Math Calculation

Charles Law

Does anyone has idea on how to calculate log number in AMX program? I think we don't have any log function to do this.

Thanks
Charles

Spire_Jeff

Ok, The mathematic nature of your question struck a nostalgic nerve in my body and drove me to understand how much math I have forgotten over the years What exactly are you trying to accomplish with the LOG function? Are you simply looking for BASE 10 Logs or do you need Natural Logs or even BASE 2 Logs? I have the beginings to a LOG function, but after much searching, it seems the most efficient way of calculating LOGS is to use a lookup table. This only works if one is using a set BASE. I did find some lookup tables and I am contemplating completing the function, but I hesitate because I don't really see myself ever using such a function.

Another answer could be: Check the Java language in DUET. I would imagine that JAVA has a LOG and a LN function built in.

Jeff

champ

There's no function I know of, but I like a challenge.

if y = a^x
then x = loga*y (log y to the base of a)

so you would have to do it something like long division

I'll have a crack at it tonight.

Hedberg

Natural log function attached

You can calculate the natural log from a lookup table of powers of e and the fact that:

ln(x+1) = x - x^2/2 + x^3/3 -x^4/4 . . .

I've created such a function that seems to work decently with a 16 bit integer input. The function could be modifed to work with a float input but the lookup table would need to be expanded to about 90 elements to accomodate values up to E39.

Once the natural log is calculated, it can be converted to any base using the relationship:

logb(x) = ln(x)/ln(b)

for common logs:

log(x) = ln(x)/ln(10) = ln(x)/2.302585

for base 2 logs:

log2(x) = ln(x)/ ln(2) = ln(x)/.693147

champ

Nice work Hedberg

So log10(100) = 2

NatLog(100)/NatLog(10) = 2

I gave up when I needed an exponential function that could raise a number to the power of a float. It's easy to raise a number to the power of an integer, but not a float.

Baskakov

I've implemeneted exp(x) function where x is a floating point number, and I suppose that series of function a^x also exists.

Hedberg

from the series expansion of e^x, you can calculate the series expansion of a^x.

I didn't do that; I looked it up:

a^x = 1+ x*ln(a) + (x*ln(a))^2/2! + (x*ln(a))^3/3! + ...

Thinking about these questions has made me wonder how calulators and computer functions actually do calculations involving logarithms. I don't know the answer to that question and a brief excursion down the information super-highway didn't reveal it. There must be some trick of which I am unaware.

Harold

Yarm

LOG function

I did this, it works quite well

DEFINE_FUNCTION float log (float base,float value)
{
local_var float epsilon
local_var sinteger integer_value
local_var float decfrac
local_var float partial

//This Log routine is accurate to 4 places
//
//Found at http://www.answers.com/topic/logarithm
//Converted to AMX by William Holt, Intelligent Homes (In Control) Ltd.
//
//Usually base is e
//e is approx 2.718282
//but for use with linear volume to db conversion, base is 10

if (value <> 0)
{
epsilon = 0.000000000001
integer_value = 0

While (value < 1)
{
integer_value = integer_value - 1
value = value * base
}

While (value >= base)
{
integer_value = integer_value + 1
value = value / base
}

decfrac = 0
partial = 0.5
value = value * value

While (partial > epsilon)
{
While (value >= base)
{
decfrac = decfrac + partial
value = value / base
}
partial = partial / 2
value = value * value
}

decfrac = decfrac + integer_value

RETURN decfrac
}
else
{
RETURN 0
}
}

Can anyone post a function to do x^y or powf(x,y) or an exp function? please.