00:00:00 --- log: started forth/05.05.17
00:32:24 --- join: Topaz (~top@sown-90.ecs.soton.ac.uk) joined #forth
00:38:45 --- quit: bbls ()
00:45:40 --- join: bbls (~bbls@80.97.121.4) joined #forth
00:49:47 --- join: segher (~segher@blueice3n1.de.ibm.com) joined #forth
01:54:11 <virl> hmm, theoretically the forth dictionary could be used as for tile graphics or other graphics. But isn't that slow?
02:03:39 --- join: Robert_ (~snofs@c-f778e055.17-1-64736c10.cust.bredbandsbolaget.se) joined #forth
02:04:23 --- nick: Robert_ -> Robert
02:05:27 <bbls> what do you mean to use it for tile graphics?
02:05:39 <bbls> you alloc thigs normally, not in the dictionary
02:05:49 <bbls> in dictionary you put just the address of the words
02:09:28 <virl> ah, yes. sry, I'm a little bit confused about forth, because I'm relative new to forth.
02:18:33 <bbls> np :)
02:27:22 --- quit: nastm (Read error: 110 (Connection timed out))
02:28:58 <virl> Forth is an amazing programming language, so simplistic
02:29:52 <bbls> simple not simplistic :)
02:31:29 <virl> ah, ok.
02:32:36 --- join: Raystm2_ (~vircuser@adsl-69-149-50-250.dsl.rcsntx.swbell.net) joined #forth
03:32:06 --- join: snoopy_1711 (snoopy_161@p54A7DA53.dip.t-dialin.net) joined #forth
03:45:25 --- quit: bbls ()
03:59:48 --- quit: Snoopy42 (Read error: 110 (Connection timed out))
04:00:09 --- nick: snoopy_1711 -> Snoopy42
04:15:12 --- join: AlexF (~Alex@196-220.suscom-maine.net) joined #forth
04:17:51 --- quit: KB1FYR (Read error: 110 (Connection timed out))
04:30:34 --- join: aum (~aum@60-234-138-239.bitstream.orcon.net.nz) joined #forth
04:31:00 --- part: aum left #forth
06:15:01 --- quit: Frek ("Client exiting")
06:15:40 --- join: Frek (~anvil@h254n2fls31o815.telia.com) joined #forth
06:54:44 --- join: bbls (~bbls@80.97.121.4) joined #forth
06:54:45 <bbls> hello
07:59:30 <Robert> Hi
07:59:40 <bbls> hi Robert
08:24:29 --- join: tathi (~josh@pcp01375108pcs.milfrd01.pa.comcast.net) joined #forth
08:54:17 --- quit: saon|smgl (Read error: 104 (Connection reset by peer))
08:56:16 --- nick: Raystm2_ -> nanstm
09:02:43 --- join: saon_ (~saon@c-66-177-224-130.hsd1.fl.comcast.net) joined #forth
09:11:23 --- join: snoopy_1711 (snoopy_161@p54A7F4E7.dip.t-dialin.net) joined #forth
09:28:18 --- quit: segher ("Leaving")
09:33:49 --- quit: Snoopy42 (Read error: 110 (Connection timed out))
09:34:02 --- nick: snoopy_1711 -> Snoopy42
09:40:56 --- join: vitaminmoo (~vitaminmo@dsl-95-113.peak.org) joined #forth
09:42:25 <warpzero> aw
09:42:28 <warpzero> no i440r
09:44:30 <Robert> Poor us.
09:47:00 <warpzero> yeah well
10:11:40 --- quit: saon_ ("Lost terminal")
10:45:15 --- join: solar_angel (~jenni@CPE0000e812679b-CM000a7362da55.cpe.net.cable.rogers.com) joined #forth
10:48:06 <Robert> Hi.
10:48:21 <solar_angel> hiya
10:52:11 <virl> what's the difference between interpret, execute and next?
10:52:41 <solar_angel> hrm... i don't think my FORTH had either interpret or execute, now that i think about it.
10:52:45 <Robert> execute - pick a word address from the stack and call that word.
10:53:12 <solar_angel> oh right
10:53:16 <solar_angel> i had that one.
10:53:23 <Robert> next - execute the next instruction/word. This can be "built into the processor" in case of a native code compiler.
10:53:51 <Robert> Otherwise it's usually a macro or small routine that is jumped to.
10:54:15 <Robert> "interpret"... is there any standard meaning? I often use that word for internal stuff in my Forths.
10:54:33 <solar_angel> i don't recall...  but then, i never was an ans-* fan.
10:55:18 <virl> me too, but I'd like to know more about what's going on under the 'hood' of a forth system.
10:55:51 <Robert> Did you read "Moving Forth"?
10:55:59 <Robert> Or perhaps "Build your own Forth".
10:55:59 <solar_angel> woo, "moving forth" rocks.
10:56:06 <solar_angel> i've read both of those... :)
10:56:10 <solar_angel> of course, i've built my own forth.
10:56:14 <solar_angel> good starting points both.
10:56:16 <virl> no, I haven't read these books.
10:56:21 <solar_angel> they're webpages, actually.
10:56:25 <solar_angel> available online
10:56:41 <Robert> Yes, both are good at explaining the concepts.
10:56:48 <Robert> As are the people in here :)
10:57:06 <Robert> When you manage to get us to stop fighting about definitions and actually explain the stuff.
10:57:30 <solar_angel> heh
10:57:55 <solar_angel> something i learned from FORTH.. shun float math.
10:58:02 <solar_angel> yesterday i wrote a blazing fast spline renderer.
10:58:07 <solar_angel> entirely without floating point math
10:58:11 <solar_angel> hell, it doesn't even use division
10:59:25 <virl> hmm, nice. I'm thinking of not to implementing those things like division/multiplication
11:00:03 <Herkamire> my memory if INTERPRET is that you pass a string or something, and it parses and executes the code
11:00:18 <Robert> Floating point can be useful, though. 
11:00:31 <Robert> And not even slow anymore.
11:00:56 <solar_angel> Robert - well true, sortof...
11:01:07 <Herkamire> yeah, on ppc floating point is supposedly faster than int
11:01:11 <solar_angel> in 3D graphics, there are a few things that work best with floats...
11:01:25 <virl> I have at the time a couple of primitive words, which won't be searched in a dictionary, they are all bytecodes because I need to implement a vm on which the forth system can run under.
11:01:27 <solar_angel> Herkamire - well that's just nasty...  int math is far more generally important.
11:01:32 <Robert> solar_angel: Don't think I ever used floating-point outside that.
11:01:48 <solar_angel> that said, there are ways of parallelizing integer math (and just plain using SIMD) that make things even faster.
11:01:59 <solar_angel> Robert - well, i use floats for things like quaternions...
11:02:07 <solar_angel> doubles, actually, just 'cause.
11:02:20 <solar_angel> (the things where accuracy is more important)
11:02:30 <solar_angel> but for just smearing lines on a screen, i'm all about integer math
11:02:40 <Robert> Yes.
11:02:54 <Herkamire> solar_angel: the SIMD on my box does floats just as well as ints
11:02:58 <Herkamire> altivec
11:03:20 <Robert> I once profiled a 3D engine, and found that the floating-point to integer conversion was where most time was spent.
11:03:26 <solar_angel> Herkamire - SSE2 does too...  but floating point math has many problems, not just speed.
11:03:44 <solar_angel> Robert - depending on the engine, that might be misleading.
11:03:48 <Robert> And that in a program doing numerous 3D transformations and polygon drawing.
11:03:59 <Robert> solar_angel: In what way?
11:04:18 <solar_angel> the x86 floating-point unit has *always* been seperately pipelined
11:04:27 <solar_angel> so CPU-specific optimizations matter a lot.
11:04:45 <solar_angel> you have to sync the two units to do a conversion
11:04:50 <solar_angel> (or at least, that was once true)
11:04:59 <solar_angel> which means you were probably seeing FWAIT's eating up a lot of time
11:05:04 <virl> how much cycles does your implementation of 'drop' takes? mine 4 cycles raw.
11:05:16 --- quit: Topaz (Remote closed the connection)
11:05:21 <bbls> solar_angel i tought fwait is not required in post 486
11:05:24 * solar_angel misses her old optimizing forth...
11:05:30 <solar_angel> it took zero cycles for drop. :P
11:05:52 <bbls> zero?
11:05:56 <solar_angel> bbls - well, it's not explicitly required no...  but that just means that it's being implicitly executed when you sync the two.
11:06:01 <solar_angel> yes, zero.
11:06:21 <solar_angel> optimizing stc forths tend to do that sort of stack juggling at compile time.
11:06:47 <bbls> you can't have zero cycles unless you inline the previously called word and remove any unused parts (the ones that would generate the TOS)
11:06:51 <Robert> solar_angel: Actually, I think I tracked down the problem to something I saw in a Pentium optimizing guide. That the C compiler inserted instructions to do _truncating_ conversion instead of running, which seriously messed up pipelining.
11:07:49 <solar_angel> oh
11:07:57 <solar_angel> okay
11:08:13 <solar_angel> bbls - my STC FORTH kept most of the stack in registers.
11:08:37 <solar_angel> so almost all the shuffling operations took zero cycles.
11:08:51 <solar_angel> swap, dup, etc all took zero cycles best case
11:08:55 <bbls> how?
11:09:00 <virl> my probably indirect forth keeps the complete stack in registers :)
11:09:23 <solar_angel> virl - unless you're using a register allocator, that's actually quite a slow way of doing things
11:09:30 <solar_angel> bbls - i used a register allocator.
11:09:49 <virl> a register allocator?
11:09:54 <bbls> how can allocate registers
11:10:01 <bbls> without
11:10:04 <bbls> inlining words?
11:10:21 <solar_angel> actually, i cheated.  the compiler was written in C, and was quite large.
11:10:25 <solar_angel> it generated some nice code though
11:10:28 <virl> hey, I program a forth for microcontroller which haves more than 50 registers ;)
11:10:38 <solar_angel> virl - ohhhh, well okay, that's a different story.
11:10:49 <solar_angel> microcontroller registers act very differently to most systems.
11:11:03 <solar_angel> usually you can even indirectly index uC registers.
11:11:22 <virl> uC?
11:11:27 <solar_angel> uC = microcontroller.
11:11:39 <solar_angel> which uC are you working with?
11:12:46 <virl> sx52 from ubicom, its a part of the xgamestation(www.xgamestation.com) and is really nice. 
11:13:10 <solar_angel> ahhh, an sx.  nice.
11:13:28 <solar_angel> pic instruction set, but way faster, am i right?
11:13:41 <solar_angel> and i never know what to make of the xgamestation...
11:14:57 <solar_angel> it looks cool, but very odd.
11:15:05 <virl> eh, for me it's the first real contact with assembler and because of this fact can you explaine me what a pic instruction set is?
11:15:50 <virl> PIC, programmable Interrupt Controller? 
11:16:01 <solar_angel> no...  the Microchip PICmicro.
11:16:06 <solar_angel> it's a type of microcontroller.
11:16:13 <solar_angel> the SX is a souped-up version.
11:16:19 <bbls> solar_angel let's suppose that you call consecutivelly two words, would you be able to optimize that?
11:16:40 <solar_angel> bbls - it supported inlining, so in many cases, yes.
11:16:53 <bbls> ok, but without inlining?
11:17:20 <solar_angel> why would you write an optimizing compiler that didn't support inlining?
11:17:35 <solar_angel> and between two out-of-line words, a drop would compile an instruction.
11:18:01 <solar_angel> it's zero instructions most times, but the register allocator emits instructions whenever there's a stack checkpoint
11:18:27 <bbls> if you have one inlined word followed by an uninlined one, could you optimize then?
11:19:10 <solar_angel> it really depends.  sometimes, yes.
11:19:12 <bbls> i suppose you could only inline after inline or inline after normal
11:19:18 <solar_angel> for instance...  "1 2 + DROP" compiles no instructions.
11:19:35 <bbls> ok, but 1 2 and + are all inlined
11:20:18 <virl> ah, thanks solar_angel, now I'm a little bit wiser than before.
11:22:01 <solar_angel> bbls - actually, the register allocator allocates a literal '1' on the stack, and a literal '2'
11:22:17 <solar_angel> then the built-in '+' operator sees two literals, adds them together, and puts the literal back on the stack
11:22:30 <solar_angel> all the while, the real (physical) stack has not been touched, and no actual registers have been allocated.
11:22:36 <solar_angel> the "drop" operator drops just the literal
11:22:39 <solar_angel> leading to no emitted instructions
11:23:28 <bbls> oh
11:23:48 <bbls> i tought you do optimizing after asm initial generation
11:24:57 <solar_angel> nope
11:25:01 <solar_angel> it generates assembly already optimized.
11:25:08 <solar_angel> it's not perfect... it's not GCC-grade optimization
11:25:15 <solar_angel> but it beats the pants off of almost any FORTH system :P
11:25:22 <bbls> :)
11:25:29 <solar_angel> oh, and another one...
11:25:39 <bbls> well, there is factor too, a forth derivative
11:25:49 <bbls> i know that i has very good optimizations
11:25:58 <solar_angel> "1 +" when it has a variable is in, say, "EAX", will emit "INC EAX"
11:26:14 <solar_angel> (as will 1+)
11:26:24 <solar_angel> and "2 +" will emit "ADD EAX, 2"
11:27:02 <bbls> i see
11:27:55 <solar_angel> yep
11:52:36 <solar_angel> hey, does anyone know where the best place to discuss the intricacies of assembly optimization and graphics programming on this network is?
11:53:56 <solar_angel> or is this the place? :P
11:54:04 <Robert> Hm, #asm and #assembler can be good, but sometimes they lack people who know what they're talking about.
11:54:23 <Robert> #osdev is usually better, but of course also a bit less on-topic.
11:54:36 <Robert> #forth would be one of the better places, yes.
11:54:58 --- quit: cmeme (Remote closed the connection)
11:56:11 <solar_angel> heh, i'm in both here and #osdev.
11:56:12 <solar_angel> so okay
11:56:48 <solar_angel> i like my spline algorithm... it's quite neat, assuming you want to generate a spline subdivided into a power of two lengths
11:57:03 <Robert> Actually, I think some people in #uuu at oftc are into that. But I haven't been there for a while, and it's even more off-topic.
11:57:05 --- join: cmeme (~cmeme@216.184.11.2) joined #forth
11:57:13 <solar_angel> ah
11:57:21 --- quit: cmeme (Remote closed the connection)
11:57:35 <solar_angel> i think i did pretty well, considering it uses, like, 3 adds and 2 shifts per step.
11:57:37 <solar_angel> and nothing more
11:57:39 <Robert> What's a spline algorithm doing, anyway?
11:57:39 <solar_angel> pure integer. :)
11:57:48 <solar_angel> well, you know what a cubic bezier spline is?
11:57:54 <Robert> No.
11:58:00 <solar_angel> (have you used Illustrator/Corel Draw/Sodipodi/etc?)
11:58:03 <Robert> I don't know what a spline is, for that matter.
11:58:05 <Robert> No.
11:58:05 --- join: cmeme (~cmeme@216.184.11.2) joined #forth
11:58:13 <solar_angel> oh
11:58:15 * solar_angel thinks
11:58:21 <Robert> ASCII graphics is about as far as I ever went.
11:58:25 <solar_angel> ah okay.
11:58:30 <Robert> Except with pen and paper.
11:58:44 <solar_angel> a spline, in graphics, is a curved line (there are also other uses, but i won't get into them)
11:58:57 <solar_angel> you take four points, two of which are endpoints, and two of which are control points
11:59:02 <solar_angel> and they define a curve
11:59:10 <solar_angel> you can glue them end to end to get curved objects
11:59:52 <bbls> why four?
11:59:55 <bbls> not a bspline
11:59:56 <bbls> much simpler
11:59:59 <bbls> just 3 points
12:00:08 <bbls> also it has nice continuations properties
12:00:19 <Robert> Ah, OK. Two shifts and three adds sounds pretty good. Including calculating the video buffer address, I assume?
12:00:23 <bbls> (you can split a 4 points line in 2 with 3 points)
12:00:23 <solar_angel> bbls - you're thinking a quadratic bezier spline
12:00:38 <solar_angel> Robert - no... that's just the spline calculation... i use a line drawing function to connect the points
12:00:52 <solar_angel> cubic splines are much more flexible
12:04:52 <solar_angel> no matter what the order, though, bezier splines aren't hard to compute
12:05:37 <Robert> What do you use it for?
12:05:50 <Robert> I mean, in the application you are working on.
12:06:32 <solar_angel> i pretty much just wrote it generically.
12:06:52 <solar_angel> i wrote a general-purpose one-dimensional cubic bezier spline computation system.
12:07:05 <solar_angel> then i glued two of them to a line drawing function, and drew some curves in my graphics library.
12:09:26 <bbls> i must go now
12:09:29 <bbls> bye everyone
12:09:31 --- quit: bbls ()
12:20:57 <virl> what graphics library solar_angel? is it public domain? =)
12:21:24 <virl> : what ." which" ;
12:21:56 <Robert> PD in for commies.
12:21:59 <Robert> is*
12:22:05 <solar_angel> the graphics library is called "Scarab", this version is codenamed "Scarab Delta"
12:22:16 <solar_angel> and it's not public domain, it's not open source, and it's not free.
12:26:18 <virl> every cool thing which was made with forth isn't accessible to poor mens like me. :-(
12:26:30 <solar_angel> my graphics lib isn't written in FORTH
12:26:34 <solar_angel> it's C and assembly.
12:26:38 <virl> oh...
12:29:12 <Robert> x86?
12:29:34 <solar_angel> the assembly is x86...  there are C versions of all the asm functions though
12:29:41 <solar_angel> so it should probably run on a different arch
12:29:48 <solar_angel> although there may be some endian issues
12:29:51 <Robert> Using any "modern" optimizations?
12:30:02 <solar_angel> someone wants to buy me a mac mini, i'll port it to mac :P
12:30:03 <Robert> Or instructions.
12:30:05 <solar_angel> modern?
12:30:12 <solar_angel> MMX and SSE
12:30:18 <Robert> As in post-386.
12:30:22 <Robert> That's modern.
12:30:42 <solar_angel> yep
12:30:51 <solar_angel> i have 3 versions of my alphablend line function
12:30:53 <Robert> I've been thinking about learning those. What reference do you use?
12:31:15 <solar_angel> intel's official guide
12:31:24 <solar_angel> they give them away as a set of pdf's now.
12:31:32 <solar_angel> the x86 reference guide
12:33:50 <solar_angel> anyway, my alphablend function, i have generic (parallelized regular instructions, does a few gorey things that are really unusual but still standard C)...
12:33:54 <solar_angel> i have the MMX
12:33:55 <solar_angel> and the SSE
12:34:39 <solar_angel> SSE added a couple MMX extensions that change the way one does that.
12:36:20 --- quit: AlexF (Read error: 131 (Connection reset by peer))
12:37:12 --- join: KB1FYR (~Alex@196-220.suscom-maine.net) joined #forth
12:39:57 <Robert> ge
12:40:51 <solar_angel> eh?
12:41:17 <Robert> (to KB1FYR)
12:41:35 <Robert> As in good evening, not Germanium.
12:42:09 <solar_angel> ah okay
12:44:42 <KB1FYR> hio
12:45:14 <KB1FYR> I knocked out a PCI card and it made me reboot... Who woulda thunk they weren't plu 'n play :P
12:47:06 <solar_angel> lol
12:47:10 <solar_angel> some of them are.
12:49:16 <KB1FYR> Not this one :P
12:49:50 <solar_angel> you need motherboard support.
12:50:22 <solar_angel> anyhow... i'm on my way.
12:50:27 <KB1FYR> bye
12:52:41 --- quit: solar_angel ("ALT-F4.  Wait, that didn't work...  Lee-Nucks, what's that?")
13:14:43 --- join: AlexF (~Alex@196-220.suscom-maine.net) joined #forth
13:14:43 --- quit: KB1FYR (Read error: 54 (Connection reset by peer))
14:22:40 <crc> hi all
14:23:18 <Robert> Hi.
14:55:31 --- join: Topaz (~top@sown-90.ecs.soton.ac.uk) joined #forth
15:08:37 --- join: I440r (~foo@209-234-161-250.gen.twtelecom.net) joined #forth
15:08:37 --- quit: I440r (Client Quit)
15:09:05 --- join: I440r (~foo@209-234-161-252.gen.twtelecom.net) joined #forth
15:42:26 --- quit: Frek ("Client exiting")
15:46:25 --- nick: nanstm -> Rastm2
15:46:53 --- join: Frek (~anvil@h254n2fls31o815.telia.com) joined #forth
15:48:11 --- quit: virl (Read error: 148 (No route to host))
15:52:09 --- quit: I440r ()
16:19:47 --- quit: Topaz (Remote closed the connection)
16:41:10 --- quit: Robert ("leaving")
17:00:11 --- join: skylan (~sjh@dialup-216-211-57-166.tbaytel.net) joined #forth
17:10:47 --- join: Sonarman (~cleetus@adsl-64-160-165-107.dsl.snfc21.pacbell.net) joined #forth
17:46:24 --- join: aum (~aum@60-234-138-239.bitstream.orcon.net.nz) joined #forth
17:46:51 --- part: aum left #forth
17:47:06 --- join: saon_ (~saon@c-66-177-224-130.hsd1.fl.comcast.net) joined #forth
17:47:21 --- nick: saon_ -> saon|smgl
17:56:20 --- quit: tathi ("leaving")
18:09:15 --- join: zol2 (~l@p549DD7F6.dip.t-dialin.net) joined #forth
18:09:31 <zol2> 'morning
18:24:05 --- part: zol2 left #forth
20:02:02 --- quit: Sonarman (Read error: 110 (Connection timed out))
21:54:14 --- quit: Frek ("Client exiting")
22:22:53 --- join: Frek (~anvil@h254n2fls31o815.telia.com) joined #forth
22:34:47 --- join: bbls (~bbls@80.97.121.4) joined #forth
22:35:01 <bbls> hello
22:39:47 <Herkamire> 'lo :)
22:39:57 <bbls> hi Herkamire
22:50:00 --- join: arke (apache@84.218.22.246) joined #forth
22:53:17 <arke> teh hies!!
22:53:39 <bbls> hi arke
22:57:20 <arke> :)
22:57:31 <arke> How's life?
22:57:59 <bbls> not too good :(
22:58:31 <arke> Oh, how come? :(
22:58:49 <bbls> maybe i'll have to go into army those days
22:58:59 <arke> :(
22:59:03 <arke> How old and what country?
22:59:08 <bbls> 22, romania
22:59:39 <arke> Oh. :(
23:00:04 <arke> Not fun.
23:00:12 <bbls> yeah :(
23:00:53 <arke> *sigh*
23:01:15 <bbls> beside this i'm also obese
23:01:24 <bbls> it will be like hell
23:02:04 <arke> Yeah, me too. I can't run 50 meters without breathing faster than a racecar. :/
23:02:16 <arke> I might have to do it. But, I hope not. 
23:02:24 <bbls> :)
23:02:26 <arke> I will definetely do everything possible to prevent it
23:02:30 <bbls> stupid romanian law
23:02:43 <bbls> there is a mandatory 1 year military service for all males
23:03:19 <arke> Germany has it too but you can substitute it with community service or something.
23:03:32 <arke> Except that it pays next to nothing. So, I don't know.
23:04:01 <bbls> well, here it pays next to nothing even if you go in the army :)
23:04:16 <bbls> stupid rules from the cold war
23:04:17 <arke> :)
23:04:29 <arke> Yeah :/
23:30:40 <Herkamire> wow, that sucks
23:30:43 <Herkamire> I wouldn't do it
23:30:52 <arke> :/
23:31:44 <bbls> you can't not to do it
23:31:48 <bbls> since it's mandatory
23:31:56 <bbls> if you refuse you go to prison
23:33:24 <arke> fun.
23:33:26 <arke> for how long?
23:33:40 <bbls> 1 year
23:33:52 <arke> i meant prison
23:34:00 <bbls> ah, many more years, not sure :)
23:35:14 <bbls> 3 years
23:35:28 <arke> ...and then you still have to do the army
23:35:29 <arke> right?
23:36:36 <bbls> i suppose yes :)
23:38:27 <Herkamire> sheesh
23:38:40 --- join: Topaz (~top@sown-90.ecs.soton.ac.uk) joined #forth
23:59:59 --- log: ended forth/05.05.17