Re: Self-mastering?

[Drew Eckhardt <drew@Poohsticks.Org>]

In message <370A0DC4.C42B4BFA@paradigm-inc.com>, rmonaco@paradigm-inc.com write
s:
>This seems like a dumb question, but has anyone had experience with any
>PCI cards which try to master to themselves?  

I've heard rumours that some of the NCR/Symbios SCSI chips do this.

>I appear to have a SCSI
>host adapter which trys to do this.  I haven't found anything in the
>specs which would preclude such operation (perhaps as a self-test?).

The NCR chips have a MOVE MEMORY instruction, which moves some number
of bytes from source to destination.

Due to hardware limitations (no indirect addressing mode (this which means 
copying a register to the address field of a direct-addressing instruction,
which could reside in the on-chip RAM of later parts (early parts executed
code from main memory) and byte at a time access to internal registers (the 
original parts could take a few hundred microseconds for each instruction, 
so a single move-memory instruction was substantially faster than the eight 
(SFBR must be either source or destination for a register operation) register 
operations)), MOVE MEMORY instructions see some use during normal operation
of the chip.

Not special-casing addresses in the chip's address spaces would 
simplify implementation of MOVE MEMORY instruction, and such bus 
accesses could be par for the course.

-- 
<a href="http://www.poohsticks.org/drew/">Home Page</a>
For those who do, no explanation is necessary.  
For those who don't, no explanation is possible.