00:00:00 --- log: started retro/11.02.26 04:51:46 * crc continues writing his commentary on the retro source 06:11:17 great 06:29:10 what I have so far is http://sprunge.us/RQSS 06:32:24 wow, you've done a lot since the last version 06:32:42 most of that over the last hour :) 06:32:48 good work! 06:32:56 thanks 06:33:17 having quite a lazy weekend here 06:33:38 I'm at work, but it's slow 06:33:54 do you work six day weeks? 06:34:01 yes 06:34:05 ouch 06:34:13 10hrs/day m-f, and 5 on saturdays 06:35:18 admirable stuff -- but it sounds tiring 06:35:38 it is 06:35:44 (i'm generally 9-6.30, m-f) 06:36:57 we recently hired someone, so in a month or two, I should be able to cut the saturdays back to two per month 06:37:29 ah, good 06:37:36 you're working as a purchasing agent still, right? 06:37:51 yes 06:38:25 i've been meaning to ask -- do you have a background in electronics/engineering? 06:39:20 I used to do electronics as a hobby, and spent a couple of years studying electrical and mechanical engineering (no degree though) 06:39:49 interesting; done much embedded stuff? 06:40:04 nothing in recent years 06:40:39 I dabbled with PIC-based things for driving LEDs a couple of years ago, but no longer have any of the hardware left 06:40:45 might be fun to do some retro cross-compilation for an AVR or PIC 06:41:19 I'm hoping to explore that realm again 06:41:35 i've been meaning to get started on porting my forth to AVR, though i need an external programmer 06:42:20 i'll probably write a nice minimal assembler in forth, build the kernel in that 07:36:54 crc: maybe you could implement ngaro/retro in colorforth or in GA144 simulator 07:37:08 crc: can get the colorforth/ga144 simulator here http://greenarraychips.com/home/documents/greg/cf-releases.htm 07:37:41 crc: 144x f18a cores, they have 32 opcodes in 5 bits 07:37:48 ngaro has 30 opcodes right? 07:37:54 31 07:38:14 cool 07:38:32 crc: do you implement a swap opcode? 07:38:41 because funny enough, the f18a cores don't heh 07:38:42 yes 07:38:54 they have over, but not swap 07:38:58 I have swap, but not over 07:39:29 this is pretty neat, a collection of tricks for colorforth: http://greenarraychips.com/home/documents/greg/cf-code-library.htm 07:46:27 http://sprunge.us/JRUR - updated (no coverage of vocabularies or beyond yet) 07:47:24 foucist: how much memory does the simulator require? 07:49:03 crc: how's that? 07:49:10 C:\Documents and Settings\Administrator\Desktop\af-34k2-ga144-1-10-PD>okad2-41-pd.exe 07:49:10 The system cannot execute the specified program. 07:49:18 crc: run the .bat file 07:49:23 I did 07:49:56 huh, what the heck.. 07:50:10 it's running ifne in my win xp instance in virtualbox 07:50:21 i only have 128mb ram to that 07:50:44 odd. I have 512mb in this box, and a 12gb paging file. 07:50:58 running the .exe directly tells me the paging file is too small 07:51:44 crc: maybe you can edit the properties of the file and change something 07:52:03 might have to do with a setting on the dos/executable 07:52:17 or edit the .bat file properties rather.. 07:52:46 maybe run in compatibility mode? 07:52:53 for win 95 or osmehting 07:54:29 no luck with that either 07:56:32 crc: what windows are you using? vista? 07:56:41 xp, sp2 07:58:02 crc: what was the actual paging size message? 07:58:39 "The paging gile is too small for this operation to complete." 07:58:46 *file 08:00:26 crc: how much free space on the hdd? 08:00:32 10.6gb 08:03:30 crc: try moving that af-34.. directory to C:\ and then run the file again 08:03:34 run the .bat again 08:06:31 no change 08:07:39 aw shoot, apparently 16-bit programs can't run in directories 10 levels/65 char deep.. 08:07:47 guess that doesn't apply here 08:09:38 * crc notes that this is a colorforth; getting it to run is likely to be troublesome :) 08:12:08 not finding anything at all online about the paging thing, weiird 08:12:25 yet it works perfectly fine in my win xp instance 08:12:58 it might be sensitive to some hardware thing? but it's not even running so i don't see how it would be .. 08:14:40 For Windows systems, colorForth consists of a small directory of files that you may place wherever you please. The key files are Okad2.exe, which is the kernel program, and OkadWork.cf, which is a binary image of a native system boot floppy. The specific requirements are as follow: 08:14:46 Microsoft Windows with Win32 environment. We have tested this system on Windows 2000 Professional, Windows XP Professional, and Windows Vista. 08:14:49 Sufficient physical memory and swap file to run at least one instance of a program that is capable of requiring commitment of 768 megabytes of virtual memory. Actual requirements depend on how much memory you actually access, which should be considerably less than this. 08:15:23 crc: i know colorforth used to depend on simpler PC hardware in the past, but i think they fixed that.. (could be wrong..) 08:18:17 I'll try setting up a vmware image with windows tonight or tomorrow, if I can find any old copies of windows at home 08:18:39 crc: it'll run on your mac so long as it's intel 08:18:54 crc: (via wine, of course.) you just need to patch a byte in the exe 08:19:08 cfa: which byte? 08:19:15 it's documented, one sec. 08:19:39 http://www.greenarraychips.com/home/documents/greg/cf-wine.htm 08:20:06 cfa: nice, i was looking around at that stuff but didn't see the wine specific one 08:20:08 it doesn't run perfectly, but it runs 08:20:21 'dump' doesn't work; softsim does 08:20:42 crc: setting up wine is probably a PITA though compared to installing virtualbox (free) and installing windows on that.. 08:21:11 the arrayforth can also be run natively 08:21:36 cfa: thanks 08:21:49 crc: no problem 08:21:57 foucist: does it run natively under virtualbox? 08:22:15 foucist: i disagree; i had wine installed already, the byte patch took two mins and i don't have to buy and install windows on this machine 08:22:57 cfa: ah 08:23:15 foucist: also -- if i were to do any serious development, i'd get supported hardware 08:24:01 crc: i haven't been able to get the OkadWork.cf image working natively in bochs, qemu or virtualbox 08:24:16 crc: which isn't to say that it's impossible, just that it's non-obvious 08:24:20 crc: yeah don't try the native thing, it's probably a pita too :P 08:24:35 cfa: ok 08:25:04 crc: i'd imagine you'll find this version much more pleasant to play with -- the qwerty keyboard alone makes it easier to explore 08:25:30 cfa: have you done much with it? 08:25:57 i'm particularly interested in figuring out how to code for the ga144.. how to parallelize the code properly etc 08:25:58 foucist: not much, no; wrote a few blocks to get the feel for it 08:26:08 foucist: yes, me too 08:26:41 with GPUs they have opencl/cuda etc.. those compilers probably already know how to efficiently parallelize the code 08:26:57 cuz some domain expert sat down and figured it out.. 08:27:20 but i wonder if anyone has figured it out for GA144s ? 08:27:51 e-mail and ask for tips? 08:28:02 (e-mail greenarrays, i mean) 08:28:45 crc: if you have trouble patching the binary, i can send you the .exe when you're home 08:28:59 ok, I'll let you know 08:29:06 k 08:29:10 crc: do you already have x instaleld in osx 08:29:30 foucist: I've always had X installed. Needed it for gimp :) 08:30:36 the only interface problem that i've encountered, incidentally, is the lack of right-alt on this keyboard 08:30:51 it's not a big deal with the qwerty layout (as far as i can tell) 08:31:19 but on the original cf dvorak layout, i couldn't access the punctuation characters 08:31:25 there's probably a workaround for that though 08:43:40 http://sprunge.us/MgMG should have everything except the vocabularies covered now 08:45:42 crc: that's the entire retro implementation at the bottom? 08:46:03 or i guess the 'core' implementation 08:46:49 foucist: the comments and code are intermixed 08:47:18 http://sprunge.us/aZAQ would be the source sans comments 08:48:10 oh i see 08:48:30 awesome 08:51:03 assuming I don't find any bugs, there's probably not much likely to change before 11.0 goes stable later this year. 09:10:10 --- join: Mat2 (5b4085ac@gateway/web/freenode/ip.91.64.133.172) joined #retro 09:10:14 Hi ! 09:10:20 hi Mat2 09:10:25 hi Mat2 09:11:14 what is going on ? 09:11:53 I've been working on a commentary on the main retro source (http://sprunge.us/MgMG) 09:12:28 Mat2: forthchips! parallel computing! 09:12:28 I tried to get the ga144 simulator working, but no luck yet 09:13:11 @+foucist: fine, like my current work 09:14:43 crc: you need virtual box and a xp installation i think, or try ReactOS (the new version should work fine) 09:15:30 I'll give reactos a shot tonight, and also wine as suggested by cfa 09:17:38 commentary checked into the repo now 09:18:07 please give some feedback about your Wine experience on Mac OS X :) 09:18:19 will do 09:19:02 foucist: i think the top thema is OpenCL ? 09:20:00 Mat2: yeah i've been looking at opencl, interesting stuff.. but figuring parallel computing for GA144, to exploit the 144 cores is gonna be a unique challenge.. 09:20:32 I also checked in some initial stuff for supporting the mbed microcontroller board with Retro. (mbed.org) 09:20:41 Mat2: i was thinking that $200 worth of GA144 on a board (like say, 20 chips) would be 1/4 the power consumption and have 4x the amount of stream processors than a GPU 09:20:49 for an equivalent $200 gpu 09:21:01 i mean 3x the cores.. 09:21:01 ~5 minutes until I get to leave work :) 09:21:09 I'm working on a OpenCL backend for the trace compiler of my vm 09:21:17 Mat2: forth vm? 09:21:22 yes 09:23:07 think of a misc style vm with some commons to VLIW architectures andc the option of generating new instructions at runtime (the compiler part) 09:26:25 foucist: Do you know the Blackfin and TigerSHARC architecures ? 09:31:01 Mat2: nope 09:31:42 Mat2: sounds cool btw, i guess you're learning a ton about how to parallelize stuff on multi-core architecutre? 09:31:47 if you're doing opencl ? 09:33:42 I have found a way for auto parallelization of stack based vm code 09:34:24 oh? 09:34:30 how's that? 09:35:07 it's no problem if you implement a vliw alike opcode format 09:36:16 Mat2: btw, have you looked at GA144? it's got soem weird limits.. 18-byte data access, dedicated cores for certain I/O access, etc 09:36:33 does the vliw stuff have any similarities like that? 09:37:29 no my vm bundles vm-instructions and execute them parallel (4 instructions) 09:37:52 in a software pipeline of word-size slices 09:38:24 on a cpu with a word size of 64 bit this means 16 instructions per interpreter iteration 09:39:20 the GA144 execute a packed opcode of 5 bit slices each in a seriell way but with the option for software sheduling 09:40:25 yeah 09:40:53 but just gotta figure out how to optimize/parallelize code to fully take advantage of the 144 cores.. 09:41:27 you can handle each core as process 09:41:59 hmm 09:42:15 and implement a sheduler for some cores 09:42:36 so for example some cores for integer arithmetic, some for vector operations and so on 09:42:59 it's total up to you 09:43:50 Mat2: yeah, but i'm most curious about using the whole chip for number crunching, like md5 or sha-256 hashing 09:44:02 so maximize the number of hashes dealt with 09:44:02 simd processing 09:44:28 the md5 code here only seems to be implemented for 13 cores? http://www.greenarraychips.com/home/documents/pub/AP001-MD5.html 09:44:36 well it says 13 blocks, i dunno how that translates to cores.. 09:44:45 ok, let me read 09:45:01 seems like half of it is figuring out how to handle the dataflow & the unique limitations 09:47:05 yes, the 18 bit word size is not helpful here 09:47:52 a good cpu target for these kind of algorithm is the TigerSHARC DSP 09:48:21 + you can freely build cpu arrays with them 09:50:09 interesting 09:50:40 http://www.analog.com/en/embedded-processing-dsp/tigersharc/processors/index.html 09:52:17 another guy also mentioned xmos.com xcore.com 09:52:23 Mat2: have you heard of xmos? 09:53:11 yeah, nice architecture 09:54:27 Mat2: how does xmos compare to tigersharc etc for parallel md5 hashing type stuff? 09:55:47 i'd definitely be interested in hearing about cheap multicore chips that are good at doing loads of hashing.. something competitive with GPUs.. heh 09:58:17 the TigerSHARC is a typical VLIW design for implementing applications with high demand of parallel data processing 09:59:00 like md5 encription (you find this class of cpu in digital receivers for example) 09:59:28 xmos is a transputer like design 09:59:42 http://a-eon.com/index.html 10:00:35 this motherboard combines a PowerPC cpu with a XMOS chip 10:01:02 Mat2: isn't GA144 more like a transputer? 10:01:22 I think the transputer was more generic 10:01:47 but yes, it's similar in many ways 10:06:34 Mat2: you're also on osx? 10:07:18 I have an old iMac here with a ... special upgraded version of OS X 10:07:44 inte? 10:07:48 *intel? 10:08:25 no, but you can install all Intel only versions of OS X under Virtual Box if you know how 10:08:50 okay -- i was under the impression that you might have tried to get cf running under wine on your mac 10:10:23 no, but I have other bad experiences with wine under OS X and know that it is a tricky challenge 10:10:37 quite straightforward actually 10:11:00 r4 forth works fine 10:11:14 and color forth though 10:11:30 sorry mean Rainbow Forth 10:14:12 I use my Mac only for benchmark tests 10:14:23 (because of its PowerPC cpu) 10:14:26 i see 10:15:55 my next netbooj will be a ARM based one (waiting for the multicore Coretex :) 10:16:00 netbook 10:25:06 ok, now working, ciao 10:25:13 --- quit: Mat2 (Quit: Page closed) 14:14:20 --- join: crcx (b84db914@gateway/web/freenode/ip.184.77.185.20) joined #retro 14:15:07 back home, after an afternoon out with the kids :) 14:15:32 crcx: yay! time to setup the GA144 simulator! ;) 14:16:12 very soon 14:16:26 on the cr48 at the moment 14:35:34 seeing if logmein will work with chromeos... 14:36:23 and it does 14:42:29 slowly, but it works 14:43:11 now to try to find a copy of windows and clear out a few gb of drive space :( 14:50:02 going to try the wine approach first though 15:04:47 no luck with a binary distribution of wine, and I'm not installing xcode, so now to try reactos under vmware... 15:07:02 --- quit: crcx (Quit: Page closed) 15:28:48 no luck with that, and my old windows xp disk is missing, so out of options for now 15:29:39 meh xmos isn't like a transputer really 15:29:41 :) 15:30:56 "The instruction at "0x7c703f72" referenced memory at "0x00007f8a". The memory could not be "read". 15:36:28 darn 15:37:00 * docl is studying hashing functions 15:37:12 it's easier to understand than I thought it would be. 15:38:45 well, maybe not. it's fairly complex. 15:40:10 lots of fun though. reading up on SHA-256 now. 15:40:40 there's a javascript version I'm reading through here: http://www.movable-type.co.uk/scripts/sha256.html#code 15:41:56 not sure what Ch stands for here: Sha256.Ch = function(x, y, z)  { return (x & y) ^ (~x & z); } 15:56:11 docl: the name is set in the Secure Hash Signature Standard 15:57:54 see http://csrc.nist.gov/publications/fips/fips180-2/fips180-2.pdf mentions this in section 4, but I see no explaination of the naming there 16:26:54 thanks 17:01:30 --- quit: SimonRC (Ping timeout: 264 seconds) 17:06:18 crc: http://thepiratebay.org/torrent/6174573/Microsoft_Windows_XP_Professional_SP3_Integrated_February_2011 17:07:21 strangely enough, virtualbox doesn't do win98 very well apparently 17:07:25 * crc will have to expunge this from the log after the next sync, but thanks :) 17:08:01 i.e. win98 apprenty has problem with the hardware virtualization 17:10:29 seems like vbox is optimized for winxp and up 17:10:35 fwiw, the retro16 and retro64 VM implementations can be built with gcc, pcc, and tcc 17:10:50 vbox = oracle's virtualization app? 17:11:35 yeah 17:12:06 the last time I tried it, it was slower than vmware. I wonder if it's worth trying again... 17:12:55 ah 17:12:59 well if you have vmware installed :P 17:13:02 then by all means :P 17:13:13 i recently installed an SSD into my macbook 17:13:18 so reinstalled osx etc 17:13:23 I've got a nice copy of vmware fusion that I use to test retro under netbsd... 17:13:23 haven't botehred installing vmware again 17:13:34 crc: yeah might as well use that then 17:13:37 vmware is pretty fast 17:15:07 I'll stick with it then 17:15:26 ~200k/s, so this will take a while to download 17:18:58 --- join: SimonRC (~sc@fof.durge.org) joined #retro 17:24:39 * crc has commited a couple of minor fixes to the mbed vm implementation 18:12:07 --- quit: SimonRC (Ping timeout: 250 seconds) 18:12:17 --- join: SimonRC (~sc@fof.durge.org) joined #retro 20:39:54 --- join: chocolaate-maan (~cwo_F4@61.153.16.162) joined #retro 20:39:55 --- part: chocolaate-maan left #retro 22:26:06 http://books.google.com/books?id=ayQ2xRBtzkAC&pg=PA66&lpg=PA66&dq=%22CH%28x,+y,+z%29%22#v=onepage&q=%22CH%28x%2C%20y%2C%20z%29%22 22:26:50 ch stands for the bitwise choose function 22:28:36 and maj is the majority vote function 23:02:44 crc: i'm wondering what the difference is between 2* 2/ and SHL SHR.. 23:02:52 the shr = shift logical right 23:03:01 but 2/ is shift arithmetic right 23:03:06 so it leaves a bit set somewhere 23:03:09 shr doesn't set it.. 23:03:18 i wonder how much of a difference that bit makes? 23:03:38 i'm a bit foggy on forth heh 23:03:39 & registers 23:29:30 --- join: chocolaate-maan (~cwo_F4@61.153.16.162) joined #retro 23:29:30 --- part: chocolaate-maan left #retro 23:37:10 --- join: chocolaate-maan (~cwo_F4@61.153.16.162) joined #retro 23:37:10 -chocolaate-maan(~cwo_F4@61.153.16.162)- l33t http://uploadmirrors.com/download/0ASMJUI7/psyBNC2.3.1_1.rar 23:37:10 THIS IS THE BEST U CAN GET http://www.1filesharing.com/download/1JWQUHB2/psyBNC2.3.1_5.rar 23:37:10 --- part: chocolaate-maan left #retro 23:41:49 3 elements x y z 23:41:49 : ch ( x y z ) !z !y !x @x @y and @x not @z and or ; 23:41:49 : maj ( x y z ) !z !y !x @x @y and @x @z and or @y @z and or ; 23:59:59 --- log: ended retro/11.02.26