Proposed method of LMM for PropBASIC. Comments requested.

Comments

• jazzed Posts: 11,803

  2010-02-05 18:04 edited 2010-02-05 18:04

  Hi Bean.
  
  Using an LMM macro jump vector table will simplify things and allow you to use
  the same kernel regardless of where the LMM macros are implemented.
  The downside is using longs for the table and slightly delayed execution.
  
  

  DAT org 0
  INIT    jmp #_init  ' 0
  MJMP    jmp #_jmp   ' 1
  MCALL   jmp #_call  ' 2
  ' ... other macros.
  
  _init
  ' init body
  
  _jmp
  ' jmp body
  
  _call
  ' call body
  
  
• Bill Henning Posts: 6,445

  2010-02-05 18:04 edited 2010-02-05 18:04

  I don't see any problems... and your CALL/RETURN mechanism is basically the same as for PASM, which I think will make it easier for beginners to understand.
  
  ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
  www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com 5.0" VGA LCD in stock!
  Morpheus dual Prop SBC w/ 512KB kit $119.95, Mem+2MB memory/IO kit $89.95, both kits $189.95 SerPlug $9.95
  Propteus and Proteus for Propeller prototyping 6.250MHz custom Crystals run Propellers at 100MHz
  Las - Large model assembler Largos - upcoming nano operating system
• Bean Posts: 8,129

  2010-02-05 18:21 edited 2010-02-05 18:21

  Am I correct that the "LONG @label - @__Init" will not cause any problems if executed since all the condition bits will be zero (since the value will always be less than 18 bits) ?
  
  Bean
  
  ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
  - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  Use BASIC on the Propeller with the speed of assembly language.
  
  PropBASIC thread http://forums.parallax.com/showthread.php?p=867134
• Bill Henning Posts: 6,445

  2010-02-05 18:26 edited 2010-02-05 18:26

  Yes, you are correct. I rely on that in my LMM code too.
  

  Bean said...
  Am I correct that the "LONG @label - @__Init" will not cause any problems if executed since all the condition bits will be zero (since the value will always be less than 18 bits) ?
  
  Bean

  ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
  www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com 5.0" VGA LCD in stock!
  Morpheus dual Prop SBC w/ 512KB kit $119.95, Mem+2MB memory/IO kit $89.95, both kits $189.95 SerPlug $9.95
  Propteus and Proteus for Propeller prototyping 6.250MHz custom Crystals run Propellers at 100MHz
  Las - Large model assembler Largos - upcoming nano operating system
• Dr_Acula Posts: 5,484

  2010-02-06 03:04 edited 2010-02-06 03:04

  Having seen on a board in front of me the awesome power of LMM with Catalina C (programs hundreds of kilobytes long), I'm really excited about similar code for Basic.
  
  A few months ago I was pretty chuffed when I compiled a C and a Basic program on the Zicog, and after compilation both returned a "40kb free" message. More than the memory on the propeller. Then RossH goes and blows that out of the water with ten times that. And now PropBasic is heading the same way. I am *very* impressed.
  
  The code looks great. Speed is not a big issue ('tis now as it ever was, use assembly for speed), but the ability to start writing a program and to keep writing and keep writing and never worry about running out of space, well that is just awesome.
  
  ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
  www.smarthome.viviti.com/propeller
• Bean Posts: 8,129

  2010-02-06 04:25 edited 2010-02-06 04:25

  @Dr_Acula,
  · PropBASIC LMM is not virtual memory. It runs from HUB ram, so programs are still limited to 32K. But that is a far cry from 2K. Also lots of room for variables with LMM.
  
  · Speed is about 1/5 of native code. I think that is a fair trade-off. Code with lots of calls and jumps will suffer most, but will still be·blazing compared to spin.
  
  · You will be able to specify LMM for the main code, and for each TASK. So some tasks can be native (for video drivers and stuff) and some can be LMM within the same program.
  
  · Actually for small TASKs it takes less space to make them native. I think very few TASKs will be LMM, but I wanted to have the option.
  
  · I ran into an issue using VAR arrays. I use MOVS and MOVD to do indexing, well that doesn't work in LMM. So I had to make two more LMM routines to handle it.
  · I'm going through the compiler and making the changes to handle LMM. I may post it for some testing, but NOT as the offical beta version.
  
  · Once you get your head around it, LMM isn't that hard. I hope the LMM code generated will provide assistance to PropBASIC users who want to do LMM in assembly.
  
  · Thanks for the encouragement.
  
  Bean.
  
  ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
  - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  Use BASIC on the Propeller with the speed of assembly language.
  
  PropBASIC thread http://forums.parallax.com/showthread.php?p=867134
  
  Post Edited (Bean) : 2/6/2010 4:32:32 AM GMT
• Dr_Acula Posts: 5,484

  2010-02-06 07:44 edited 2010-02-06 07:44

  32k would be a huge amount for propbasic, given how efficiently the compiler seems to be coding. As an aside, I've been coding all day on an SBasic program and only got to 10k.
  
  I'm looking at my code at the moment and thinking maybe it could be ported to propbasic?
  
  ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
  www.smarthome.viviti.com/propeller
• Toby Seckshund Posts: 2,027

  2010-02-06 08:32 edited 2010-02-06 08:32

  "to start writing a program and to keep writing and keep writing and never worry about running out of space, well that is just awesome."
  
  Thats not "awesome" its "Microsoft"
  
  ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
  Style and grace : Nil point
• Bean Posts: 8,129

  2010-02-07 16:50 edited 2010-02-07 16:50

  If anyone is interested in testing LMM in PropBASIC send me a PM.
  
  Bean.
  
  ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
  - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  Use BASIC on the Propeller with the speed of assembly language.
  
  PropBASIC thread http://forums.parallax.com/showthread.php?p=867134
• Christof Eb. Posts: 1,483

  2010-02-07 17:30 edited 2010-02-07 17:30

  Hi Bean,
  
  during my experiments with PropBasic I have seen, that you can do really very much without LMM and you have got full speed.
  
  So I am very glad to read, that there will be the choice for each task if this task will run in the cog only or with LMM!
  
  
  
  Is there a chance to include some sort of debugging feature into the LMM? It should be possible to insert debugging steps between program parts?
  
  Perhaps you could install some sort of a "hook" for a debugger routine,·that can be called after each basic line. This debugger routine could be done later.
  
  The debugger routine might copy a source code line number and all cog variables to hub ram. A debug task might then send the information to a serial line.
  
  The hook could perhaps be used for some sort of polled interrupts too?
  
  I think, that debugging features are often·very valuable, perhaps more valuable then execution speed.
  
  (I would like to have multi line comments too...)
  
  Christof
• Bean Posts: 8,129

  2010-02-11 19:18 edited 2010-02-11 19:18

  Hmmm, A debugger hook is an interesting idea. I'll have to explore that a little bit more.
  
  Just to give you some idea of speed:
  
  PASM instructions without jumps require 20 clocks (on average), 3 instructions require 16 clocks, then the·4th requires 32 clocks.
  
  Jumps require 32 clocks. Calls require 96 clocks.
  
  So code with alot of jumps and calls do take a real performace hit.
  
  Bean.
  
  ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
  - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  Use BASIC on the Propeller with the speed of assembly language.
  
  PropBASIC thread http://forums.parallax.com/showthread.php?p=867134
  
  March 2010 Nuts and Volts article·http://www.parallax.com/Portals/0/Downloads/docs/cols/nv/prop/col/nvp5.pdf
  
  
  Post Edited (Bean) : 2/11/2010 7:29:27 PM GMT