IRC log started Wed Sep 22 00:00:01 1999 [msg(TUNES)] permlog 1999.0922 -:- eihrul [lee@usr5-ppp181.lvdi.net] has joined #tunes -:- core [core@mcp.suntech.fr] has joined #tunes people hi * core/#tunes is overwhelmed :) 01:50am -:- SignOff Zhivago: #TUNES (Ping timeout for Zhivago[th.merddin.com.au]) -:- Zhivago [brian@th.merddin.com.au] has joined #tunes core: there? * eihrul/#tunes sighs. 02:20am -:- eihrul [lee@usr5-ppp181.lvdi.net] has left #tunes [] -:- eihrul [lee@usr5-ppp181.lvdi.net] has joined #tunes -:- SignOff eihrul: #TUNES (Leaving) -:- AlonzoTG [Alonzo@216-164-131-242.s242.tnt2.lnhva.md.dialup.rcn.com] has joined #tunes -:- SignOff core: #TUNES (ircII EPIC4pre2.001-NR3 -- This is *not* the default quit message) -:- FareWell [fare@d026.paris-121.cybercable.fr] has joined #Tunes -:- smokie [tw026024@zaalf13.twi.tudelft.nl] has joined #tunes -:- core [core@core.suntech.fr] has joined #tunes huwo hi ho core -:- smokie is now known as smoke` hey smoke! :) how's your OS coming ? great.. about to be able to run non-networked and not-too-complicated unix applications :) that'll be a good base for further expansion :) (and no, i'm not going to use that for native interfaces :) 11:30am -:- NetSplit: tolkien.openprojects.net split from sterling.openprojects.net [11:30am] -:- BitchX+Deb1an: Press Ctrl-F to see who left Ctrl-E to change to [tolkien.openprojects.net] -:- Netjoined: tolkien.openprojects.net sterling.openprojects.net -:- smoke` [tw026024@zaalf13.twi.tudelft.nl] has joined #Tunes -:- hcf [nef@me-portland-us946.javanet.com] has joined #tunes -:- tmf [s720@barkebille.ii.uib.no] has joined #tunes -:- SignOff FareWell: #TUNES (Ping timeout for FareWell[d026.paris-121.cybercable.fr]) -:- SignOff smoke`: #TUNES (BitchX-75+Deb1an -- just do it.) -:- SignOff core: #TUNES (Ping timeout for core[core.suntech.fr]) -:- Downix [down@d-gnaps-101.ici.net] has joined #tunes morning all 01:20pm -:- eihrul [lee@usr5-ppp04.lvdi.net] has joined #tunes hi lo what's new? playing around with thread migration and it's taken off 40 instructions off my IPC path (of 110 instructions or so) so i'd say the concept is rather successful :) ok * Downix/#tunes is digging into ways of improving his chat server a bit found a routine for adding in a 3D interface to it. is that not extraneous? :) as these are intended for chatting, no? otherwise they can chat in Quake I know, but I'm trying to figure out ways of making it easier for non-UNIX guys to handle it. Right now raw telnet is slightly warped non-unixish guys can handle telnet not all, I deal with idiots all day looking at a possible plug-in for those who cann't handle it as/is 01:40pm they will remain such if you don't give them a reason to learn w/o removing the interface we have now. more features for those who go with the native interface. so how are you proposing to do your 3D? I'm not, I found someone else who made a 3D plug-in for his server so I was looking to see how hard it would be to add it into my own gah, that's no fun My chat server isn't for fun I do it because it'll help friends out so why can't it be fun and helpful at the same time? I have enough real work to do w/o going into detail on a smaller chat server project fair enough I do more work w/ F-CPU anyways that's where my fun is 8) * Downix/#tunes just finished up the ALU how is that project advancing? woo hoo! it's progressing quicker than I thought it would our assembler is done i've heard lots of conflicting stories on it... such as it's going dead, it's not quite dead yet... I have to submit the ALU once I've simulated it to death assemblers are a < 1 week job they're a harder job when you don't have an assembly language you made an assembler without an instruction set and/or encoding? no, we made both at once well, all at once that makes sense... or not quite we're developing it all kinda hap-hazard parts of it are dead others are going strong but with the assembler done, making a working CPU is possible right now, but not necessarily a great one simulated CPU anyway? right will check out the project again... fcpu.tux.org wasn't it? f-cpu.tux.org right off by one ok thou must stop saying 'ok' 01:50pm alright 8) much better alright 8) anyways right now I'm doing tests on possible ALU's we could use so far the straight and simple one works best as long as no math is done well, low-level math only the second mul, div, sqrt show up, it's in trouble it's not much difference than intel cpus i would guess... it takes way too many cycles for a mul/div as is yup 1 cycle for an add, close to 40 for a div, 10 or so for a mul I'm looking at removing all math functions from the INT core and putting them into the FPU the FPU's ALU is more akin to a vector unit's core so it runs very fast, more cycles per second than the INT unit on the same freq. so hopefully even tho it takes 12 cycles for a mul function, it'll get the results sooner than otherwise they have a logo... yup I like it isn't that premature? not really the logo is for the project we're never going to produce a chip ya know so is tunes' logo i would suppose only the cores for other people to make them would anyone even endeavor to make them? I know of a few universities in europe that are looking at it, and my company is going to be as well wouldn't it be expensive though? depends on how it's produced no sub instruction? hrm? in design docs not sure, I didn't look it all over, only took the instructions as is and made a straight ALU based on them 02:00pm I know we probably could use some more instructions, but as is I've not yet found a single function I could not handle oh... sub is in there ok OH! I got what you ment what i meant about what? sub? zero I thought when you said no subn instruction that we were missing sub-functions within instructions oops hehe, actually you could have no sub instruction if you coded add right to sub you merely make the second integer a negative from the design docs, the add instruction just has a negate flag right all you'd need saves transistors and speeds up the design by 0.5% i suppose the same is true of the shift instruction as well? Correct would make for a fun assembly language I know, it's a bit odd yup the GCC people on the project all say it'll be a lot more powerful for higher-level programming too you could just make flag/opcode combinations different mnemonics, but that would be too over-done * Downix/#tunes nods we didn't go overboard it would make assembler construction much simpler, though we simply took the most obvious routes, and then tested for speed and ease of use * Downix/#tunes nodsd do you actually have an emulator or simulator built yet? semi I'm doing a simulator while working on the ALU 02:10pm making a flexible one, so I can change pipeline depth, etc, with only a few changes that way I can better test speed * eihrul/#tunes wishes he knew more about how cpus are constructed... any particularly good books on the subject? get the book "COmputer Architecture, A quantitive approach" i own that :) most of us on F-CPU have it and swear by it 8) i snagged it while browsing around in bookstore smart just haven't had an adequate chance to read it yet right now we're also looking at making a modular surface interface for the core so you can mount multiple cores, multiple FPU's, etc. into a single die, PLUS allow you to change what the bus interface is to whatever you choose be is Socket-7, EV-4, etc. in theory tho UI'm not so sure it'll work but if they want to try it, go for it I'll stick to the F-Bus (modified MIPS SysAD bus) * eihrul/#tunes likes the logic instruction. * Downix/#tunes does too tis quite clever speedy too out does the not instruction work though? I'd explain, except X just died along with adobe grr one register is just not used? I believe that was the idea that register being %b? same idea that MIPS and Alpha ah used 02:20pm can't remember the reasons why, but something about speeding up the system if' you'd notice, we're mostly concerned about speed or at least a good part of us are another thing we're worried about is ease of implimentation no MTL, TTL, etc. which are? ways of handling the core functions Multiple Threaded Logic, Tiny Threadded Logic and Command Threaded Logic we're using CTL, we understand it better * AlonzoTG/#tunes ignites a flamethrower and eyes Downix most CPU's are CTL, Alpha is TTL, and SH-3/4 is MTL hey AL how do the three differ? eihrul: How threads are handled. i know... but how? :) ok, in MTL, the core has a pre-proicessor to set up the signals as threads, in TTL the core itself threads then as they go through it, in CTL you send commands in a specific pattern in order to thread it namely how much control the compuler has attempting to grok concretely... ok 02:30pm so MTL is just a bit less decentralized form of TTL? er more decentralized :) right, it's also a lot cheaper to make since you don't have to design the core so accurately and CTL depends on parallelism between different units, somewhat like pipelining? no, CTL depends on the software for all threads define "commands in a specific pattern" :) ok, say you have one command that requires the results of a mul function w/ a 12 cycle functioning. in CTL, the compiler would make sure that 11 commands slide between the command and the function okay... so i wasn't totally lost... right but that does kind of depend on parallelism between disimiliar units, no? yep, but that's another case very easy to make a simple emulator for this :) yup what exactly is the purpose of the size flags in loadi? can't the assembler do that? 02:40pm yes, but it makes high-level languages run a bit smoother how so? it's a constant*constant multiply :) I forgot why, but it did loopentry? that's an odd one... couldn't most compilers do that? not all intel compilers seem to get away fine without it 02:50pm hrmmm, how does one go about doing an unconditional jump? look it over eh? There's a flag for jump, I believe I'll look over the ALU ah... well it doesn't descirbe it in documentation :) I think so, I'll check or go to #f-cpu and ask mathias i'll just take your word for it 03:00pm -:- SignOff tmf: #TUNES (Ping timeout for tmf[barkebille.ii.uib.no]) downix: do you know of any good multi-windowed irc clients for X? not really, I use xchat client i'm using doesn't like multiple channels too well which one? irc2 ok, try BitchX then bitchx is too colorful i like monochrome better 03:10pm -:- eihrul_ [eihrul@usr5-ppp04.lvdi.net] has joined #tunes -:- SignOff eihrul: #TUNES (Leaving) -:- eihrul_ is now known as eihrul xchat is pseudo-multi-windowed yup, but it's nice 03:30pm -:- SignOff hcf: #TUNES (Leaving) -:- hcf [nef@me-portland-us843.javanet.com] has joined #tunes -:- NetSplit: lackey.openprojects.net split from sterling.openprojects.net [03:40pm] -:- BitchX+Deb1an: Press Ctrl-F to see who left Ctrl-E to change to [lackey.openprojects.net] -:- Netjoined: lackey.openprojects.net sterling.openprojects.net -:- AlonzoTG [Alonzo@216-164-131-242.s242.tnt2.lnhva.md.dialup.rcn.com] has joined #tunes -:- core [core@core.suntech.fr] has joined #tunes * AlonzoTG/#tunes ignites a flamethrower and eyes core hey core 04:10pm Gakuk! * AlonzoTG/#tunes is worrying about exactly how to get functions in his OS to call other functions... * Fare/#Tunes suggests the assembly instruction CALL okay.... call far lamefunc not all assembly sets have that command CALL NEAR leetfunc Downix: use a macro, and be done But How do I pass parameters? What calling conventions do I use? ATG: what language are you using? I am thinking about the C convention... the same as the surrounding language! But I would also like to experiment with LISP and any OS needs some ASM. hmm TUNES needs an assembly core... Please specify it so we can start writing it!!! 04:20pm -:- SignOff core: #TUNES (Ping timeout for core[core.suntech.fr]) -:- liar [brand@p0wer.qzx.com] has joined #tunes hey liar hey eihrul: u here? 05:00pm he's on f-cpu erm, #f-cpu here too I think he's ignoring this channel oh, ok eihrul: kevin says the newest version of bochs should work fine in the debugger ok eihrul: did u get the latest bfe? 05:10pm eihrul: zhivago had to fix a bug in structures.c that made it seg fault when u loaded more than 2 strucs or lists i am not infallible :) i have no idea how it was working tmp = (struct mem_struct *) realloc (memlist, sizeof(struct mem_struct) * (++nr_strucs)); in the sizeof i had (struct mem_struct *) er u had i did? ya and it was working for 2 lists alignment saves the day! alignment? either that or the data it was overwriting wasn't critical heh hmm * Downix/#tunes wonders whatever happened to that warhol painting that was done for Apple computer they probably used it to insulate their cheesy casing hehe Downix: i talked to several ppl and i dont think u should use the name commodore the image that came to their mind when i mentioned it was crappy old computer liar: I've been looking into a lot of things, and yes, that image is there that being the idiots that i talked to so is the image of cheap PC's for the home public image is only what you display to the public plus that name has gone down with 2 companies now, kinda bad luck if ya ask me I know where they failed same way that Apple failed marketing Nope, they were too large for anyone but the founder to handle in Commodore's case, they didn't re-hire the original CEO, and it went bankrupt 05:20pm in Apple's case, they did, and look at it now vs what was happening to it same time-table in both companies experts said if Jobs had not come back, Apple would have gone under last year and they are right, no inspiration, no central authority yup but commodore had very poor marketing tactics Commodore, when they fired their CEO, was the first billion dollar personal computer company they didn't before the firingg notice the well pointed ads with William Shatner and how they pulled the poorly designed PET before it ruined their image in the end, it's all who leads the company IBM's led by Nabisco's former CEO, and look at how it's tactics are working now -:- hcf [nef@me-portland-us843.javanet.com] has left #tunes [] and Apple is led by the biggest showman in the computer field PT Jobs 8) jobsum hehe heh * AlonzoTG/#tunes mows down everybody in the room... aahh! CharlieTG at 12 o clock, call in the napalm! haha WAR, glorious WAR!!!! * Downix/#tunes switches into war mode anyone know howto configure bochs for dynamic translation? * Downix/#tunes grabs his AKM and WWII army hat and charges om with his bayonet 05:30pm oh damn look at all the new debugger commands in the newer bochs fast-forward :) wonderful, latest bochs has errors 05:40pm fun 06:00pm bochs with errors, say it isn't so u any good finding out why source dont compile? sometimes it's a case-by-case basis usually how big a download is bochs again? i fixed one error were he assumes we will all use the external fpulib emulaton 500k u need to edit fpu/Makefile and remove the extra -I if u dont link to the linux kernel emulation hmm that was all? ya but when it tries to build the bochs binary it fails lots of missing functions during linking ok cookies...aggg brb eating hehe 06:10pm * eihrul/#tunes is wondering if liar used ./configure hehe so far only one frivolous compiler error ok ooo... 5 more K&Risms.... * eihrul/#tunes shudders. K&R? old-school C ok * Downix/#tunes didn't start C till he got Linux i haven't done any K&R C either... but it's something you just learn about actually, not till my talker went nuts because people, much like intel, are still trying to keep backwards compatability with decrepid old compilers though K&R prototypes do have their perks ok 06:20pm odd... somebody forgot to implement a few instructions in bochs it seems odd will just have to put some dummy instructions in and hope nothing breaks what instructions it's is mysteriously entitled ESC going to look at the opcode map to see what instruction it is really supposed to be the FPU escape? escape key? maybe it's an instruction to Bochs only? -:- SignOff AlonzoTG: #TUNES (Ping timeout for AlonzoTG[216-164-131-242.s242.tnt2.lnhva.md.dialup.rcn.com]) wow... you're right, they're in the cpu ah, there we go it seems someone forgot a single '.' in a makefile and everything puked geesh grrr, netscape crashed fun just as i was 100+ mails into the list too 06:30pm -:- AlonzoTG [Alonzo@216-164-131-242.s242.tnt2.lnhva.md.dialup.rcn.com] has joined #tunes -:- SignOff Downix: #TUNES (BitchX: shaken, not stirred) i have not followed the list this month.. 07:10pm -:- SignOff AlonzoTG: #TUNES (Have Nice Day :)) all-clean make clean only cleans the main dir i found that out too :) ugh so does rm `find | grep '\.o'` 07:20pm okay, got the undefined references think i should email kevin about it or u think u can fix it? gimme a few k i guess just mail him k 07:30pm -:- lar1 [LarMan@1Cust230.tnt31.sfo3.da.uu.net] has joined #tunes -:- lar1 [LarMan@1Cust230.tnt31.sfo3.da.uu.net] has left #tunes [] -:- hcf [nef@209.94.148.84] has joined #tunes -:- lar1 [LarMan@1Cust230.tnt31.sfo3.da.uu.net] has joined #tunes Hye hye :) Hows bfe, liar? its good Kevin P. Lawton wrote: > Looks cool! I didn't test it, but the screen snapshot on > your website looks sweet. Some day I was hoping to integrate > a GUI in the debugger, but so much to do... Heh, cool! -:- SignOff eihrul: #TUNES (Ping timeout for eihrul[usr5-ppp04.lvdi.net]) 08:00pm -:- eihrul [eihrul@usr5-ppp06.lvdi.net] has joined #tunes what int value is used for a good exit 0? k I'm not sure thogh sounded good to me lar1: EXIT_SUCCESS likewise, EXIT_FAILURE is use for a non-successful exit 08:10pm -:- SignOff eihrul: #TUNES (Ping timeout for eihrul[usr5-ppp06.lvdi.net]) eihrul: And what does EXIT_SUCESS equal? he is having inet troubles tonight :) :) kewl my new stack feature is working good auto selects TOS and keeps it in view -:- pyro [tcn@cci-209150250119.clarityconnect.net] has joined #tunes pyro! how's it been? bfe is finished watch abi answer that :) -:- eihrul [lee@usr5-ppp76.lvdi.net] has joined #tunes lar1: u figure out the resize problem yet? :) No, I haven't looked You want me too? ya Ok, I'll look tonight I booted in to win cause I needed somthing for my servers network card.. so I'll have to reboot ;) 08:20pm hmm lar1 & liar... l1ar eihrul: i dont think popen will fail i changed the name of BOCHS_PATH so bfe would fail and it didnt s/of/in/ pyro actually.. lar1 & liar == la`0 hey cool, I got a bunch of mail on retro eihrul: we got a problem here pyro: from zhivago?