Lua interpreter with arbitrary precision math.

[Thorham]

Here's a build of the Lua interpreter from Lua 5.2.3: lua.lha

Works with Kicstart 3.0+ and 68020+ (only tested with 68030). Compiled with SASC6.58, so might work with Kicstart 2+. Does NOT require an FPU, ixemul.library, etc. Only needs standard Workbench libraries. This build of Lua doesn't support loading dynamic libraries.

To use, simply unpack the archive anywhere you want, and start from Workbench. You can also run lua.exe from the CLI directly. When starting from the CLI, use lua.exe -l mapm to start the interpreter with the math library.

Example usage (type into the interpreter command line directly):

mapm.digits(100)
a=mapm.sqrt(12345)
b=mapm.sqrt(54321)
print(a*b)
c=mapm.number"1"
d=mapm.number"3"
e=c/d
print(e*d)

You can also use things like mapm.sin, mapm.cos, etc.

Currently there's no documentation, but you can get some pointers here:

mapm: http://www.tc.umn.edu/~ringx004/mapm-4.9.5a.tar.gz
lmapm: http://webserver2.tecgraf.puc-rio.br...2/lmapm.tar.gz

Mapm is the arbitrary precision floating point library. This contains documentation about the functions. Lmapm is the code that links mapm to lua. The file lmapm.c will show you which functions are available in the interpreter. Sorry, but it will have to do for now.

[Thorham]

No one's interested in Lua here?

[tolkien]

I like LUA. I have the book Programming in lua. Like the concept but I dont use it.
I hope your work can be used by the talented coders we have around.

[Lonewolf10]

No offense, but I have never heard of Lua until now (not surprising considering there's over 500+ programming languages - consult the official 'Hello World' collection).

I will probably give it a try sometime over the next few days - I'll need a break from trying to get the 4096 tile in Coagulus' game and from drawing new GFX for my own game.

[Thorham]

Originally Posted by tolkien

I hope your work can be used by the talented coders we have around.

The intention is only to use it as a programmable calculator, nothing fancy.

Originally Posted by Lonewolf10

there's over 500+ programming languages

[Leffmann]

I prefer Python, but Lua is probably slimmer and faster to run directly on the Amiga.

I've also built the Lua interpreter and compiler with SAS/C, but they give run-time errors for certain programs. My guess is that SAS/C isn't implementing the standard C library correctly, which Lua is entirely dependent on.

[Thorham]

Originally Posted by Leffmann

I prefer Python, but Lua is probably slimmer and faster to run directly on the Amiga.

Probably a lot easier to build as well.

Originally Posted by Leffmann

I've also built the Lua interpreter and compiler with SAS/C, but they give run-time errors for certain programs. My guess is that SAS/C isn't implementing the standard C library correctly, which Lua is entirely dependent on.

I had that problem with things like print(0x1234567a). Just hangs. Fixed it by playing with the compiler options (I'll upload the fixed version later). SASC also shows lots of warnings when compiling Lua. Tried to do it with GCC (no warnings), but I can't seem to link the Lua interpreter with the mapm library properly without ixemul (both work fine without ixemul, combine them, and mapm doesn't work properly). Very annoying (probably something small as I'm a compiler n00b).

[jPV]

I like Lua too. I did learn it recently, because there's a native implementation of Lua included in MorphOS by default, and it can be used to replace ARexx. I converted some of my ARexx scripts to Lua and also wrote some new ones. You can find them here if interested.

Lua is still missing something compared to ARexx, but OTOH it's more powerful in many other areas. Nevertheless I enjoy making scripts with both languages.

[alkis]

I've compiled lua with gcc and then run this

-- sieve.lua
-- the sieve of Eratosthenes programmed with coroutines
-- typical usage: lua -e N=500 sieve.lua | column

-- generate all the numbers from 2 to n
function gen (n)
  return coroutine.wrap(function ()
    for i=2,n do coroutine.yield(i) end
  end)
end

-- filter the numbers generated by `g', removing multiples of `p'
function filter (p, g)
  return coroutine.wrap(function ()
    for n in g do
      if n%p ~= 0 then coroutine.yield(n) end
    end
  end)
end

local start = os.clock()
c=0
N=N or 500		-- from command line
x = gen(N)		-- generate primes up to N
while 1 do
  local n = x()		-- pick a number until done
  if n == nil then break end
  --print(n)		-- must be a prime number
  c=c+1
  x = filter(n, x)	-- now remove its multiples
end
print(c)
print(string.format("elapsed time: %.2f\n", os.clock() - start))

I've run it with the gcc compiled exe and with the Sas compiled exe.
Here are the results:

3.Work:t/amath> lua.exe sieve.lua 
95
elapsed time: 4.34

3.Work:t/amath> lua.exe sieve.lua 
95
elapsed time: 4.33

3.Work:t/amath> luagcc sieve.lua 
95
elapsed time: 3.26

3.Work:t/amath> luagcc sieve.lua 
95
elapsed time: 3.26

Gcc (3.4.0), looks like, produced code almost 25% faster than SAS C!
Didn't expect that much.

[Thorham]

Originally Posted by alkis

[/CODE]Gcc (3.4.0), looks like, produced code almost 25% faster than SAS C!
Didn't expect that much.

I get 2.15 on my A1200 with Blizzard 1230 IV using my own SASC compiled version. Compiled for 68020 with some options to fix a bug, and without optimizations.

[alkis]

Originally Posted by Thorham

I get 2.15 on my A1200 with Blizzard 1230 IV using my own SASC compiled version. Compiled for 68020 with some options to fix a bug, and without optimizations.

I've put luagcc in the zone. If you could measure it please, I am curious!

[Thorham]

Originally Posted by alkis

I've put luagcc in the zone. If you could measure it please, I am curious!

It crashes when I run the test script, but it somehow still completed the script the first time I tried, and I think it was 1.90 (crashed quickly afterwards).

[alkis]

Weird. I don't get crashes (emulated A1200). Maybe it overuses the stack. I think I was playing with 500k stack (was going for 50k but I typed one zero too many).

[Thorham]

Originally Posted by alkis

Weird. I don't get crashes (emulated A1200). Maybe it overuses the stack. I think I was playing with 500k stack (was going for 50k but I typed one zero too many).

Yeah, tried with stack 1000000, and it works fine. Result for the test is 1.81.

[alkis]

ok, gcc6 is GODLIKE

3.Work:t/amath> lua.exe sieve.lua
95
elapsed time: 5.31

3.Work:t/amath> luagcc sieve.lua
95
elapsed time: 4.64

3.Work:t/amath> /lua sieve.lua [gcc6 68000 version]
95
elapsed time: 3.48

3.Work:t/amath> /lua sieve.lua [gcc6 68020 version]
95
elapsed time: 3.34

3.Work:t/amath> /lua /for.lua [gcc6 68020 version]
5000050000
11.7 seconds
3.Work:t/amath> lua.exe /for.lua
5000050000
26.632 seconds
3.Work:t/amath> luagcc /for.lua 
5000050000
24.58 seconds

!!!! 11.7 down from gcc3.4.0's 24.58, down from 26.6 of SAS C !!!

Code for for.lua

s = os.clock()
sum = 0
for i=1,100000 do
        sum = sum + i
end
print(sum)
print(tostring(os.clock()-s) .. " seconds")

lua for 68000 and lua020 for 68020 in the linked zip

https://drive.google.com/open?id=1dg...SxkL-YLLzDbXQA

[hooverphonique]

Originally Posted by Thorham

Yeah, tried with stack 1000000, and it works fine. Result for the test is 1.81.

1MB of stack to run an amiga program!? That's nuts!

[meynaf]

Huge stacks are common for programs coming from other platforms (understand : pc).
I've found that usually 300k is enough, though.

[alkis]

Originally Posted by hooverphonique

1MB of stack to run an amiga program!? That's nuts!

It's an interpreter You can run all sort of things, with various levels of recursion and such.

For example the last version I linked runs happily the "for.lua" with 4k stack.
"sieve.lua" crashed with 50k stack, worked with 100k.

[Krashan]

Originally Posted by Thorham

This build of Lua doesn't support loading dynamic libraries.

Lua can be made to support dynamic libraries in a form of system libraries. I've done it in MorphOS port and probably backporting it to classic Amiga is straightforward.

[MartinW]

Apologies for resurrecting an old thread but any idea what hardware the above tests were performed on?

For example " 11.7 down from gcc3.4.0's 24.58, down from 26.6 of SAS C "

I'm just trying to guage my results with the lua i just built.