[Qemu-devel] Fix (Was: segfault in current cvs)

[Qemu-devel] Fix (Was: segfault in current cvs)

[Renzo Davoli]

The fix for the problem on ppc is:

in exec-all.h:

#if defined(__alpha__)
#define CODE_GEN_BUFFER_SIZE     (2 * 1024 * 1024)
#elif defined(__powerpc__)
//#define CODE_GEN_BUFFER_SIZE     (6 * 1024)   This is the value in the CVS
#define CODE_GEN_BUFFER_SIZE     (6 * 1024 * 1024) //This sounds better!
#else
#define CODE_GEN_BUFFER_SIZE     (8 * 1024 * 1024)
#endif

Otherwise in exec.c CODE_GEN_BUFFER_MAX_SIZE becomes negative and all
the tb_alloc code get mad!

I hope this helps.

ciao
	renzo

[Qemu-devel] Re: Fix (Was: segfault in current cvs)

[Renzo Davoli]

[-- Attachment #1: Type: text/plain, Size: 181 bytes --]

On Wed, Jan 14, 2004 at 05:57:29PM +0100, Renzo Davoli wrote:
> The fix for the problem on ppc is:
With the fix W98 boots! (and seems to work)

see the attachment.....

ciao
	renzo

[-- Attachment #2: snapshot.png --]
[-- Type: image/png, Size: 49682 bytes --]

Re: [Qemu-devel] Re: Fix (Was: segfault in current cvs)

[Fabrice Bellard]

Yes, Windows 98 mostly works now, but there are still bugs in the 
install. Windows 2000/XP still have problems in the install (a strange 
IDE related error) and I have never tried to boot them.

What kind of performance do you get on PPC ? Do you feel it is more 
usable than Bochs ? On PPC, I was forced to remove register variables 
usage in the system emulation, so it will be possible to be 2 or 3 times 
as fast very soon.

Fabrice.

Renzo Davoli wrote:
> On Wed, Jan 14, 2004 at 05:57:29PM +0100, Renzo Davoli wrote:
> 
>>The fix for the problem on ppc is:
> 
> With the fix W98 boots! (and seems to work)
> 
> see the attachment.....
> 
> ciao
> 	renzo

Re: [Qemu-devel] Re: Fix (Was: segfault in current cvs)

[Renzo Davoli]

On Thu, Jan 15, 2004 at 11:26:12AM +0100, Fabrice Bellard wrote:
> Yes, Windows 98 mostly works now, but there are still bugs in the 
> install. Windows 2000/XP still have problems in the install (a strange 
> IDE related error) and I have never tried to boot them.
I have started Win 98 using an image that I installed with bochs.
I have tried also to install Win 98 SE and it worked at least to me....
I have seen just two main problems:
- The CD-ROM unit is recognized and used at boot time but windows simply
  does not see it. On "System Resource" there are no CD-ROMS.
- The network unit is presented as a generic "Remote Access Device". I have not
  tried to install a specific driver as I'd need the CD-ROM to do it.
  I don't know if the "Remote Access Device" means that no network
  interface has been detected so it wants to use a modem or is the
  NE2000 card that appears with a generic name. I'd vote for the former.

I haven't heard the boot sound, maybe because I had the dsp busy with
other applications.
If I try to open the floopy and there are no images mounted it takes a
long time before to return again operational.
> 
> What kind of performance do you get on PPC ? Do you feel it is more 
> usable than Bochs ? On PPC, I was forced to remove register variables 
> usage in the system emulation, so it will be possible to be 2 or 3 times 
> as fast very soon.

This is a great news.
Now the performance is quite good with repoect to bochs. I think that
it is about 10 times at least. Win98 boot takes about 1 minute (65 
seconds) on qemu (Powerbook G4 1Ghz).
If it is possible to boost it up 3 times.... I'll call my titanium
"PentiumBook G4" ;-)
Apart from jokes, the system is usable but some operations are already
slow.
It is the case for example of the menu animations. Clicking on "start"
the menu arises slowly slowly from the bar.... 
Disk access is very fast (a few seconds for a scandisk on 500MB).
Computation is reasonable. 
Maybe the bottleneck is on the VGA graphic board emulation?

Anyway it is a great work.
Let me use this message to give a public "thank you" to Fabrice for
what he is doing for the free software.

ciao
	renzo