00:00:00 --- log: started retro/12.11.29
05:47:44 --- quit: harrison (Ping timeout: 256 seconds)
10:12:43 --- join: beretta (~beretta@cpe-107-8-120-84.columbus.res.rr.com) joined #retro
10:46:14 --- join: beretta_ (~beretta@cpe-107-8-120-84.columbus.res.rr.com) joined #retro
10:48:50 --- quit: beretta (Ping timeout: 252 seconds)
11:26:33 --- quit: karswell (Read error: Operation timed out)
11:32:42 --- join: karswell (~coat@93-97-29-243.zone5.bethere.co.uk) joined #retro
13:44:53 --- join: Mat2 (~claude@91-65-144-133-dynip.superkabel.de) joined #retro
13:44:58 <Mat2> hello
13:56:04 <beretta_> ello
13:56:10 --- nick: beretta_ -> beretta
13:56:17 <Mat2> hello beretta
14:00:14 <Mat2> which leads you here ?
14:00:49 <beretta> nothing much really.  Just poking around
14:01:30 <Mat2> do you find interest in programming ?
14:01:55 <beretta> yup.  I've made several FORTHS
14:03:59 <Mat2> nice :) where can I found the sources ?
14:06:05 <Mat2> I am currently working on my own system in retro fashion (hope that's the correct translation)
14:07:42 <beretta> I've only released one: http://sourceforge.net/projects/cocoboot/
14:09:22 <Mat2> hey, a forth system for tandy coco as it seems ! My favoured 6809 based system
14:10:01 <beretta> Yup!
14:10:57 <beretta> I'm revising the system for a VM version similar to Retro's VM.
14:13:07 <Mat2> hmm, you mean based on the current ngaro version I think
14:13:09 <beretta> I'm virtualizing the machine a touch more... so I can compile forth code on my x86 for the 6809.
14:13:25 <beretta> yes ngaro... I forgot it's name.
14:15:25 <Mat2> I see. Take a look at my vm design, I think it can give you some inspirations for performance optimizations at vm-code level
14:16:36 <tangentstorm> welcome beretta :)
14:17:22 <tangentstorm> Mat2: do you already have a ngaro -> navm instruction translator?
14:18:03 <Mat2> no, because the retro compiler need to be rewritten anyway for navm
14:18:33 <beretta> cocoboot is Token Threaded.  But I'm rewriting for and engine very similar to ngaro.
14:22:51 <beretta> newly defined words also got added to the token table... which quickly ran out of space (it was 8 bits)
14:27:00 <beretta> https://www.assembla.com/wiki/show/navm  is that yours?
14:27:14 <Mat2> I have 
14:27:23 <Mat2> sorry, yes
14:27:59 <tangentstorm> beretta: https://www.assembla.com/code/avm/git/nodes/master/metro/backport/avmInterpreterAtom64.pas
14:32:00 <tangentstorm> Mat2: what does the LI instruction do?
14:32:43 <Mat2> like LIT in ngaro, push an immediate value onto the data stack
14:32:56 <tangentstorm> LIT ah! duh. :)
14:33:35 <tangentstorm> is it possible to put TOS, DSP, etc in the actually assembly code instead of %rcx, etc?
14:34:14 <Mat2> not with the inline assembler in a portable way
14:34:41 <tangentstorm> hmm.. not even with macros?
14:34:45 * tangentstorm tries it
14:34:59 <Mat2> *not in a portable way*
14:35:21 <tangentstorm> why not?
14:35:30 <beretta> is there some sort of flow control ?  
14:37:04 <tangentstorm> beretta: BA and BEA are jumps, if that's what you mean
14:37:17 <Mat2> because freepascal reparses the assembler sources so these macros
14:37:21 <tangentstorm> Branch Always and Branch if Equal?
14:38:45 <Mat2> ok, execution is trace based. This means an opcode (for 64 bit 16 instruction in 8 slots) can select the execution of another trace (or the same) conditionally
14:39:23 <tangentstorm> trace?
14:40:06 <Mat2> yes, the trace can be compiled into subroutine threaded code for interpretation before execution or native-code (if a new instruction is generated)
14:40:33 <tangentstorm> but what's a trace?
14:41:08 <Mat2> a chunk of vm-code (in my vm a trace is identical with an opcode bundle)
14:42:20 <tangentstorm> so a trace is a 64-bit number containing 8 opcodes, or a mix of opcodes and literals / addresses
14:42:29 <Mat2> the BEA instruction for example select the compilaton of another opcode if TOS=SND, which is then executed
14:43:14 <Mat2> ^yes
14:43:49 <tangentstorm> Mat2: and by "not in a portable way" ... you mean macros aren't portable to other pascal compilers?
14:43:56 <Mat2> right
14:45:08 <Mat2> there exist some incompatibilities between gas and as for example (and freepascal will expand only gas conform macros)
14:46:36 <tangentstorm> Mat2: i mean freepascal's own macros, like this: https://github.com/sabren/b4/blob/master/ng_ops.pas#L9
14:47:29 <Mat2> is the freepascal preprocessor protable ?
14:48:09 <Mat2> sorry I mean portable
14:48:22 <tangentstorm> i'm not sure. i know gnu pascal has macros. probably delphi does, but i don't know.
14:49:38 <Mat2> probably I can compensate though inline procedures.. er I think that should not work
14:53:27 <Mat2> beretta: My vm design combine two 4 bit instructions into an bytecode. All combinations which are not useful - like dup+drop are replaced with higher order instructions (like branch conditionals). This way the interpretation overhead is halved independent of the threading method
14:54:12 <Mat2> and on the fly compilation is simplified
14:55:14 <Mat2> that works well even for 6502 cpu's so I think 6809 based interpreters would benefit also
14:56:29 <tangentstorm> Mat2: hang on. why does it need to be portable? wouldn't you have to rewrite it anyway if you want to target another architecture?
14:57:13 <tangentstorm> and as long as you're targeting x86, you have fpc, right?
14:57:20 <tangentstorm> so what does it need to be portable *to*?
14:57:43 <tangentstorm> amd64 i guess, not x86
14:58:24 <Mat2> I would need a good cross compiler
14:58:46 <tangentstorm> to what?
14:59:07 <Mat2> to ARM v6 for example
15:00:37 <tangentstorm> Mat2: okay, so freepascal appears to support ARMv6 as an output. http://wiki.freepascal.org/FPC_New_Features_2.6.0#ARM_systems
15:00:56 <Mat2> oh good to know, thanks
15:01:00 <tangentstorm> But, you'd need to change the actual assembly used for ARM, right?
15:01:14 <Mat2> so I need only an assembler for MIPS
15:02:14 <tangentstorm> http://wiki.freepascal.org/Native_MIPS_Systems
15:04:06 <Mat2> ok, I will try it out
15:04:42 <tangentstorm> Mat2: I'm completely clueless here... I'm just trying to understand what you're saying. :)
15:05:20 <tangentstorm> You want to run this on a bunch of different processors?
15:05:35 <Mat2> yes
15:06:02 <tangentstorm> Wouldn't you have to rewrite it for each processor, because they use different instruction sets?
15:06:23 <tangentstorm> I mean assembly language is specific to one machine, right?
15:07:10 <Mat2> yes, but I can reuse code for cpu's of the same architecture
15:07:16 <Mat2> like x86 and amd64
15:07:26 <tangentstorm> right
15:08:08 <tangentstorm> ... You know freepascal supports both, right?
15:08:40 <Mat2> I know, sadly the 64 bit gas version have some incompatibilities with the 32 bit gas one
15:08:56 <Mat2> (as I know)
15:09:07 <tangentstorm> okay, and you're using gas as the external assembler?
15:09:16 <Mat2> right
15:09:21 <tangentstorm> why?
15:09:50 <Mat2> because freepascal don't support the Intel syntax for amd64 (completely)
15:10:43 <tangentstorm> aha. what's broken?
15:11:32 <Mat2> there just work on finishing the preparser for 64 bit assembler code
15:11:47 <Mat2> so the only alternative would be an external procedure
15:11:55 <Mat2> this way I could use nasm
15:14:31 <tangentstorm> so gas32, gas64, and nasm all support intel syntax, but their macro systems are incompatible, right?
15:17:24 <Mat2> that's the case
15:18:16 <tangentstorm> Okay, so then the question is: Can we use fpc's macro system with assembly code that's being passed to an external assembler?
15:18:36 <tangentstorm> that's what you're asking me :)
15:18:42 <Mat2> right
15:18:55 <tangentstorm> see, i'm slow but I get there eventually :)
15:19:00 <Mat2> :)
15:20:11 * tangentstorm runs off to do an experiment.
15:27:53 <Mat2> beretta: Your token-threading approach reminds me of the one implemented in open firmware
15:30:32 <Mat2> Do you using TTC because of size restrictions ?
15:32:10 <Mat2> I'm just wondering because the Ngaro ISA is cell based
15:34:56 <tangentstorm> Mat2: what's the command to compile avmInterpreterAtom64.pas ?
15:40:26 <Mat2> fpc avmInterpreterAtom64.pas
15:42:01 <Mat2> you will get a note because .align is target specific
15:42:16 <Mat2> ^hint, hint: that's the problem
15:42:37 <tangentstorm> no i get like 2 pages of error messages :/
15:43:08 <tangentstorm> i tried:  fpc -Rintel -Aas avmInterpreterAtom64.pas
15:43:14 <tangentstorm> but it dies on the %
15:43:30 <Mat2> can it be that you try to compile it on an x86 and not amd64 system ?
15:43:37 <tangentstorm> ... :)
15:45:19 <Mat2> the other problem is the -Rintel switch, because the assembler source is AT&T syntax
15:47:20 <tangentstorm> I have an atom based netbook too, but mine isn't 64 bits. I was thinking they were the same. Thanks :)
15:47:46 <Mat2> that must be a very old netbook :)
15:59:18 <Mat2> you can create a virtual machine with qemu, that's what i use because qemu emulating x86 amd64, sarc, mips and arm systems
15:59:48 <Mat2> you don't need a real system for testing
16:00:24 <Mat2> (plus you can use its internal debugger)
16:04:06 <tangentstorm> no, it's fine. i have two 64-bit hex-core servers... used to run a web hosting company
16:21:08 <tangentstorm> Mat2: it doesn't look like macros work inside assembler blocks. :/
16:24:27 <Mat2> I feared it
16:26:38 <Mat2> by the way, can you please run some benchmark tests on your servers if I'am finshed the vm ? That would help me with further optimizations
16:37:19 <tangentstorm> sure, i'd be happy to.
16:42:19 --- nick: tangentstorm -> tangentafk
16:45:20 <Mat2> thanks
16:48:38 <Mat2> I will go to sleep
16:48:41 <Mat2> ciao
16:48:45 --- quit: Mat2 (Quit: Verlassend)
17:28:01 --- nick: tangentafk -> tangentstorm
21:05:34 <beretta> Mat2: TTC is for size restrictions.  I was asked to write a FORTH that would fit on the CoCo's standard boot sector: 4608 bytes.
21:51:42 --- quit: beretta (Read error: Connection reset by peer)
23:42:56 --- join: saper_ (saper@wikipedia/saper) joined #retro
23:44:05 --- quit: saper (Read error: Connection reset by peer)
23:59:59 --- log: ended retro/12.11.29