I have 54 data files, each of which is several GB in size.

Each file contains 15 columns of (x,y,z) values.

My task is to read those files, compute autocorrelation, and plot graphs.

How should I read the files?

1. Open each file
2. Read 15 columns
3. Prepare plot
4. Write the plot to a Bitmap file

Or,

1. Open each file
2. Read one column
3. Prepare one plot
4. Accumulate plot
5. go to step #1
6. do this 15 times
7. Finally, write the plot to a bitmap file

Which one will be faster?

Which one will be memory-efficient?