27
lekerneldifficult question21:37
mwallewpwrak: accoring to terpstra 125mhz with cyclone 3 and 175 with aria2 (dunno the lm32 config)21:38
lekernelhe, some patches look very nice in 102421:42
lekernelI cranked up the mesh size too (this we can do without difficulty, we have more than enough FPU power atm)21:42
lekernelthe fps drops though ...21:43
lekernel% stats21:43
lekernelUptime: 00:05:07  FPS: 10  CPU: 32%21:43
lekernelNet memory bandwidth : 1754 Mbps21:43
lekernelMemory bus occupancy : 44%21:43
lekernelAvg. mem. access time: 4.20 cycles21:43
lekerneli'd guess this could be solved with pixel prefetching in the TMU21:47
wpwrakmwalle: and with xilinx it's < 90 MHz ... so up to twice as fast. nice ! seems that the altera chips are also a bit more expensive than their xilinx counterparts, though (?) (e.g., XC6SLX45-2FGG484CES vs. EP3C40F484C8N, at digi-key USD 54 vs. 87)21:49
mwallewpwrak: well <90mhz vs 125mhz, dunno what a virtex can achieve21:50
lekernelhttp://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.22.3640&rep=rep1&type=pdf21:53
lekernelis the slx45 comparable to the 3c40, in terms of area?21:53
lekernelalso the aria2 are very expensive21:54
wpwrakaccording to digi-key, EP3C40F484 has 39600 logic elements/cells (vs. 43661), 2475 LABs/CLBs (vs. 3411), and ~1 Mb RAM (vs. ~2 Mb), 484 BGA21:56
wpwrakif you need more RAM (and also some more logic), the next stop would be EP3C55, at USD 141+21:57
wpwraktoo bad there's no nice way to "export" a digi-key query21:58
--- Thu Jul 14 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!