Pi calculation

[Galileo]

I'm back.

Calculation of "some" pi digits.

Code: [Select]

// Cálculo de pi --- Pi calculation
// Translation from Phix code to Yabasic by Galileo, 09/2018.

n = 35001 : a = 10000 : dim f(n)

for i = 1 to n : f(i) = 2000 : next i
 
for c = n to 14 step -14
d = 0
for b = c to 1 step -1
d = d * b
g = b * 2 - 1
d = d + f(b) * a
f(b) = mod(d, g)
d = int(d / g)
k = e + int(d / a)
next
k2$ = "000000" + str$(k)
print right$(k2$, 4);
e = mod(d, a)
next

print "\nTime insumed: ", peek("millisrunning") / 1000, " seconds."
In my machine, delayed about 55 secs. Phix code about 21 secs.

[Tomaaz]

It looks familiar. I think that the Pix code is a translation of my code. 

[Galileo]

It looks familiar. I think that the Pix code is a translation of my code. 

Could you show the code? What dialect did you use? What's the performance?

By the way, the same code in Lua 5.3 takes about 65 seconds (on my machine). I didn't expect it.

[Tomaaz]

https://openeuphoria.org/forum/m/132983.wc

Euphoria and Lua.

Pirx == Tomaaz. I had problems with registering on that forum and tried several times. Finally they activated that Pirx account. Luajit less than 1 sec.

[Galileo]

But LuaJIT doesn't belong in the same league. Or am I wrong?

[Tomaaz]

But LuaJIT doesn't belong in the same league.

And why would that be?

[ZXDunny]

About the only reason I can think of is that LuaJIT executes compiled code (in the form of a dynarec) as opposed to a bytecode or pure interpreting engine.

[Tomaaz]

About the only reason I can think of is that LuaJIT executes compiled code (in the form of a dynarec) as opposed to a bytecode or pure interpreting engine.

Yes, but by accepting this point of view you agree that we can't compare almost anything. BaCon, Pure Basic, NaaLaa, Yabasic, Luajit - they all do things in a different way. You can't even compare 32-bit version of FreeBASIC to it's 64-bit equivalent, cause one translates to C and use C compiler, while the other don't (forgive me, if I'm wrong). And I think it's more interesting to compare different things than to compare similar things.

[ZXDunny]

To be honest, I've been thinking about using a Dynarec for SpecBAS. There's only so far a stack-based bytecode implementation will take you for speed concerns. But then, SpecBAS doesn't really need it as nobody is using it for CPU-intensive tasks (and that's not a job for a BASIC in any case, IMO), and as all the heavy lifting is done via my library of support routines, the language itself is just a method for calling that at the end of the day.

To be fair, I suppose if LuaJIT is BASIC-alike enough in its language implementation (and that's a very, very wide field these days - we have strayed so, so far from Dartmouth) then it can indeed be compared to others.

[Galileo]

Phix-to-Yabasic conversion again. Little faster.

Code: [Select]

t1 = peek("millisrunning")

chki = 1
a = 10000
c = 35000 // 8400
fr = int(a / 5)
dim f(c)

for i = 1 to c : f(i) = fr : next

while(c > 0)
    g = 2 * c
    d = 0
    b = c
    while(b > 0)
        d = d + f(b) * a
        g = g - 1
        f(b) = mod(d, g)
        d = int(d / g)
        g = g - 1
        b = b - 1
        if not(b = 0) d = d * b
    wend
    c = c - 14
    n4$ = str$(e + int(d / a),"%04g")
    print n4$;
    e = mod(d, a)
wend

print "\nTime elapsed: ", (peek("millisrunning") - t1) / 1000, " seconds"