Re: Can GTK+ make use of XSHM extension from MIT

["Bin Chen" <binary chen gmail com>]

On Jan 12, 2008 9:46 AM, Tristan Van Berkom <tvb gnome org> wrote:
> On Jan 11, 2008 11:01 PM, Bin Chen <binary chen gmail com> wrote:
>
> > > Gdk does all image operations in 32-bits, N800/N810 display is 16-bit.
> > > Whether you use SHM or not can be of less concern than this from the
> > > performance point of view.
> >
> > OK, so your point is because the convert process must be done on every
> > operation, so the performance will be bad? But I think if the XSHM is
> > used, at least the transfer overhead will be cut, right?
>
> What are you using to compose the image ? if you are doing direct
> pixel operations on the permanent GdkImage that you created
> and occasionally tell the X server to update a window based on
> your shared image, this should be fast.
>
> Data transfers over unix sockets generally go:
>    while (not finished writing and no errors)
>       write (8kb chunks of remaining data);
>
> If you are using shared memory, then you still need to wake up
> the X server process for it to copy the data internally in its own
> process - unless these are really huge images, you are not
> gaining much by using XSHM.
>
> Its really hard to say what will be optimal for your situation
> without a little more context on what you want to achieve.
>
OK, from the thread above I have point out the main purpose of me to
start this thread is I want to write a video player in a slow machine,
I have tested the X and framebuffer case that the difference is large,
so I am going to find the root cause so I can make the two results
closer.

Thanks.
Bin