00:00:00 --- log: started retro/12.12.08
02:30:56 --- quit: tangentstorm (Ping timeout: 260 seconds)
02:40:28 --- join: tangentstorm (~michal@108-218-151-22.lightspeed.rcsntx.sbcglobal.net) joined #retro
06:22:55 <tangentstorm> input: BEGIN ! 1 + 2 END.
06:23:58 <tangentstorm> output: 1 2 + putn
06:26:22 <tangentstorm> yep yep yep. i do believe that is what they call a compiler. HOO RAH!
08:00:27 <crc> cool :)
10:09:42 <tangentstorm> crc: i was saying yesterday that retro has no word for a loop that runs 0 or more times...
10:10:02 <tangentstorm> like "while" is more like do...while in c
10:10:28 <tangentstorm> did i miss one?
10:11:36 <tangentstorm> it's easy enough to make one up, but what's the "retro way"?
11:01:21 --- join: Mat2 (~Claude@91-65-144-133-dynip.superkabel.de) joined #retro
11:01:27 <Mat2> hello
11:04:01 <Mat2> I'm happy my compiler basical works. It compiles AVM opcodes into native code at runtime
11:06:52 <Mat2> now I'm trying to reanimate GIT access to my repro
11:08:08 <Mat2> (that seem to be the hard part)
11:12:21 <tangentstorm> hey Mat2 :) awesome
11:12:28 <Mat2> hi tangentstorm
11:12:48 <Mat2> just read your compiler is working too :D
11:12:53 <tangentstorm> my compiler also works! I've got a basic arithmetic calculator working :)
11:12:56 <tangentstorm> yeah
11:13:15 <tangentstorm> actually have notes for all of pl/0 but i've been distracting myself all morning
11:13:32 <tangentstorm> so you have a JIT basically?
11:13:40 <Mat2> yes
11:14:42 <Mat2> but thanks to the AVM binary format without the need of typical memory and processor ressources for this task
11:15:11 <Mat2> the whole compiler fits into 16 kByte ram at current
11:17:38 <Mat2> was a surprise for me, because I had designed AVM original for efficient interpretation, not compilation
11:19:23 <tangentstorm> huh. i guess that makes sense. you were pretty close to assembly already
11:20:03 <Mat2> I've found the code generation benefit from instruction fusion a lot which simplifies everything
11:23:05 <Mat2> and lead to good code quality without the need for excessive optimization analysis because bundled instructions hold there dependancy state already
11:33:25 <Mat2> the opcode bundle (liDUP ldOVR liADD ldBR) for example compiles to:
11:33:30 <tangentstorm> Mat2: I've never really thought about how JIT works... is that virtual stuff you were asking about yesterday to allow you to be able to just create instructions in ram and jump to them on the fly?
11:34:39 <Mat2> basical yes
11:35:51 <Mat2> for executing generated native code, the dynamic allocated code buffer need to be marked for execution (the NX bit in Intel terminology of the allocated page entries need to be set)
11:37:54 <Mat2> then it is save to let compile the FPC compiler a call into this buffer like every other PROCEDURE or function
11:38:21 <Mat2> is callable though a pointer
11:39:58 <Mat2> sorry I mean, than it's save to call into the buffer exactly the same way as PROCEDURES and FUNCTION's are callable though a pointer in Freepascal
11:43:12 <Mat2> the opcode bunlde (liDUP ldOVR liADD ldBR) for example can be compiled into such a prepared buffer. Actually the compiler generate the following code:
11:43:53 <Mat2> MOV R8,  1000h  ; liDUP
11:44:08 <Mat2> MOV R9,  R8
11:44:27 <Mat2> MOV R9,  [R9]   ; ldOVR
11:44:38 <Mat2> MOV R10, R9
11:44:52 <Mat2> ADD R10, 2000h  ; liADD
11:45:04 <Mat2> MOV R10, [R10]
11:45:13 <Mat2> RET
11:45:30 <Mat2> at offset 0.
11:45:43 --- join: kumul (~kumul@66-50-75-241.prtc.net) joined #retro
11:48:46 <Mat2> to execute this code, you need only creating a call frame and let compile a call to it - et voilá: That's it
11:49:06 <Mat2> see you later
11:49:17 * Mat2 coding#
11:49:23 * Mat2 coding
11:49:39 --- nick: Mat2 -> Mat2-coding
12:13:15 <tangentstorm> sorry. i'm a big ball of distracted today.
13:16:25 --- nick: Mat2-coding -> mat2
13:16:34 --- nick: mat2 -> Mat2
13:16:47 <Mat2> no problem
14:53:54 --- nick: tangentstorm -> tangentgone
15:11:40 <Mat2> ciao
15:11:44 --- quit: Mat2 (Quit: Verlassend)
16:18:36 <crc> tangentgone: apart from repeat/again with a manual condition/break, not yet
17:14:57 --- quit: docl (*.net *.split)
17:16:53 --- join: docl (~docl@98.158.26.222) joined #retro
17:16:54 --- quit: docl (Changing host)
17:16:54 --- join: docl (~docl@unaffiliated/docl) joined #retro
17:55:03 --- join: arescorpio (~arescorpi@186.18.36.190) joined #retro
20:03:03 --- quit: arescorpio (Quit: Leaving.)
20:20:43 --- join: foucist (~foucist@ps14150.dreamhost.com) joined #retro
20:20:43 --- mode: ChanServ set +v foucist
20:21:01 <foucist> hello folks!
20:21:18 --- quit: tangentgone (Ping timeout: 265 seconds)
20:38:45 --- join: tangentgone (~michal@108-218-151-22.lightspeed.rcsntx.sbcglobal.net) joined #retro
23:53:35 --- quit: kumul (Quit: WeeChat 0.3.9.2)
23:59:59 --- log: ended retro/12.12.08