SwapContext hook back

SwapContext hook back

[Westermann Fu]

[-- Attachment #1.1: Type: text/plain, Size: 578 bytes --]

Hi,

I found that for DRI callback 'SwapContext' hook handler, seems no driver
interested with it, except for old glint video driver do real work in it.
Does it mean context switch concept outdated for current graphics hardware,
but I found that intel hardware has logical context save/store command for
these stuff? As I guess, context switch design strategy was taken out behind
background of old-style DMA graphics device such as GLINT. But how the
performance boost by using context switching rather than by reg-by-reg
emiting at that old time?   Thanks very much.

Regards

[-- Attachment #1.2: Type: text/html, Size: 605 bytes --]

[-- Attachment #2: Type: text/plain, Size: 345 bytes --]

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev

[-- Attachment #3: Type: text/plain, Size: 161 bytes --]

--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: SwapContext hook back

[Ian Romanick]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Westermann Fu wrote:

> I found that for DRI callback 'SwapContext' hook handler, seems no
> driver interested with it, except for old glint video driver do real
> work in it. Does it mean context switch concept outdated for current
> graphics hardware, but I found that intel hardware has logical context
> save/store command for these stuff? As I guess, context switch design
> strategy was taken out behind background of old-style DMA graphics
> device such as GLINT. But how the performance boost by using context
> switching rather than by reg-by-reg emiting at that old time?   Thanks
> very much.

My recollection is that this was to allow for the X server to handle
graphics context switching.  That method is certainly undesirable.  If a
driver were to implement some sort of optimized switching, the kernel is
the right place to do it.  What we've found in the past with most
hardware is that the cost of tracking dirty state tends to outweigh the
benefit of not emitting redundant state.  Different applications don't
generally have any of the expensive state in common.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAkuL8NAACgkQX1gOwKyEAw9OEQCgkZRX9T5zmOBeCNvDXx4u/OZW
8XcAnArdE7Jqnw1ihDX6vZpyq4Gr5e9n
=I2NG
-----END PGP SIGNATURE-----

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--