[Qemu-devel] Re: [PATCH 12/16] vmware_vga: remove !EMBED_STDVGA code

[Juan Quintela <quintela@redhat.com>]

andrzej zaborowski <balrogg@gmail.com> wrote:
> 2009/10/14 Juan Quintela <quintela@redhat.com>:
>> andrzej zaborowski <balrogg@gmail.com> wrote:
>>> 2009/10/14 Juan Quintela <quintela@redhat.com>:
>>>> It don't compile.  And the trivial fixes (change vga.foo field to foo field
>>>> don't work either.  No output
>>>
>>> Just a note that there's supposed to be no output because your guest
>>> OS (and BIOS) expect the standard VGA functionality to be there.  The
>>> define is only useful for debugging vmware_vga or OS driver.  It's ok
>>> to drop it if you want.
>>
>> I preffer to drop it because it has no chance of working (suspend/resume
>> code is not there for instance).
>>
>> Once that we are there.  I did a fast try at enabling DIRECT_VRAM with
>> the same not working result.  Any idea if it would make things
>> better/fast/... whatever?
>
> Currently it probably makes things slower.  I have not digged through
> the newer SDL code deep enough, to tell if it's possible to create a
> SDL surface directly from guest RAM provided it's contiguously mapped
> on host -- this is what VMware does and it's one of the expected
> benefits from using vmware_vga that isn't there, and I think using DGA
> saves them another unneeded copy.  This would need reimplementing the
> pieces guarded by #ifdef DIRECT_VRAM.

Without DIRECT_VRAM, my un-scientific test (run glxgears) show that
cirrus_vga is a bit faster (207 vs 190 fps), neither the test of the
results something to write home about.

I have another problem with the driver: depth.

I have to change s->depth to 32 in vmsvga_reset() to make it work
correctly on my setup (default qemu.git tree, i.e. nothing fancy),
running with sdl.  Any clue here?

Anthony thinks that the problem happens at the memcpy() in
vmsvga_update_rect(), but I haven't had the time to look at how to fix
it.  Any idea here?

Later, Juan.