🔒 Locked

How could I optimize this code?

Started by darkforest Jun 1, 2006 at 2:50 PM 12 replies 1.8k views

Original Post

Jun 1, 2006 at 2:50 PM


void Byteswapping(unsigned char* aDest, unsigned char* aSource, int size)
{
  int i, j;
  int iTimesSize;

  for (i=0; i<size; i++)
  {
    iTimesSize = i * size;
    for (j = 0; j<size; j++)
    {
      aDest[iTimesSize+j]=aSource[j*size+i];
    }
  }
}

Is there a easy way to optimize this?

darkforest

Jun 1, 2006 at 2:54 PM

void Byteswapping(unsigned char* aDest, unsigned char* aSource, int size){  aDest = aSource;}

Whould this mean the same?

DrEvil

Jun 1, 2006 at 2:54 PM

Looks relatively optimized to me, short of doing assembly level stuff. Possibly use the restrict keyword on the char * parameters.

rip-off

Jun 1, 2006 at 3:01 PM

Quote:
Original post by darkforest
void Byteswapping(unsigned char* aDest, unsigned char* aSource, int size){  aDest = aSource;}
Whould this mean the same?

Nope.

It only changes the local value of aDest.

Perhaps memcpy, or std::swap if you are using c++?

Fruny

Jun 1, 2006 at 3:04 PM

Quote:
Original post by darkforest
Is there a easy way to optimize this?

Not that I can see. Your compiler probably already does strength reduction and loop unrolling for you.

dbzprogrammer

Jun 1, 2006 at 3:17 PM

What is the goal of your function? I'm messing around with it, and it's not really copying a C string correctly... Are you trying to flip bytes in memory, so that a C string would be spelled backwards?

dbzprogrammer

Jun 1, 2006 at 3:25 PM

He'd have to flip his code around, it'd work more like this then:

void Byteswapping(unsigned char* aDest, unsigned char* aSource, int size){  int i, j;  int iTimesSize;  i = 0;  while(i<size)  {    iTimesSize = i * size;    j = 0;    while(j<size)    {      aDest[iTimesSize+j]=aSource[j*size+i];      ++j;    }  ++i;  }}

Besides, I have a feeling that leaving it as a for loop would be faster due to automatic compiler optimizations.

BlueHabu

Jun 1, 2006 at 3:31 PM

MMX instuctions will copy things really fast if you have that option

Quick google result
http://webster.cs.ucr.edu/AoA/Windows/HTML/TheMMXInstructionSeta2.html

dbzprogrammer

Jun 1, 2006 at 3:33 PM

Quote:
Original post by Fruny
Quote:
Original post by darkforest
Is there a easy way to optimize this?

Not that I can see. Your compiler probably already does strength reduction and loop unrolling for you.

I don't quite get the concept of Strength reduciton, could you explain it?

JohnBolton

Jun 1, 2006 at 4:06 PM

What is your function supposed to do? It looks like it "transposes" a size by size array of bytes.

If size is a multiple of 4, then you could speed it up by reading or writing 4 bytes at a time.

BTW, the iTimesSize variable is unnecessary. Any decent compiler will do that for you.

Scourage

Jun 1, 2006 at 4:26 PM

Quote:
Original post by Anonymous Poster
erm third time lucky

Normally pre increment is faster than post increment.(machine dependent optimization)

Generally true, but most importantly on expense items to create (such as iterators ). On most architectures, simple types such as ints, produce similar code for post and pre increment.

Cheers,

Bob

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.