First, you have not answered Robin's question: Is this homework? If it isn't homework, where is this data coming from and how will the output be used? Without an answer to these questions, this thread will be closed.

Then, more questions based on your responses above:
1. No, you have not shown us any of your code (except for a sed command I gave you which seems to have no use in this thread, and a grep that gives you output you don't know how to process. I agree with Robin, we need to see the code you have written (even if it isn't working). And, we need to see a more representative sample of the input files this script is supposed to process. Searching for text matching a pattern in a file of arbitrarily formatted data is somewhere between hard and impossible. Searching for text in a text file where the format of all of the data in the file is well described is somewhere between easy and trivial. Make all of our lives easier by describing you input files fully.

Your first post said in file1 I have something like this

Code:

Min X = -250.000000, Max X =  250.000000
Min Y = -95.000000, Max Y =  95.000000
Min Z = -95.000000, Max Z =  136.500000

The following is something like that:

Code:

MinX = -250.000000,MaxX =  250.000000
Min y=-95.000000, Max Y =95.000000
miNz=-95.000000,maxZ=136.500000

but writing REs that recognize "something like" both of these is a much, much, more difficult task than writing an RE to match something that is "exactly" like the the first format. Are min values always negative? Are max values always non-negative? Are max values always positive?

If level0 is a "pre-defined integer value", why do you "have to enter some value"? Please explain.

2. What are you doing where getting 13 or more significant digits in the results from calculating a quotient from the sum of two input values divided by a third integer is insufficient? We can calculate arbitrary (but specified) precision results using bc or dc, but we can't print arbitrary (unspecified) precision output. (If xMin is -1, xMax is 1, and level0 is 3, your script will spend the rest of our lives printing an unending number if you insist on arbitrary precision output.) You need to clearly specify what ranges of values x, y, z, and level0 can take; and for some representative real values of those variables, show us the exact output that should be produced (not comments saying calculations should be performed) AND don't choose "nice" numbers that will produce integral results; show us real world values demonstrating the precision you need in your output. And, give a clear English description of how the precision is to be determined.

3. Maybe case doesn't matter to you, but we need to know if your script has to perform case insensitive searches for your variables in your input file and if it has to perform case translations to get the required output. We need a clear specification of what is in your input files and exactly what you want in your output file to help you write code that can find what you want in your input files and produce what you want in your output files.

4. This doesn't address the question I asked. Can the output be:

Code:

xMin -250.000000;
xMax 250.000000;
xg  result from calculating abs((250.000000-(-250.000000))/$level0);

yMin -95.000000;
yMax 95.000000;
yg  result from calculating abs((95.000000-(-95.000000))/$level0);

zMin -95.000000;
zMax 136.500000;
zg  result from calculating abs((136.500000-(-95.000000))/$level0)";

instead of the order you said you want. It will simplify your code if we can produce the x related results when we find the input line:

Code:

Min X = -250.000000, Max X =  250.000000

instead of having to save data to print at the end after finding all three significant input lines.

5. That didn't answer the question, but since you need more precision that C double-precision computation provide, it probably doesn't matter. Also, what OS are you using for this project?

And, just for fun, why is speed so important in calculating three values? Are there hundreds of input files to process? Is it the calculations that have to be fast? Or is it searching HUGE input files for your three important data lines that has to be fast? In most projects I've worked on, the speed of this part of a larger project would be insignificant unless we did some really gross stuff to slow it down unless this code will be in a loop processing lots of files.