The scsi v40.12 (Kickstart 3.1) and the v43.45 (Boing Bag 2) use big loops for copying files. It's IDE, not SCSI, all is done by the CPU !
On the A4000D, these loops give a faster transfer rate when using .w (16 bit) than .l (32 bit)
On the A600/A1200, .l is faster...
Maybe I found why big loops .w are faster than .l in the read/write function on the A4000D scsi driver : PM me for explanations if you know how are working U901 & U902.
These two Gals need a simple fix (I hope) for working faster using .l !
I use this funny trick to get all ax+0 / ax+1 / ax+2 / ax+3 using big loops .l all the time in my scsi beta :
- ax+0 => loop .l
- ax+1 => .b + .w => loop .l => .b
- ax+2 => .w => loop .l => .w
- ax+3 => .b => loop .l => .w + .b
... and I wanna use it into the A4000D scsi driver, of course !!
