Subject: Re: Optimizing the kernel
To: Trevor Blackwell <tlb@eecs.harvard.edu>
From: Chris G Demetriou <Chris_G_Demetriou@BALVENIE.PDL.CS.CMU.EDU>
List: port-alpha
Date: 08/14/1995 18:50:44
> I was able to boot a kernel compiled with -O, on a 3000/400.  I
> couldn't get it to work using the native compiler, but it worked with
> the cross compiler. I have no idea why.

interesting.  i'm hoping to upgrade the compiler tools, etc., soon.


> Generally, the 950811 release seems pretty solid. I think changing
> MB()/MAGIC_READ to wbflush() helped a lot. Nice work, Chris.

thanks!  but...

wbflush() is the new name for MB()...  same code (just inlined now).
while working on the TGA driver, i realized that MB() is just ... too
common a #define to use on something strange -- and it makes more
sense as "megabytes" anyway.

MAGIC_READ is something entirely different (and it's still there).
It (according to various system programmer's manuals) ensures that
writes to the TC are complete, sort of a wbflush for the TurboChannel.
The DEC 3000 SPM said that it's a good idea to read from that magic
address, when checking TC interrupt registers.



chris