[Bug 38800] glXSwapBuffersMscOML is slow on AMD Fusion but not on Intel 945 w/Atom - bugzilla-daemon

From: bugzilla-daemon@freedesktop.org
To: dri-devel@lists.freedesktop.org
Subject: [Bug 38800] glXSwapBuffersMscOML is slow on AMD Fusion but not on Intel 945 w/Atom
Date: Thu,  7 Jul 2011 12:06:40 -0700 (PDT)	[thread overview]
Message-ID: <20110707190641.0AC97130052@annarchy.freedesktop.org> (raw)
In-Reply-To: <bug-38800-502@http.bugs.freedesktop.org/>

https://bugs.freedesktop.org/show_bug.cgi?id=38800

--- Comment #34 from Mario Kleiner <mario.kleiner@tuebingen.mpg.de> 2011-07-07 12:06:40 PDT ---
On Jul 7, 2011, at 5:18 PM, Alex Deucher wrote:

> 2011/7/7 Michel Dänzer <michel@daenzer.net>:
>> On Don, 2011-07-07 at 11:00 -0400, Alex Deucher wrote:
>>> On Thu, Jul 7, 2011 at 10:45 AM, Mario Kleiner
>>> <mario.kleiner@tuebingen.mpg.de> wrote:
>>>> Jerome, can you please forward this to the bugzilla discussion? My
>>>> web-access to bugzilla is dead at the moment [Internet problems]:
>>>>
>>>> Need to run to a meeting for two hours or so, will be right back  
>>>> after the
>>>> break ;-)
>>>>
>>>> What about this? It seems to address our issues:
>>>>
>>>> We could use the 'vline' irq's if Alex thinks they are reliable and
>>>> supported on all asic's. If we program vline irq's to trigger  
>>>> immediately
>>>> after end of vblank, e.g., scanline 1, we could do deferred  
>>>> pageflip
>>>> programming there, and wait for flip completion/timestamping/etc.
>>>
>>> The only tricky thing with vlines is that the ddx uses them for tear
>>> free rendering so a command stream may change the vline at any time.
>>> However, we could adjust the CS checker to override the vline from
>>> userspace with whatever the kernel happened to need.
>>
>> What about the vsync interrupts? Are those still available as of  
>> AVIVO?
>
> Here's a list of the available display int srcs on r6xx+:
> VBLANK
> VLINE
> SNAPSHOT
> FORCE_COUNT
> FORCE_VSYNC
> TRIG_A
> TRIG_B
> V_UPDATE
> GRPH_PFLIP
>
> 7xx+ adds:
> VSYNC_NOM
> SET_V_TOTAL_MIN
>
> V_UPDATE looks like the most useful.  It generates an interrupt when
> the double buffered display regs switch (falling edge).
>

That sounds good. V_UPDATE irq would mark the end of the portion of  
vblank that we care about, so we could return a finished pageflip  
event earlier and don't need to mess with the vline and ddx.

Btw. i've added Simon, and forwarded the last e-mails, we "lost" him.  
Webinterface works again for me, should we move back to bugzilla?
-mario

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel