[Mesa-dev] [PATCH] i965: Skip register write checks if cmd_parser_version >= 2.

[Daniel Vetter]

On Tue, Oct 25, 2016 at 11:16:56AM -0700, Francisco Jerez wrote:
> Kenneth Graunke <kenneth at whitecape.org> writes:
> 
> > If the kernel advertises a new enough command parser version, then we
> > can just assume that register writes will work and not bother executing
> > commands on the GPU to test it.
> >
> > This should speed up context creation.
> >
> > From the command parser version history (i915_cmd_parser.c):
> >
> >   * 1. Initial version. Checks batches and reports violations, but leaves
> >   *    hardware parsing enabled (so does not allow new use cases).
> >   * 2. Allow access to the MI_PREDICATE_SRC0 and
> >   *    MI_PREDICATE_SRC1 registers.
> >
> > Both of the things we check for were added before version 2, but
> > 1 doesn't sound sufficient.  The point is to skip work on modern
> > kernels, so requiring version 2 will work.
> >
> > Signed-off-by: Kenneth Graunke <kenneth at whitecape.org>
> > ---
> >  src/mesa/drivers/dri/i965/intel_extensions.c | 8 ++++++++
> >  1 file changed, 8 insertions(+)
> >
> > diff --git a/src/mesa/drivers/dri/i965/intel_extensions.c b/src/mesa/drivers/dri/i965/intel_extensions.c
> > index 66079b5..2a5a3b7 100644
> > --- a/src/mesa/drivers/dri/i965/intel_extensions.c
> > +++ b/src/mesa/drivers/dri/i965/intel_extensions.c
> > @@ -46,6 +46,10 @@ can_do_pipelined_register_writes(struct brw_context *brw)
> >     if (brw->gen != 7)
> >        return true;
> >  
> > +   /* No need to execute commands to check if the kernel advertises it. */
> > +   if (brw->screen->cmd_parser_version >= 2)
> > +      return true;
> > +
> >     static int result = -1;
> >     if (result != -1)
> >        return result;
> > @@ -106,6 +110,10 @@ can_write_oacontrol(struct brw_context *brw)
> >     if (brw->gen < 6 || brw->gen >= 8)
> >        return false;
> >  
> > +   /* No need to execute commands to check if the kernel advertises it. */
> > +   if (brw->screen->cmd_parser_version >= 2)
> > +      return true;
> > +
> 
> I'm not sure this gives you any guarantees that the command parser is
> going to be active?  Has the command parser been enabled by default on
> all kernel versions exposing cmd_parser_version >= 2?  Wouldn't this
> break things if the command parser is disabled in the kernel command
> line?  [Currently it would just cause some extensions to be disabled]

We have the stance that touching any module option is an unsupported
configuration, don't do that (and we'll ingore bug reports). Either way,
the kernel reports 0 when the cmd parser is disabled.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch