Re: [Qemu-devel] [Patch] Ali Chipset support for PC [+ questions about alpha-softmmu target]

[Brian Wheeler <bdwheele@indiana.edu>]

On Wed, 2008-09-10 at 09:30 +0200, Tristan Gingold wrote:
> On Sep 9, 2008, at 5:53 PM, Brian Wheeler wrote:
> 
> > Hello!
> >
> > Long time listener, first time caller...
> >
> > Since the ES40 Alpha Emulation project seems to have stalled, I've  
> > been
> > looking at what it would take to get QEmu to a state where it could
> > emulate an alpha machine.
> 
> Hi,
> 
> I had the same idea too.  Qemu should be much faster than es40 emulator.
> 

That was my thought.  Plus the devices are more fully tested.  I wrote
the IDE stuff (and misc other devices) for es40 and it was really hit or
miss on timing.


> > The es40 is basically a PC with an alpha chip instead of an x86 and it
> > uses the Ali1543C south bridge.   This basically is a copy of  
> > piix_pci.c
> > so the chipset reports the ALI vendor and device codes to qemu when  
> > the
> > "alipc" machine type is selected.
> >
> > I've booted Knoppix and it seems to work ok...which is no surpise
> > considering how minor the changes are.
> >
> > So, now to the questions:
> > * Is anyone else working on a softmmu target for the alpha?
> 
> I plan to work on it.  But first I have to finish the TCG conversion  
> of target-alpha.
> 

Cool.

> > * What state is the alpha cpu emulation in, and what's the best way to
> > test it?
> 
> Still beta.  It was not maintained nor tested for a while.  And bugs  
> were recently found.
> 

Anything that I can do to help?  I worked on es40 (mostly pc-style
hardware stuff) so if I can help I'd like to.

> > * Looking over some of the softmmu targets, it looks like the bulk of
> > target work (after the cpu emulation, of course) is just attaching all
> > of the devices in the right places.
> 
> Booting might be an issue too as alpha boots from icache.
> It looks like es40 has an IO-TLB which is not present in standard  
> PC.  I suppose Sparc64 has also
> an IO-TLB but I didn't look further.

Yeah, the icache thing will be interesting.  Looking at es40, the icache
looks like its only used during the rom decompression stage.  es40
creates a decompressed rom file on the first run and it looks like the
first uint64 is the PC, and the second uint64 is the PAL_BASE.  That
should be enough to get the system up and running (after a software-only
call_pal instruction is implemented)...and it doesn't seem to mess with
the icache in that case.

I may look at a standalone rom extractor, so you can give it the
cl67srmrom.exe file and drop out a ready-to-run rom file.

Brian

> Tristan.
> 
> 
>