[mythtv] Re: pvr-350 performance patch - simplified

[John Harvey]

I'd thought about it but I would prefer to move to the installed header
rather than a copy of the header file so this was a start in that direction.
If you have an installed header file this functionality was there always so
the only problem would be saying you want to compile for 0.2 and then
reverting to a 0.1.9 driver so it was really an assumption that this would
work, but I am not that worried and yes in this case we could detect and
revert to the old code.
In the original patch there were other things that worked around bugs in
0.1.9 which couldn't be detected at run time in the same way so the #ifdef
was used for changes in the behavior of the prep_dma call as well where
0.1.9 needed page aligned starting points and 0.2 doesn't.

So I guess I'm not that worried. I don't really want to maintain ivtv header
files in every dependent project (and myth isn't the only one) but now that
ivtv.h is simplified in the driver it would be easier than it was before
since we should be able to just copy a new version in occasionally.


John

> -----Original Message-----
> From: mythtv-dev-bounces at mythtv.org [mailto:mythtv-dev-bounces at mythtv.org]
> On Behalf Of David Engel
> Sent: 03 February 2005 03:56
> To: Development of mythtv
> Subject: [mythtv] Re: pvr-350 performance patch - simplified
> 
> On Wed, Feb 02, 2005 at 10:39:24PM -0000, John Harvey wrote:
> > Attached is the re-worked pvr-350 patch. I have removed all the region
> > update code so this now has an option to include installed ivtv header
> file
> > in settings.pro and if its there it uses the BLT_FILL ioctl instead of
> the
> > memset /prep_frame method. Some of the other code is re-organised to
> make
> > the use of the BLT_FILL a bit easier.
> 
> John, have you considered making your change use a run-time decision
> instead of a compile-time one?  It looks like it should be fairly
> easy, eg.
> 
>    Replace the old ivtv-ext-api.h in Myth with ivtv.h.
> 
>    Initialize a flag assuming IVTVFB_IOCTL_BLT_FILL is supported.
> 
>    If IVTVFB_IOCTL_BLT_FILL fails with EINVAL, clear the flag and
>    fallback to using memset/IVTVFB_IOCTL_PREP_FRAME.
> 
> David
> --
> David Engel
> gigem at comcast.net