John Patrick Poet
john at BlueSkyTours.com
Thu Dec 16 16:19:17 UTC 2004
On Thu, 16 Dec 2004, Daniel Thor Kristjansson wrote:
> On Wed, 15 Dec 2004, John Patrick Poet wrote:
> ]I have another request: Please take all the asserts out of the code.
> I generally agree so I'll go on an assert hunt in the backend, I only
> added them when I thought there was no provision for a errored return.
> This assert was in the original code in another form.
> We should probably record video without audio anyway, right?
Probably. I only had that assert happen the one time, so I don't think it
is common, but....
> ]P.S. I am working a new version of the ringbuffer code. Not sure if I can get
> ]it to handle a simultaneous compile job, but I am trying.
> It's an improvement as is. If you don't think there will be major
> changes maybe it should be committed as is?
My original cut at this read directly from the device driver into the
ringbuffer, but employed a second buffer when reading *from* the
ringbuffer. It can be argued that this is inefficient, and that extra
memcpy is wasteful. The version of the ringbuffer patch you are using
processes data directly out of the ringbuffer.
However, in "real world" testing, I have never been able to get this
"efficient" version to work as well as the version which "double buffered".
I have been working on cleaning up that original "double buffer" version,
and getting it ready to post. I started up three shows recording before I
left for work. Depending on how that test goes, I may post it tonight.
More information about the mythtv-dev