OpenEuphoria

2. Re: Pi Benchmark Game

• Posted by Pirx Aug 23, 2018
• 1690 views

Sorry, but are you sure you're using Luajit and not the standard Lua interpreter?

4. Re: Pi Benchmark Game

• Posted by _tom (admin) Aug 23, 2018
• 1693 views

I do not have any Lua installation, only LuaJIT.

For the 32bit atom, Bodhi Linux:

$ luajit 
LuaJIT 2.0.5 -- Copyright (C) 2005-2017 Mike Pall. http://luajit.org/ 
JIT: ON CMOV SSE2 SSE3 ATOM fold cse dce fwd dse narrow loop abc sink fuse 
>  

For the 64bit i5, Mint Linux:

$ luajit 
LuaJIT 2.0.5 --Copyright (C) 2005-2017 Mike Pall. http://luajit.org/ 
JIT: ON CMOV SSE2 SSE3 SSE4.1 fold cse dce fwd dse narrow loop abc sink fuse 
> 

Where do I look next?

_tom

6. Re: Pi Benchmark Game

• Posted by Pirx Aug 23, 2018
• 1679 views

This is really strange. I don't understand why on my machine LuaJIT is 10x faster than compiled Euphoria and 20x faster than interpreted Euphoria, while on yours it's the opposite. I hope that someone knows why and will explain it to me.

8. Re: Pi Benchmark Game

• Posted by Pirx Aug 23, 2018
• 1744 views

OK. Please, run this code. It calculates 10000 digits of PI.

Euphoria:

include std/math.e 
include std/text.e 
 
atom n = 35001, a = 10000, d, g, k, e = 0 
sequence f = repeat(2000, n) 
 
for c = n to 14 by -14 do 
	d = 0 
	for b = c to 1 by -1 do 
		d = d * b 
		g = b * 2 - 1 
		d = d + f[b] * a 
		f[b] = mod(d, g) 
		d = floor(d / g) 
		k = e + floor(d / a) 
	end for 
	sequence k2 = "000000" & sprint(k) 
	puts(1, k2[(length(k2) - 3)..(length(k2))]) 
	e = mod(d, a) 
end for 

Lua:

n, a, d, g, k, e = 35001, 10000, 0, 0, 0, 0  
f = {} 
for x = 1, n do 
	f[x] = 2000 
end 
 
for c = n, 14, -14 do 
	d = 0 
	for b = c, 1, -1 do 
		d = d * b 
		g = b * 2 - 1 
		d = d + f[b] * a 
		f[b] = d % g 
		d = math.floor(d / g) 
		k = e + math.floor(d / a) 
	end 
	k2 = "000000" .. tostring(k) 
	io.write(string.sub(k2, string.len(k2) - 3)) 
	e = d % a 
end 

As you can see, it's the same algorithm. On my machine Luajit takes 0m0.941s (less than one second) to calculate and print 10000 digits of PI, while interpreted Euphoria takes 24 sec.

10. Re: Pi Benchmark Game

• Posted by Pirx Aug 23, 2018
• 1634 views

OK. These numbers make sense.

12. Re: Pi Benchmark Game

• Posted by Pirx Aug 23, 2018
• 1600 views

And what's interesting, in your example compiled Euphoria is just 2x faster than the interpreted one. This is max. I can get on my laptop. At last I think we are getting somewhere with these numbers.

14. Re: Pi Benchmark Game

• Posted by Spock Aug 23, 2018
• 1601 views

Irv,

Changing the type to 'integer' and then compiling might also improve EUC. On my system it improved EUI from 8s to 7.5s

Spock