crctest.c

// ok --> "gcc crctest.c -o testapp -std=c11 -march=armv8-a+crc -mfpu=crypto-neon-fp-armv8" 
// fails --> "gcc crctest.c -o testapp -std=c11 -mcpu=cortex-a53 -mfpu=crypto-neon-fp-armv8" 

#include <stdio.h>
#include <math.h>
#include <arm_acle.h>
#include <time.h>

unsigned int CRCTable[ 256 ];
typedef unsigned long long u64;
typedef signed long long s64;
typedef unsigned int u32;
typedef unsigned int uint32_t;
typedef unsigned short u16;
typedef unsigned short uint16_t;
typedef unsigned char u8;
typedef unsigned char uint8_t;

#define CRC32_POLYNOMIAL     0x04C11DB7

unsigned int CRCTable[ 256 ];

u32 Reflect( u32 ref, char ch )
{
	 u32 value = 0;

	 // Swap bit 0 for bit 7
	 // bit 1 for bit 6, etc.
	 for (int i = 1; i < (ch + 1); ++i) {
		  if(ref & 1)
			value |= 1 << (ch - i);
		  ref >>= 1;
	 }
	 return value;
}

void CRC_BuildTable()
{
	u32 crc;

	for (int i = 0; i < 256; ++i) {
		crc = Reflect( i, 8 ) << 24;
		for (int j = 0; j < 8; ++j)
			crc = (crc << 1) ^ (crc & (1 << 31) ? CRC32_POLYNOMIAL : 0);

		CRCTable[i] = Reflect( crc, 32 );
	}
}

u32 CRC_Calculate( u32 crc, const void * buffer, u32 count )
{
	u8 *p;
	u32 orig = crc;

	p = (u8*) buffer;
	while (count--)
		crc = (crc >> 8) ^ CRCTable[(crc & 0xFF) ^ *p++];

	return crc ^ orig;
}

u32 CRC_CalculateARM( u32 crc, const void * buffer, u32 count )
{
        u8 *p;
        u32 orig = crc;

        p = (u8*) buffer;
        while (count >= 8) {
		crc = __crc32d(crc, *((u64*)p));//(uint64_t)(p[4] | p[5] << 8 | p[6] << 16 | p[7] << 24) << 32 | (p[0]  | p[1] << 8 | p[2] << 16 | p[3] << 24) );
		p += 8;
                count -=8;
	}
        if (count >= 4) {
                crc = __crc32w(crc, *((u32*)p));//(p[0]  | p[1] << 8) );
                p += 4;
                count -= 4;
        }
	if (count >= 2) {
		crc = __crc32h(crc, *((u16*)p));//(p[0]  | p[1] << 8) );
		p += 2;
                count -=2;
	}
	if (count == 1)
		crc = __crc32b(crc, *p);

        return crc ^ orig;
}

u32 CRC_CalculatePaletteARM(u32 crc, const void * buffer, u32 count )
{
	u8 *p;
	u32 orig = crc;

	p = (u8*) buffer;
	while (count--) {
		crc = __crc32h(crc, *((u32*)p));
		p += 8;
	}

	return crc ^ orig;
}

u32 CRC_CalculatePalette(u32 crc, const void * buffer, u32 count )
{
        u8 *p;
        u32 orig = crc;

        p = (u8*) buffer;
        while (count--) {
                crc = (crc >> 8) ^ CRCTable[(crc & 0xFF) ^ *p++];
                crc = (crc >> 8) ^ CRCTable[(crc & 0xFF) ^ *p++];

                p += 6;
        }

        return crc ^ orig;
}

int main(void)
{
    const u32 size = 1024;
    u8 bytes[size];
    int i = 0;
    u32 old_crc = 0xFFFFFFFF;
    u32 new_crc = 0;
    CRC_BuildTable();
    for(int j = 0;j<100;j++)
    {
        for(int i = 0 ;i<size;i++)
        {
            bytes[i] = (((i *i)-2*i+8)*j);
            //printf ("%x ",bytes[i]);
        }
        //CRC_BuildTable();

        new_crc = CRC_Calculate(0xFFFFFFFF,&bytes,1024);

        if (old_crc == new_crc)
           printf("match\r\n");

//        printf("%u %u\r\n",new_crc, CRC_Calculate_ARM(0xFFFFFFFF,&bytes,1024));
        old_crc=new_crc;
        //printf("%u\r\n",crc32(0xFFFFFFFF,&bytes,size));
    }
    
    clock_t begin = clock();
    for(int k=0; k<50000;k++){
        new_crc = CRC_Calculate(0xFFFFFFFF,&bytes,1017);
    }
    clock_t end = clock();
    double time_spent = (double)(end - begin);
    printf("%f %x\r\n",time_spent,new_crc);

    new_crc=0;

    begin = clock();
    for(int k=0; k<50000;k++){
        new_crc = CRC_CalculateARM(0xFFFFFFFF,&bytes,1017);
    }
    end = clock();
    time_spent = (double)(end - begin);
    printf("%f %x\r\n",time_spent, new_crc);

    printf("test\r\n");
    return(0);
}