> I also suspect that problem is small size of transfer buffer, and > read/write error checking per each transfer, but also wish to point that > there is no need for checking weather are clusters adjacent.
Sure there is. There's a big difference if you write 16 kbytes of data as four writes of a 4 kbyte block (assuming a pcfs with 4 kb cluster size), or you perform a single 16 kbyte write after you've noticed that the four clusters that should be written are all adjacent. Especially if the internal block size of the flash memory is bigger than the pcfs' cluster size; in this case the four 4 kbyte writes would internally translate to four read, modify, write cycles. -- This message posted from opensolaris.org
