PCI bus access method

[Gabor Kerenyi]

Hi!

I'm curious about which one of the following code
segments is executed faster. They are in an interrupt
handler, the readings and writings are done on the PCI bus.

__u32 tmp;
int i;
for (i=0;i<128;i++) {
	tmp=readb(dev->dpmem+i);
	if (!tmp) continue;
	writeb(0,dev->dpmem+i);
	some_code;
}

256 PCI bus access in the worst case, but fewer code

or

__u32 tmp;
int i,j;
char *p=&tmp;
for (i=0;i<128;i+=4) {
	tmp=readl(dev->dpmem+i);
	if (!tmp) continue;
	writel(0, dev->dpmem+i);
	for (j=0;j<4;j++)
		if (*(p+j)) {
			some_code;
		}
}

64 PCI bus access in the worst case but more code

Thanks,

Gabor