Thread: Optimise code (noob questions)

Hello, this is my first post

I've done a simple file encryption program, but it takes too much cpu power and i will appreciate any advice you can give me to optimise my code. This is the part that needs to be improved:

Code:

while(1){ 

            err = fread( &byte_buff  , sizeof(unsigned char) , 1 , input_file );

            if( err != 1 ){                                    
                if( feof( input_file ) ){                   
                    printf("\n\n Operation successful!\n");   
                    break;
                }                                            
                else{                                       
                    fprintf(stderr, "\n Error occured while reading from file!!!\n");
                    break;
                }
            }


            //-- Encrypt / Decrypt the byte with Bitewise NOT
            byte_buff = ~ byte_buff;

            err = fwrite( &byte_buff  , sizeof(unsigned char) , 1 , output_file);

            if( err != 1 ){                                 
                fprintf(stderr, "\n Error occured while writing!!!\n");
                break;
            }

            if( (i % ( file_size / 20 ) == 0) && (i!=0) ){
                printf("\b\b%d", percentage += 5); fflush(stdout);
            }
            i++;
            
        }

Also i have a question about fread() and fwrite(). Is there a better way than read/write one byte at a time so i can boost the speed ? Because if i read more than 1 byte, lets say 2 but the file is 3 bytes, i'm not sure if fread() will read the last byte and when reach eof stop, or it will give me the last byte + 1 extra junk byte
English is not my native language, so excuse me for my mistakes

I said it was simple , and the cpu usage is 25% which i think is too much for such program, thats my i wanted to know if i can optimise it

^_^

I don't know about the rest of you, but my money is on the original poster having a 4-core CPU and a "Windows" operating system.

Soma

^^^^ You're so cruel! I love it. ^^^^

Dimebag, have you considered reading from the file, a block of chars at one time, and then encrypting/decrypting all of that block at once, before writing it to a file. Your logic feels like a roller coaster - I'm down reading a teeny little bit from a file, then I'm up en/de crypting, then I'm back down again writing a teeny little bit of data, to a file.

I believe your intuition is right - go big with this, and ditch the read/do a tiny something/write, loop you have now.

A profiler would be nice, here.

You need SOME measurement of performance as it is now, so any changes can be compared on the same test data.

The cpu time is being consumed by all the calls to fread() and fwrite() transferring 1 byte at a time. Try reading 65536 bytes to 1048576 bytes at a time. (You'll need to allocate a buffer that is the same size).

Originally Posted by Salem

All that attempting to load a large file into memory in one go will achieve would be the initial transfer of the entire file into memory, and then immediately to swap space, and then another transfer out of swap space back into memory.

Unless the file size is 1GB or more, it seems unlikely that the file data in the programs memory would immediately end up in the swap space.

Originally Posted by Salem

Standard C already declares BUFSIZ in stdio.h.

For a PC (at least with visual studio), BUFSIZ is defined to be 512 bytes, the size of a sector. If the actual file I/O (reads and writes) are done 512 bytes at a time, that's fairly inefficient. Most PC's have 4GB or more of ram, and even with a 32 bit system, a program can usually allocate 1GB of ram without issues.

@Salem: Good point!

Actually the OS and even the hd will take care of buffering, caching, etc (I think the library is free to implement fread()/fwrite() anyway it likes, as long as it is buffered) .

However, if most of the files are reasonable sized then swapping is much less probable to happen anyway. Also, the less calls to fread()/fwrite() the less probable I/O blocking overhead will take place.

All in all, a profiler (along with enough test cases) should be used for safe conclusions, before applying either approach. Also, setvbuf() can also be used for optimization experiments, etc. Knowing before hand the size of the file to be handled, is often a plus for an educated decision.

There's no easy way to vote for one or the other, without testing first on the specific platform, for the specific dataset and for the specific application.