Description shmuel siegel 2002-01-23 10:23:16 EST

The virtual machine on many embedded devices does not do runtime optimation. It 
would therefore be nice if the compiler performed certain optimizations which 
although they could have been done by the programmer they would interfere with 
the understanding of the code. In the following class (initialization and error 
checking code has been removed)fastCRC runs twice as fast as slowCRC on a 
Motorola i85s phone, but no programmer wants to write code like fastCRC. 
Eclipse's compiler should have an option to perform this kind of optimization.

public class CRC32
{
/**
* Table of CRC-32's of all single-byte values
*/

static int crc_table[];
public static int crc;

    static public int slowCRC(byte[] buf, int offset, int len)
    {
        for(;len>0; offset++,len--)
        {
            crc = crc_table[(crc ^ buf[offset]) & 0xff] ^ ((crc >> 8)&0xffffff);
        }
        crc = crc ^ 0xFFFFFFFF;

        return crc;
    }

    static public int fastCRC(byte[] buf, int offset, int len)
    {
        int lcrc = crc ^ 0xFFFFFFFF;
        int[] ltable = crc_table;
        for(;len>0; offset++,len--)
        {
            lcrc = ltable[(lcrc ^ buf[offset]) & 0xff] ^ ((lcrc >> 8)&0xffffff);
        }
        crc = lcrc ^ 0xFFFFFFFF;

        return crc;
    }

}