**** BEGIN LOGGING AT Fri Sep 30 02:59:57 2005 Sep 30 03:05:03 * lennert wonders about implementing an arm core Sep 30 03:07:01 * dyoung wonders about ARM Corp disliking that. Sep 30 03:07:36 we need a "leg" up Sep 30 03:08:49 * dyoung wonders about a FFT/DFT RT Processor. Sep 30 03:09:25 when I get back...got it "covered" Sep 30 03:09:34 i wonder what it is that the previous one was infringing on Sep 30 03:09:43 we can always do a ppc or sparc or something if they keep being jerks Sep 30 03:12:17 Did you see that gaisler sparc core ? Sep 30 03:12:41 ESA is using it. Sep 30 03:14:03 yeah Sep 30 03:14:04 the Leon Sep 30 03:14:12 the Leon eval boards cost megabucks Sep 30 03:14:18 kilobucks, rather Sep 30 03:14:28 Hmm Sep 30 03:14:33 yeah, but the fun is to write a core yourself Sep 30 03:14:44 Yeah. Sep 30 03:15:06 I need a good excuse to buy that V4 dev board Sep 30 03:15:22 me too :-/ Sep 30 03:15:39 my other half is ok with it but it's like $80 shipping to here Sep 30 03:15:52 The M L401 ? Sep 30 03:15:57 $80 ?! Sep 30 03:16:44 no, the other one Sep 30 03:16:50 the digilent v2pro board Sep 30 03:17:13 http://www.digilentinc.com/info/XUPV2P.cfm Sep 30 03:17:15 Oh... Sep 30 03:17:25 I was looking at http://www.xilinx.com/xlnx/xebiz/designResources/ip_product_details.jsp?sGlobalNavPick=null&sSecondaryNavPick=null&category=&iLanguageID=1&key=HW-V4-ML401-USA Sep 30 03:20:54 this board is pretty neat. Sep 30 03:21:10 yeah Sep 30 03:21:12 theres so many ports on it I dont know what I would do. Sep 30 03:21:13 it's $500 though Sep 30 03:21:18 and the digilent one is $300 Sep 30 03:21:39 and the digilent one has a dimm slot and sata so i like it more Sep 30 03:21:43 I dont know what the salient differences between the v4 and v2pro are. Sep 30 03:21:47 what does the ml401 have that the digilent does not? Sep 30 03:22:32 Pretty much nothing, except its a V4 Sep 30 03:22:55 Platform Flash thingo too Sep 30 03:23:13 I th ink I mean SystemACE Sep 30 03:23:27 I'm not up with the terminlogy Sep 30 03:25:46 I guess V4 is lower power 90nm and "multi platform" whatever that means. Sep 30 03:25:57 does the v4 have powerpc cores too? Sep 30 03:26:49 Either one or 2 of them. Sep 30 03:27:32 For the FX. DUnno about the LX/SX yet Sep 30 03:27:43 hmm Sep 30 03:27:55 Yep, only the FX has the PPC. Sep 30 03:28:14 for a project here at the university It Would Be Nice to be able to do genetic algorithms in an fpga Sep 30 03:28:36 so the ML401 would not have the PPC hard core. Sep 30 03:28:54 Lennert, there was an actricle about that in the last Xcell magazine. Sep 30 03:29:08 "genetic sequencing using a FPGA Coproessor" Sep 30 03:29:12 fluffy just points me to this Sep 30 03:29:13 http://www.enterpoint.co.uk/moelbryn/broaddown2.html Sep 30 03:29:20 dyoung: genetic sequencing is totally different, no? Sep 30 03:29:27 i know that people used the ixp2000 for genetic sequencing Sep 30 03:31:24 Fluffy ? Sep 30 03:31:38 oh, sorry Sep 30 03:31:40 Ben Sep 30 03:31:43 Ben Dooks is his name Sep 30 03:31:45 another arm kernel hacker Sep 30 03:31:58 Okay, I had to Re-read what you just said. Sep 30 03:32:09 Genetic Algorithm. Sep 30 03:32:46 my friend (which was over at my place last night) does genetic algorithms Sep 30 03:32:59 Anyhow, yes the digilent $300 board seems like a better value. Sep 30 03:33:04 and they want to buy a cluster for evaluating them quickly Sep 30 03:33:17 but i figure that it can be done much faster with a bunch of fpgas Sep 30 03:33:39 Yes, in the same issue of Xcell there was something similar. Sep 30 03:33:49 (assuming I understand what a genetic algorithm is) Sep 30 03:34:24 well, something that has a pool of candidate solutions, a fitness evaluation function for a candidate solution, and a way of creating new candidate solutions from existing ones Sep 30 03:34:28 taking things like chromosomes and determining all the possible outcomes; using a bigass tree. Sep 30 03:34:31 using mutation, crossover, whatever Sep 30 03:34:53 Okay close enough. Sep 30 03:35:02 Let me grab that magazine Sep 30 03:35:14 dyoung: you talked about genetic sequencing, isn't that something completely different? Sep 30 03:35:47 Yes. Thats another animal Sep 30 03:37:00 genetic algorithms is like: i know the inputs, i know how to evaluate who is better, let evolution figure out how to get me where i want Sep 30 03:37:07 yup Sep 30 03:38:41 http://www.enterpoint.co.uk/moelbryn/minican.html Sep 30 03:38:47 E145 for a pci board! Sep 30 03:38:59 http://www.nallatech.com/?node_id=1.2.2&id=2 Sep 30 03:41:11 then you slap some of these on Sep 30 03:41:12 http://www.nallatech.com/?node_id=1.2.2&id=11 Sep 30 03:41:59 dyoung: aren't those way expensive? :) Sep 30 03:42:14 probably Sep 30 03:42:33 i'll just fill out the form, see what they say Sep 30 03:43:59 i'd really like the uni to start using fpgas for this stuff Sep 30 03:44:02 okay I get it Sep 30 03:44:26 So you feed your GA some basic information and a set of parameters. Sep 30 03:44:56 Let it sit there and run through all possibilities then you have the best answer for the specificied specification Sep 30 03:45:27 sort of like what chess game would do Sep 30 03:45:50 s/all/a subset of all possible/ Sep 30 03:48:20 Cool. Sep 30 03:48:54 a FPGA should be good at that. Sep 30 03:49:15 evaluating lots of stuff in parallel, yes Sep 30 03:49:29 you can just make a very simple core (say, an arm core) and put 32 of them in one fpga Sep 30 03:52:33 you can even make it do async memory references :P Sep 30 03:55:15 it has to be something that has gcc support Sep 30 03:55:29 otherwise people won't grok it Sep 30 03:55:49 is it possible to get arm licenses for such "small" projects? Sep 30 03:55:49 and it shouldn't be x86 because i won't grok the resulting vhdl anymore after a week :P Sep 30 03:56:10 vmaster: a friend tells me that arm are just really being twats and can't really stop you Sep 30 03:56:21 vmaster: you're not infringing on anything unless you implement thumb, apparently Sep 30 03:56:44 ah, ok, so you use an arm isa, but not an actual core? Sep 30 03:57:49 well, that would be the plan Sep 30 03:58:18 arm happens to be what i'm familiar with Sep 30 03:58:38 but if arm keep being twats i can do mips or sparc instead Sep 30 04:13:42 reading this digilent product page for the v2p board jogged my memory. Sep 30 04:14:17 at some point I need to write/make a doo dad that takes a video input and is able to sample a arbitrary line of data from the signal and store it. Sep 30 04:14:48 That sounds like something the V2P can do. Sep 30 04:14:53 right? :-) Sep 30 04:18:57 yeah, i think so Sep 30 04:19:04 the v2p board plus the analog input module :) Sep 30 04:19:44 Yep. Sep 30 04:19:52 so you'll have your toys paid for you Sep 30 04:19:53 Okay, *now* I have a an excuse to buy one. Sep 30 04:19:55 damn ;) Sep 30 04:20:12 dual powerpc cores, you can run linux on it too! Sep 30 04:20:19 Yep! Sep 30 04:20:34 Theres a appnote for Microblaze/uClinux too Sep 30 04:20:35 *cough* bastard *cough* :) Sep 30 04:20:43 Yep. Sep 30 04:20:45 :-) Sep 30 04:20:49 microblaze is evil proprietary $$$-ware, isn't it? Sep 30 04:20:56 Probably. Sep 30 04:21:07 we need to write the OpenJTAGCpu Sep 30 04:21:15 LennertBlaze is going to be way cooler. Sep 30 04:21:18 hah Sep 30 04:21:29 MicroDerek doesn't have that ring to it, i agree Sep 30 04:21:30 * lennert hides Sep 30 04:21:35 Hey... Sep 30 04:22:24 i'd write an ARM core and call it the PhukTart Sep 30 04:22:34 and then wait until arm send their nastygrams my way Sep 30 04:23:10 (and then file a complaint about them sending me obscene letters) Sep 30 04:26:15 hehe Sep 30 06:17:52 g'nite Sep 30 06:18:29 nitey. Sep 30 06:18:45 Nice Email btw. :-) Sep 30 06:18:56 hehe, why so? Sep 30 06:19:26 "Your web form is brainless, but anyways..." :-) Sep 30 06:19:33 hehe, yes :) Sep 30 06:21:25 i mean, you give a list of 20 fields, * marks the fields that are mandatory, and all fields have a * ! Sep 30 06:21:52 Yep. Sep 30 06:22:12 Probably written by some part-timer or consluttant. Sep 30 06:22:19 consluttant :) Sep 30 06:23:16 hehe one of my friends has been trying to talk me into making a move to san francisco to do consluttant work. Sep 30 06:23:40 she keeps saying, "you dont need to be a total slut, just someone's bitch for a while. its not that bad!" Sep 30 06:24:08 haha Sep 30 06:32:45 dyoung-zzzz: and she enjoys being someone's bitch for a while? Sep 30 06:33:10 I think she has a normal job. Sep 30 06:33:17 bitch. :-) Sep 30 06:34:09 does her husband agree? Sep 30 06:34:47 not married. Sep 30 06:35:11 but hmm. Sep 30 06:35:26 still sounds slightly wrong :) Sep 30 06:36:18 I offered to be her bitch, but that wasnt appealing enough. Sep 30 06:36:24 hehe Sep 30 06:36:30 other way round of course! Sep 30 08:21:48 dyoung-zzzz: i received a response from nallatech, they want to know my location Sep 30 08:29:59 no word about prices? Sep 30 08:30:52 ah, in case you didn't see the usenet posting - the enterpoint devices will be on discount Sep 30 08:32:20 vmaster: oh.. Sep 30 08:32:26 vmaster: when, where, whom? Sep 30 08:32:50 Prevue - FPGA Dev Board Sale Sep 30 08:32:50 From: Sep 30 08:32:50 John Adair Sep 30 08:32:50 Date: Sep 30 08:32:50 Friday 30 September 2005 16:46:29 Sep 30 08:32:52 Groups: Sep 30 08:32:55 comp.arch.fpga Sep 30 08:32:57 no references Sep 30 08:33:44 GBP£160 (BD2-400) and GBP£190 (BD2-1000) Sep 30 08:33:52 due to a "stock clearance" Sep 30 08:34:33 i found the post Sep 30 08:35:36 so that'd make the bd2-1000 board E272,- Sep 30 08:35:50 and the -400 E229,- Sep 30 08:36:37 about 30% off Sep 30 08:36:49 yeah Sep 30 08:37:16 <[g2]> 30% off ??? Sep 30 08:37:24 <[g2]> morning guys Sep 30 08:37:29 yeah Sep 30 08:37:47 the enterpoint pci boards are on sale, apparently Sep 30 08:37:59 <[g2]> url ? Sep 30 08:38:08 <[g2]> new rev coming ? Sep 30 08:38:35 he also writes: Sep 30 08:38:36 We also have a product coming that is aimed at students and is going to be Sep 30 08:38:36 aimed at the cheap end of the market. It will be PCI based and should be Sep 30 08:38:37 capable of supporting some of Broaddown2's DIL header add-ons. Sep 30 08:38:41 so that is extra good news Sep 30 08:50:18 <[g2]> I wonder if enterpoint needs a US distributor :) Sep 30 08:57:33 <[g2]> lennert how's the vhdl coding going ? Sep 30 08:58:25 <[g2]> A USB 1.1/2.0 analyzer would be way cool too Sep 30 09:00:04 would a fpga be fast enough for that task? Sep 30 09:30:54 <[g2]> 1.1 way easy Sep 30 09:31:10 <[g2]> 2.0 a reach Sep 30 09:32:29 yeah, that's what i think Sep 30 09:32:35 11mbps - no problem Sep 30 09:32:39 480? tough Sep 30 09:33:01 <[g2]> How fast is the LVDS stuff Sep 30 09:40:36 Point-to-point LVDS signaling is possible at speeds of up to 622 Mb/s. Sep 30 09:40:41 from a xilinx app note Sep 30 09:44:51 ah, ok, this could be a lot easier Sep 30 09:45:14 <[g2]> what there's something in the app note ? Sep 30 09:45:24 with usb 2.0, you use a transceiver Sep 30 09:45:46 <[g2]> so you just need to handle the 480Mbs bit stream Sep 30 09:45:48 that takes care of the high-speed serial signal, and offers a parallel interface Sep 30 09:46:05 <[g2]> sync'd to a clock comming from the transceiver Sep 30 09:46:22 yep Sep 30 09:46:35 60mhz parallel isn't a problem, i guess Sep 30 09:46:53 <[g2]> 8-bits wide Sep 30 09:46:56 how much are commercial usb 2.0 analyzers? Sep 30 09:46:57 yes Sep 30 09:47:07 <[g2]> at least $1K Sep 30 09:48:17 could indeed be a nice job for an fpga Sep 30 09:48:54 <[g2]> having a DDR or SATA intf would be quite cool Sep 30 09:49:27 <[g2]> 60MBs is quite a heavy data rate Sep 30 09:49:48 <[g2]> or 15M 32-bit word Sep 30 09:50:00 ellisys uses 32mb fifo, and an usb 2.0 uplink to the host pc Sep 30 09:50:55 <[g2]> right that gives you 1 second for up and down Sep 30 09:51:14 <[g2]> or 2 seconds Sep 30 09:51:24 <[g2]> isn't usb half-duplex ? Sep 30 09:51:27 i guess you'd filter the data inside the fpga, and transmit only what is interesting Sep 30 09:52:02 <[g2]> well there's really 2 classes of problems Sep 30 09:52:23 <[g2]> 1 protocol issues where filtering would work and just grabbing all the data on the bus Sep 30 09:52:38 <[g2]> 2) usb PHY issues Sep 30 09:52:48 ah, ok Sep 30 09:52:49 <[g2]> which the transceiver would mask Sep 30 09:53:34 <[g2]> a firewire 800 chip for the data store would be cool :) Sep 30 09:57:21 hehehe, yeah Sep 30 13:23:59 [g2]: vhdl coding is going fine Sep 30 13:24:09 [g2]: my current board is just slightly limited Sep 30 13:41:50 <[g2]> the S3 ? Sep 30 13:42:22 <[g2]> you've got a 200 or 400 ? Sep 30 13:43:27 xc3s200 iir Sep 30 13:43:28 c Sep 30 13:43:47 i'm not running out of logic but out of ports to use :) Sep 30 13:44:01 vga is working, serial is more-or-less working, ps/2 will be working soon Sep 30 13:44:09 that's really all the board has :) Sep 30 13:44:50 You have those 40-pin I/O connectors. Sep 30 13:44:56 yeah Sep 30 13:44:59 but what to hook up to them? Sep 30 13:45:05 a pci edge connector maybe Sep 30 13:45:19 USB Adaptor? Sep 30 13:45:28 hmm Sep 30 13:45:35 hmmm Sep 30 13:45:35 <[g2]> a 14-pin and 20-pin ARM JTAG connector Sep 30 13:45:36 more serial ports Sep 30 13:45:59 [g2]: i know where you're going ;) Sep 30 13:46:03 a IDE Hard Drive? Sep 30 13:46:27 i'd really like pci and/or a gigabit ethernet connection Sep 30 13:46:30 Its a minor pain that the S3 inputs are not TTL compatible. Sep 30 13:46:45 dyoung-zzzz: not? Sep 30 13:47:30 <[g2]> what's the voltage level on the I/Os can we connect directly up to the JTAG on the IXP4xx ? Sep 30 13:47:35 Nope. Sep 30 13:47:38 They are LVTTL only Sep 30 13:48:04 [g2] LVTTL Sep 30 13:48:10 <[g2]> dyoung-zzzz that should work for us right ? Sep 30 13:48:23 Depends on what youre trying to do. Sep 30 13:48:31 If its talk to a IXP4xx, yes. Sep 30 13:48:32 <[g2]> JTAG flashing and debugging Sep 30 13:48:39 <[g2]> nod on the IXP4xx Sep 30 13:49:15 Thats what the CPLD would be for in past desgins we've talked about. To increase the i/o range. Sep 30 13:49:22 dyoung-zzzz: can't you just add a resistor and be done with it? Sep 30 13:50:02 <[g2]> dyoung-zzzz right. However currently I'm only working with Xscales Sep 30 13:50:10 <[g2]> IXP4xx's even for now Sep 30 13:50:33 <[g2]> I've got couple Cirrus boards around though Sep 30 13:50:35 lennert, that would be pretty UBE; but I suppose you could. Sep 30 13:51:42 dyoung-zzzz: hmmm Sep 30 13:52:01 Hmm Sep 30 13:52:03 Actually... Sep 30 13:52:13 the S3 has SelectIO Sep 30 13:52:15 indeed, no TTL Sep 30 13:52:26 so you can do something weird. Sep 30 13:52:36 GTL, HSTL, LVCMOS, LVTTL, PCI, SSTL Sep 30 13:52:49 theres an app note about using SelectIO with TTL using OC Sep 30 13:52:50 (and some differential standards) Sep 30 13:54:33 Aha, and the S3 has internal clamp diods. so you can just limit the current to < 10ma Sep 30 13:54:53 300ohm should be enough. Sep 30 13:55:49 dyoung-zzzz: ? Sep 30 13:58:58 http://direct.xilinx.com/bvdocs/publications/ds099-2.pdf Sep 30 13:59:09 page 2 Sep 30 14:00:19 yeah i just don't know what the heck you're talking about Sep 30 14:01:37 When the input voltage is greater than VCCO +0.5V, the upper clamp diode turns on and conducts a reverse current into the associated supply (VCCO for non-dedicated I/O; VCCAUX for dedicated I/O). Because this reverse current can damage the I/O, the Spartan-3/-3E I/O is not 5V-tolerant. Sep 30 14:01:42 Spartan-3/-3E I/O can be made 5V-tolerant by using an external series current limiting resistor to limit the current into the upper clamp diode to 10 mA. This makes the input 5V-tolerant, but an I/O configured as an output still cannot drive 5 Volts and the resulting VOH does not meet the input specifications of the 5V device Sep 30 14:02:12 so using a resistor in series to limit the current to < 10ma total will work as you suggested. Sep 30 14:15:53 i was thinking about a voltage divider actually Sep 30 14:16:22 then again, i still don't know what i'm talking about Sep 30 14:40:01 s'pose you could do that too; should be fine for "slow" stuff. Sep 30 14:40:15 Its still UBE. :-) Sep 30 14:40:26 i don't know enough about electronics, really Sep 30 14:41:44 <[g2]> that makes two of us :) Sep 30 14:42:02 dyoung-zzzz: where did you learn your electron foo? Sep 30 14:50:53 Um... Sep 30 14:51:06 Shocking the hell outta myself mostly... Sep 30 14:51:43 sounds good Sep 30 14:51:50 <[g2]> heh... and living to tell about it :) Sep 30 14:51:55 Actually when i was a child, my father and I together made a crystal radio. Sep 30 14:52:03 [g2]: that's the 'natural selection' part Sep 30 14:52:03 I thought it was the coolest thing since sliced bread Sep 30 14:52:15 after that; justu a lot of playing around. Sep 30 14:52:18 [g2]: either that or 'intelligent design', which'd mean that the intelligent designer likes dyoung Sep 30 14:52:23 (and zapping the eff outta myself) **** ENDING LOGGING AT Sat Oct 01 02:59:56 2005