00:00:00 --- log: started forth/04.07.10 00:07:40 --- join: arke (arke@melrose-251-251.flexabit.net) joined #forth 00:07:45 hi :) 00:08:37 hello 00:09:22 kc5tja: you didnt reply to my email :) 00:09:28 hi foobar 00:12:27 --- quit: Herkamire ("bed") 00:13:07 kc5tja may be 9asleep 00:13:41 naah, he'd be loged off 00:13:48 hes probably fetching food or reading a bok 00:13:54 bok bok 00:14:05 ah i got nothing, n/m 00:14:54 book* 00:15:43 he was discussing K and J languages earlier 00:17:24 similar to APL 00:18:44 hmm 00:18:48 forth is better. 00:18:52 hugh, i have spoken. 00:19:12 sorry, theres just something VERY intriguing about sitting on the toilet with grampas old laptop playing with pygmy 00:19:39 someday he would like to add J's matrix-handling powers to forth 00:24:20 --- quit: arke ("bye") 00:24:41 laptops rule! i'd hate to go back to a desktop machine 01:02:25 on the phone 01:18:54 --- join: qFox (C00K13S@cp12172-a.roose1.nb.home.nl) joined #forth 01:48:12 * aum waits for a wristwatch running forth, with GSM and 802.11, plus neural interface 01:55:02 Well, I think I'm going to go to bed. 01:55:14 That was a really long phone call. :/ 01:56:04 --- quit: kc5tja ("THX QSO ES 73 DE KC5TJA/6 CL ES QRT AR SK") 02:51:30 --- join: Serg (~z@193.201.231.126) joined #forth 02:51:50 re.. 02:52:48 --- quit: Serg (Client Quit) 03:10:35 over to you, 4onetom_ 03:11:12 --- part: foobar left #forth 04:13:14 --- join: crc (crc@0-1pool176-15.nas6.philadelphia1.pa.us.da.qwest.net) joined #forth 04:28:01 --- join: Topaz (~top@spc1-horn1-6-0-cust217.cosh.broadband.ntl.com) joined #forth 04:28:52 Hello Topaz 04:28:53 --- join: T0paz (~top@spc1-horn1-6-0-cust217.cosh.broadband.ntl.com) joined #forth 04:49:02 --- quit: T0paz ("Leaving") 04:49:07 lo 04:53:44 What's up? 05:03:24 tis thundererering outside :D 05:04:43 It's clear (and HOT) where I'm at :-) 05:05:02 At the moment, I'm starting to work on my assembler again... 05:06:09 i'm going to play with postgresql some more i think 06:09:44 --- join: snowrichard (richard@adsl-068-209-159-248.sip.shv.bellsouth.net) joined #forth 06:09:54 * crc greets snowrichard 06:09:57 hi 06:10:13 --- join: I440r_ (~mark4@168-215-246-243.gen.twtelecom.net) joined #forth 06:10:33 I think the dpans94 document could cure my insomnia pretty quick 06:11:04 Hi snowrichard, I440r_ 06:11:13 ! 06:11:18 snowrichard: :) I take it you're not an ANS fan? 06:11:53 I have a limited attention span. I don't think I could sit here long enought to read the whole thing 06:12:15 * Robert is not so sure any sane person would. 06:12:31 who said I was sane :) 06:14:12 Hm, right, you DO hang in #forth... 06:15:05 think I'll go get another ~==== 06:17:38 * Robert interprets that as a hairy snake, or possibly a joint. 06:24:21 lol 06:30:41 no joints since august 2003 06:30:54 just a cigarette 06:31:55 and coffee refill ~0) 06:32:11 :) 06:32:24 * Robert thinks about what else a Forth system might need. 06:32:31 a coffeepot 06:32:48 Hm, yes, I'll implement one of those. 06:32:59 : quit ." ~O)" recurse ; 06:33:17 coffee not found, operator halted 06:34:13 my kitten won't sit anywhere else in the room but my lap 06:34:46 How cute..and possibly annoying in the long run. :) 06:35:30 what is annoying is when she walks on the keyboard 06:37:08 is that recurse tail optimized so you don't blow the stack? 06:37:56 Nope. 06:38:00 And I blame I440r_ for that. 06:38:21 My sister's hamster sometimes gets to walk on my keyboard, but it can't even press the keys down. :) 06:38:55 I had a hamster once but it died an untimely death being transported in a hot car 06:40:55 :( 06:41:12 My sister's first hamster ran into a hole in the wall and never returned. 06:41:25 freedom 06:42:28 lol 06:42:38 #hamster 06:42:52 we dont need no stinkin recursive coffee! 06:43:54 You don't like tail recursion, do you? 06:43:59 we have a new resident who is a bit of a redneck and he has a great sense of humor 06:44:45 Nice. Does he use Forth? 06:44:55 he doesn't use computers 06:45:05 Oh... :) 06:45:21 just had a good laugh listening 06:45:42 tail optimizations are good when done BY HAND 06:45:48 isforth has a word to do just that 06:45:58 goto? 06:46:01 yes 06:46:13 Can you goto to the current word? 06:46:16 i might add ;goto :) 06:46:19 yes 06:46:31 erm maybe not lo 06:46:52 err i probably wont add ;goto 06:46:55 could call it spaghetti 06:47:14 goto isnt like goto in basic 06:47:18 basic had NOTHING BUT goto 06:47:31 goto just converts a call/ret into a jmp 06:47:37 yea 06:47:39 actually into a jmp/ret lol 06:47:49 the exit is still compiled in by ; 06:47:56 actually, i can make it NOT compiled in 06:48:13 goto can set a flag and ; can test the flag and only compile the exit if there was no goto 06:48:32 but why add that extra complexity 06:48:45 who cares about 4 bytes of code that will never be executed lol 06:49:17 Coding stuff without subroutine calls is pretty annoying. 06:49:51 basic had goto and gosub/return 06:50:33 I wrote a preprocessor for the Coco Basic one time. 06:50:38 to add structure 06:50:56 used the : to indent 06:50:56 :) 06:51:07 one of microsofts old basics was coded in forth 06:51:11 might have been quickbasic 06:52:01 the coco floppy only held 156K 06:52:29 Cool. 06:52:34 I440r_: I wasn't refering to BASIC 06:52:39 40 tracks single sided 06:52:40 oh 06:52:50 brb 06:52:54 I440r_: But I had a bet with a guy, about how much RAM you need for writing a PONG game. 06:53:12 I440r_: He didn't think 40 bits was enough. The fool. 06:53:39 And that machine we simulated (4-bit) had no subroutine calls or indirect jumps. 06:53:40 the kim-1 chess program ran with 1.1K 06:53:52 That's pretty good. :) 06:53:59 Including an AI player? 06:54:13 it played ok 06:54:23 slow, 06:54:35 1 mhz clock 06:55:01 What computer? 06:55:06 6502 single board 06:55:19 1977 06:56:32 6502 is like 284562865429 * better than a z80 06:56:34 Nice. 06:57:10 not much real estate for your stack though 256 bytes 06:57:23 that helps 06:57:34 your not likely to use recursion when you have limited stack :) 06:58:04 6052s are fun... 06:58:09 65816s are better though :D 06:58:32 Steve Wozniak bought a copy of my assembler 06:59:35 back when he was still working in his garage :) 07:00:26 Heh, neat. 07:01:02 talked to him on the phone 07:01:11 So we have (in part) you to thank for that IMB/Microsoft won the war. ;) 07:01:36 this was before there was a pc remember 07:01:49 Yes, I'm just kidding. 07:02:11 my dad's uncle patented Ti 07:02:15 's first calculatore 07:02:25 Jerry Merryman 07:02:42 --- quit: Topaz ("Leaving") 07:02:49 * Robert has a TI-82.. with a BASIC interpreter *shrug*. 07:03:08 I had an HP-41C one tiem 07:03:10 time 07:03:38 it was RPN notation with alphanumeric capabilities 07:04:47 Nice.. Never had any HP calc. 07:05:32 sold it when I was on my way out of the navy 07:07:44 I had a little rinkydink pocket pc with basic from Radio shack too 07:09:41 --- quit: snowrichard ("Leaving") 07:19:55 --- join: snowrichard (richard@adsl-068-209-159-248.sip.shv.bellsouth.net) joined #forth 07:23:52 --- quit: snowrichard (Client Quit) 07:27:10 it was a hard long fight, but finally both me and the stairs are red now, but he got more! 07:28:47 :D 07:29:07 So back to Forth now? 07:29:25 can i take a break first? 07:29:33 any idea how much work it is to paint a goddamn stairs? red? 07:29:47 i kinda underestimated it, took me 5 hours straight 07:30:40 that includes leaning against the stairs and then against the wall forcing you to clean the wall asap without making an even bigger stain 07:30:54 btw, red + white makes pink. 07:30:55 :p 07:32:40 >:) 07:33:33 but you know whats worst 07:33:43 i taped the sides of the walls where i had to paint 07:33:55 so i wouldnt go paint the walls in corners and such 07:34:09 but at one point i went over the tape (wasnt very wide) 07:34:11 just a little 07:34:17 but i knew you'd see it for ever and ever 07:34:23 and you cant just rip off the tap 07:34:24 e 07:34:29 because ... well duh 07:34:41 and you cant wait till you can, because by then it'll be there, for ever and ever 07:34:53 so you go get paintremover and some cloth 07:35:22 but the stain is so close o the rest of where you painted that the cloth drags thru it, polluting more 07:35:27 aaaaargh 07:35:33 Poor, poor qFox. 07:35:52 But at leat you found something that's worse than working in F. 07:36:01 agreed 07:36:23 physical labour < mental labour 07:37:38 Heh. 07:37:46 That can be interpreted in many ways. :) 07:37:47 anyways, i have been thinking, and i think i've thought of how to do it 07:38:12 so at least i can code and be in my own happy place for a while now :p 07:38:26 :D 07:39:05 i'll define a misc set of machine forth words 07:39:21 and use that as a basis, threading everything else 07:39:45 i'll use F's own execute 07:40:08 Sounds nice... 07:40:17 one of the walls i hit last night 07:40:25 was the fact that i'm using a different stack then F 07:40:42 so i couldnt just use DUP and stuff like that, since they would not have the effect i need :) 07:42:31 it wont be particulary fast, but whatever 07:42:37 Hehe 07:42:47 Have fun ;) 07:42:55 muhu 07:43:12 If you want you could code a MachineForth -> native code compiler. 07:43:33 nah thats ok 07:52:07 --- join: onetom (~tom@novtan.bio.u-szeged.hu) joined #forth 07:52:09 --- quit: onetom_ (Read error: 54 (Connection reset by peer)) 08:09:40 --- quit: I440r_ (Read error: 110 (Connection timed out)) 08:30:50 --- join: I440r_ (~mark4@168-215-246-243.gen.twtelecom.net) joined #forth 08:32:29 --- part: qFox left #forth 08:32:29 --- join: qFox (C00K13S@cp12172-a.roose1.nb.home.nl) joined #forth 08:50:02 --- join: Serg (~z@193.201.231.126) joined #forth 08:51:12 hi ! 08:51:20 did anyone crack w/ debugger ? 08:51:32 Nope 08:51:36 i need to remove 30-second limit from old Fraunhofer MP3 coder 08:52:48 i set breakpoint on ExitProcess in Numega SoftICE, it breaks... but how can i go back by callstack and find the 30sec check ? 08:52:56 * crc finally made progress on the RetroForth assembler 08:53:31 Nice 08:54:27 This is going to take quite a bit of time to write... 08:58:57 --- quit: crc (Read error: 104 (Connection reset by peer)) 09:00:10 --- join: crc (crc@0-1pool176-20.nas6.philadelphia1.pa.us.da.qwest.net) joined #forth 09:06:13 --- nick: TomasuDlrrp -> Tomasu 09:06:15 * Tomasu is back (gone 10:28:37) 09:20:21 * Tomasu is away: Just casue 09:29:20 --- join: Herkamire (~jason@h000094d30ba2.ne.client2.attbi.com) joined #forth 09:33:22 --- quit: Serg () 09:39:36 --- quit: crc (Read error: 110 (Connection timed out)) 09:41:53 * Tomasu is back (gone 00:21:31) 10:01:51 --- join: kc5tja (~kc5tja@66-74-218-202.san.rr.com) joined #forth 10:01:59 --- mode: ChanServ set +o kc5tja 10:34:39 --- quit: I440r_ ("Leaving") 10:41:22 --- join: snowrichard (richard@adsl-068-209-159-248.sip.shv.bellsouth.net) joined #forth 10:41:27 hello again 10:42:28 Hi snowrichard 10:43:04 got my little peer to peer voicemail program working with a Glade Gui 10:47:36 --- join: tathi (~josh@pcp02123722pcs.milfrd01.pa.comcast.net) joined #forth 10:48:02 * snowrichard greets tathi 10:49:14 Hi snowrichard 10:51:27 learning sockets programming and gtk 2.0 10:51:57 Is gtk worth learning? 10:52:23 it seems to be a lot easier to use than java swing. that is the only other toolkit I have tried. 10:52:54 I rather like gtk 10:53:11 --- join: SDO (~SDO@67-23-111-213.clspco.adelphia.net) joined #forth 10:53:24 Well, a lot of the little glue macros aren't documented...but you can usually just guess those. 10:53:31 OK, I had to do some MFC at school. 10:53:52 as in Micro$haft? 10:55:08 I had Visual C++ for Windows once. 10:55:13 did not get far 10:55:34 excuse me C not C++ 10:56:02 Was in the programming class... When I should do some graphics stuff I couldn't find any good way to get a framebuffer to work with. 10:56:07 So I left and never returned. 10:56:26 drop or just leave? 10:56:44 Just didn't show up anymore. 10:56:56 But the teacher was nice. 10:56:57 doesn't help the gpa 10:57:19 He gave me a good grade anyway. That course was for braindead monkeys anyway. 10:57:34 So by just showing I understand what a computer is I passed. 10:57:44 you mean a MCSE? 10:58:16 No, it was just an ordinary programming class in high school, and they happend to use Visual C++ 10:58:36 The morons in the government forces ALL school to teach GUI programming. 10:58:43 when I was in High school there weren't even PC's yet 10:59:15 So then I guess the teachers knew at least a little bit about what they were teaching. 11:00:08 in my networking curriculum the teacher and I both took the Novell CNA and he flunked it. I think I missed 1 question 11:00:30 they said it was the highest score they had seen at that testing center 11:01:23 Hehe... People still haven't got used to computers. 11:01:51 i need a sandwich be back in a bit 11:01:56 OK, see you 11:02:09 * Robert is trying to hack together a Makefile 11:15:32 make? eeew 11:16:55 Hehehe. 11:17:23 It gets worse - I'm using C (just like, ehm, certain other persons in here). 11:17:43 if C is the right tool for the job, then why not use it? 11:18:00 i'm opposed to C being used to write programs like evolution and apache though. 11:18:15 evolution? eeew ;) 11:18:32 I'm apposed to programs like evolution 11:18:35 yeah, its so bloated 11:19:16 There's nothing too wrong with make. 11:19:33 using tab -vs- spaces for syntax 11:19:37 tathi: except it's handling uf subdirectories 11:19:48 Yeah, as long as you don't use it recursively. 11:19:52 ant has some nice concepts 11:19:53 But that's not so hard to work around. 11:20:00 but they don't implement them well enough 11:20:07 I was looking at ant recently, looked neat. 11:20:22 Well, for some part of this projects (dealing with SDL graphics) I think C does a good job, but for others (a Forth VM) I'd rather use something else. 11:22:31 --- quit: snowrichard ("Leaving") 11:26:43 --- nick: lalalim_ -> lalalim 11:36:48 Robert: i also made tile forth a friend of sdl graphics 11:37:02 Robert: what r U using it 4? 11:38:25 Dude, I used Evolution, and good grief, I've *never* said this before, but I actually overwhelmingly prefer Outlook to Evolution. Overwhelmingly. Evolution is BBBAAAAAAADDD. 11:38:51 --- join: I440r_ (~mark4@168-215-246-243.gen.twtelecom.net) joined #forth 11:39:33 Well, I think some pretty massive changes will be made to the FTS1001 microprocessor design. 11:39:36 onetom: You mean SDL? A game. 11:39:47 I'm really glad I played with this architecture in an emulator before committing to hardware. 11:39:56 There are a number of things I'm going to do differently. 11:40:15 First, you can pretty much kiss the XOR-masked effective addresses goodbye. 11:40:48 It's just too hard for people to grok 11:41:04 hehe 11:41:26 so no more free jumps? 11:41:34 I have not yet decided if I'm going to use PC-relative addressing or not. I'd LIKE to, but it does introduce an additional layer of logic. 11:42:11 kc5tja: Nope. 11:42:47 While free jumps were nice for streamlining fetching of subsequent instructions, it made it such that a jump to anywhere else suffered a one cycle penalty (this includes CALLs and returns too). 11:43:07 So I'm going to invest the transistors to make calls/returns faster, which as a natural consequence, makes all branches faster. 11:43:21 Note that jumps CAN still appear in any slot though. 11:43:30 Even though the assembler doesn't put them there. :) 11:43:55 The next big change is how the P register is interpreted. 11:44:22 The contents of the P register will most likely point at the *currently* executing instruction, rather than the *next* instruction. 11:44:29 Robert: yup, i meant sdl. &what kinda game? 11:44:40 This also contributes to making calls/returns much faster. 11:46:04 The P register will have a dedicated +1 adder circuit bound to it to facilitate this. More transistors, yes, but also much faster instruction. 11:46:17 And finally, the various entry points for interrupts will all disappear. 11:46:26 Reset will force the P register to $00000000 as it currently does. 11:46:34 onetom: You go aroud in the solar system. :) 11:46:46 However, IRQ and NMI interrupts will cause the contents of the P register to be logically *exchanged* with another register, IP. 11:47:26 The contents of the data and return stacks are left completely unchanged. 11:48:14 Interrupt response times are guaranteed to not exceed five clock cycles. 11:48:25 (and can be as little as one clock cycle). 11:48:47 The reset code is responsible for initializing the initial value of the IP register. IP maintains itself thereafter. 11:49:08 * Tomasu is away: Just casue 11:49:34 A new instruction will be added to the CPU to support returning from the interrupt handler to the mainline code ("xip"), which merely (logically) exchanges the IP and P registers. 11:49:46 Hence, an interrupt handler is not a subroutine, but rather, a co-routine. 11:50:20 [20:42:11] kc5tja: Nope. 11:50:24 hearing voices? 11:50:25 :p 11:50:36 when they say jump, dont. 11:50:38 11:30 < qFox> so no more free jumps? 11:50:56 i know ;) 11:50:57 I'm hearing yours. 11:51:11 jump 11:52:01 anyway, this is one of the reasons why I wrote the CPU emulator -- to learn what needs to be fixed before I commit the thing to actual hardware. 11:52:23 Robert: it sounds very serious, coz i suppose its gonna b realistic, right? 11:52:53 Somewhat, I hope. 11:53:13 As much as my programming skills and CPU power allows. ;) 11:53:41 as long as no file reading is involved, you'll be fine :p 11:56:02 * qFox out 11:59:07 qFox: ? 12:08:20 --- join: snowrichard (richard@adsl-068-209-159-248.sip.shv.bellsouth.net) joined #forth 12:11:16 --- quit: snowrichard (Client Quit) 12:16:26 maaan. I *liked* the XOR-mask branch stuff. 12:17:37 Well, I can certainly keep it. 12:17:48 But it seems hard for people to understand. 12:17:55 I like it too 12:18:12 yeah, I can see that. 12:18:17 It was neat though :) 12:18:27 Well, I'll keep it, I guess. 12:18:40 Whatever seems to make the most sense... 12:18:50 I mean, you two are the only other people who've come the closest to actually doing anything with the emulator, so you get priority. :D 12:19:03 yay :) 12:19:14 --- join: snowrichard (richard@adsl-068-209-159-248.sip.shv.bellsouth.net) joined #forth 12:19:18 * Robert feels guilty 12:19:36 Robert: Why? 12:19:53 we had our dinner a bit early today. barbecue pork and chicken with rice 12:19:53 I don't see what's so complicated about it.... the branch tarked is IP xor MASK 12:20:02 mask is the bits that are left in the instruction 12:20:22 s/tarked/target 12:20:41 kc5tja: Because I haven't been very active on the Kestrel front... 12:20:51 But I stand by my decision to make P point to the currently executing instruction instead of the next. 12:21:10 Robert: Well, don't worry about it. Kestrel is NOT yet ready for prime time. Lots of things are still changing. 12:21:24 what's the logic behind that? 12:22:20 tathi: If P were specified to point to the next instruction, then any branch instruction would take two cycles. In cycle 1, the value P+1 is evaluated while concurrently fetching the instruction at P. In cycle 2, P := P+1. :) 12:22:59 With P pointing to the current instruction, then P and the address bus can be loaded from the same source, fetching the desired instruction packet concurrently to loading P. 12:23:11 Then P+1 can be evaluated while the fetched instruction is evaluating. 12:23:28 Thus allowing true, 1-cycle control flow. 12:23:45 one-cycle control flow is cool 12:24:21 Herkamire: And I think this is why PowerPC lacks a post-increment update on registers. Notice that the scheme I just described with the program counter is essentially how PowerPC's "load and update" system works. 12:24:49 Only in my system, the offset is fixed at the value '1'. :) 12:25:34 --- quit: snowrichard ("Leaving") 12:26:09 heh. It _constantly_ annoys me that PPC lacks a post-increment update. 12:26:33 Yeah, my A, B, and R registers do proper post-incrementing. 12:26:38 But it's not a requirement for the PC register. 12:26:42 right 12:27:11 Actually, I'm quite surprised with how stable the FTS1001 design actually is. 12:27:47 I mean, the XOR mask issue was for the benefit of the human programmers, and had no concrete technological consequence. 12:27:52 --- quit: I440r_ (Read error: 104 (Connection reset by peer)) 12:28:12 There is a certain nicety to being able to use absolute addresses right in the encoding though. :D 12:28:40 I figured that since I no longer depended on the "implicit jump", I could do away with the XOR mask idea. 12:29:03 yeah, that makes sense then. 12:29:37 And then there is the interrupt handler support in the FTS1001 design. Man, I'm so glad I didn't start to implement the old approach in the emulator yet. I *LIKE* the co-routine idea A LOT. 12:29:50 It simplifies the interrupt-related logic a huge amount. 12:30:02 :) 12:30:40 Leave the data and return stacks as-is, and just switch between which instruction pointer to use. P for regular program instructions, and IP for interrupt handling instructions. One new instruction is needed to toggle between them. 12:30:53 how have you gotten rid of the implicit jump? 12:30:55 P and IP both become special purpose registers as a direct result. 12:31:16 Herkamire: A separate state machine in the electronics will monitor when to fetch the next instruction. 12:31:31 kc5tja: ok 12:31:52 kc5tja: does that make it faster? 12:32:23 kc5tja: or do you still need a cycle to fetch the next instruction word? 12:34:36 Well, in my naive implementation, it'll be about as fast as the old scheme, but it'll be intelligent about which instruction pointer to use to fetch instructions with (which the old scheme couldn't deal with at all), and allows the optimization of fetching the next instruction one cycle early *IF* the last instruction in a slot is not a memory accessing instruction. 12:35:26 Here's my plan for the FTS1001 CPU: 12:36:04 * The FTS1000 "CPU" is just the raw execution engine. Instruction pointer, two stacks, three address registers, minimum instruction compliment to deal with these resources. Have a nice day. 12:36:55 The FTS1000 defines the "base architecture" on which all other CPUs in the same family are built on (e.g., like the PowerPC 601 was for all the 32-bit PowerPCs, and most of the 64-bit PowerPCs). 12:37:26 It also has the S! and S@ instructions to set and fetch special purpose registers. However, no SPRs are defined for the FTS1000. 12:37:54 * The FTS1001 CPU is FTS1000-compliant, but supports interrupt handling and partial address decoding. 12:38:27 This way, it makes it easy to document, I think, too. 12:38:48 I can just write a single document on the FTS1000, and a separate, smaller, document for the FTS1001 that just details its differences. 12:39:17 This also lets me expand the line up even higher: the FTS1002 need not have the same interrupt handling facilities as the FTS1001. This lets me implement memory protection if I wanted to, for example (And I will, but not that soon.) 12:39:43 So, basically, what I have running in my CPU emulator is an FTS1000. 12:39:59 (if that makes any kind of sense. :D) 12:40:34 cool cool 12:40:56 how are the three address registers used? 12:41:04 However you want. 12:41:14 wow 12:41:19 I devote almost half the opcode space to working with them in as orthogonal a manner as I could muster. 12:41:30 A, B, and the top of the return stack, R, are all used as address registers. 12:41:39 My design goal was to facilitate 3-operand *vector* operations. 12:41:55 that sounds very powerful 12:42:03 "A" is also used to address the CPU's SPRs too. 12:42:17 A! S! 12:42:18 or 12:42:22 A! S@ 12:42:36 I see 12:42:57 I am very glad I implemented at least A and B though -- they *GREATLY* simplified the implementation of displaying font data to the screen. 12:43:23 will stores/fetches using the address on the return stack pop the address off the rstack? 12:43:29 No. 12:43:36 R is considered just another address register. 12:43:58 neat 12:44:04 There is no R! or R@ instruction though, like there is for A and B. So : R@ R> DUP >R ; but other than that... it's the same. 12:44:34 right 12:45:07 what incrementing/decrementing stores/fetches do you have? 12:45:11 For this reason, I consider R to be a "destination"-optimized register, while A and B are considered "source" registers, since things like A@ N + A! are quite common. 12:47:04 Herkamire: Now, the first generation of the Kestrel is entirely CPU-bound -- the CPU does everything. 12:47:31 Future models will actually have an expansion interface (assuming there is demand for such a future model). 12:47:55 what sort of things could it be expanded to do? 12:48:12 are we talking new instructions? joysticks? 12:48:28 I mean is there a store that increments the address and stores? 12:48:28 The I/O architecture will be based a mainframe concept of "channels." 12:48:56 The memory accessors are: @A @B @R @A+ @B+ @R+ !A !B !R !A+ !B+ !R+ 12:49:29 ahh, that's what I was looking for. so they all have post-increment stores and fetches? 12:49:36 Yes. 12:49:44 cool 12:49:51 how many instructions do you have now? 12:49:55 As far as expansion, it'll be expandable with plug-in cards. 12:49:59 About 48 or so. 12:50:53 Each card is dumb -- no bus master circuitry on it at all. 12:51:26 Instead, high bandwidth access to main system memory will be through a coprocessor specialized to the task. Memory fetches are expected to be approximately as fast as a normal DMA channel would be. 12:52:23 This satisfies my goal of having the system be trivial for users to hack and build cool things with EASILY and AFFORDABLY. 12:53:18 yay :) 12:53:24 (since each slot will be under the control of the program running on the I/O processor(s), the CPU is free to do other things while the I/O processor's program blasts data to the appropriate devices) 12:53:44 So I can build a video card, for example, without having any frame buffer memory on it at all. :) 12:54:20 For the first-gen Kestrel, you might actually see shades of this architecture already, since I've yet to design the video hardware. 12:55:00 Imagine a Kestrel having 8 serial ports that can run at 115200bps each, but not a single UART anywhere in the system's I/O map! :D 12:56:22 And how about a printer port that sends data to the printer independently of whatever the CPU is doing? 12:57:43 Of course, the option for CPU-based I/O will always remain. 12:58:00 But having a dedicated I/O coprocessor (or set thereof) is an awfully powerful concept. 12:59:03 I expect I/O coprocessors to be a transport triggered architecture though. Stack is just too slow for that application. 13:01:23 sounds like you have plenty to keep you busy for a while 13:01:55 sounds awesome 13:03:57 I said I was sick of the PC architecture. 13:04:00 I fucking meant it. 13:04:11 :) 13:04:17 so people could make plugin cards that would add to to the i/o proccessor? 13:04:23 eg a graphics card 13:04:29 I WANT MY AMIGA BACK, and I want it better than it was before! KESTREL! 13:04:52 Yes, the I/O processor would reside on the motherboard as a standard part of the computer. 13:05:16