Am Montag, den 27.08.2018, 12:17 +0200 schrieb Erik Faye-Lund:
> 
> On ma., aug. 27, 2018 at 9:49 AM, Gert Wollny <gert.wollny@collabora.
> com> wrote:
> > Am Montag, den 20.08.2018, 14:10 +0200 schrieb Erik Faye-Lund:
> > >  This is quite useful for debugging shader-transpiling issues in
> > >  virglrenderer.
> > Isn't this coverted by ST_DEBUG=tgsi? 
> > 
> 
> Ah, I wasn't even aware of this. 
> 
> There's one minor difference here, though: virgl does some
> transformations on the TGSI before passing it to the host. It can
> sometimes be useful to see the difference.
> 
> > Also, virglrenderer has a variable vrend_dump_shaders in
> > vrend_renderer.c that enables dumping all the TGSI + the created
> > GLSL
> > shaders. 
> 
> Yeah, and that's also useful. But there's two differences here:
> - vrend_dump_shaders dumps the TGSI *after* parsing, comparing the
> over-the-wire TGSI and the parsed TGSI has helped me in the past to
> find flags being culled during parsing etc in the past.
> - vrend_dump_shaders needs a recompile of virglrenderer, which makes
> it a lot more inconvenient from a turn-around point of view if
> running on qemu for instance.
> 
> I don't know, perhaps this is a bit too many similar features? I can
> certainly drop this patch for now.
Given these differences it actually makes sense to have this output
available, so you have my R-b for this patch.

> Another (maybe more useful) option could be to allow a per-client
> override of the vrend_dump_shaders-functionality that could be
> enabled from the host?
I think adding the same kind of debug capability to virglrenderer that
can be enabled via environment variables would be very useful. 

Best, 
Gert 



_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev

Reply via email to