On Fri, Oct 27, 2017 at 1:18 AM, Iago Toral <ito...@igalia.com> wrote:

> On Wed, 2017-10-25 at 16:25 -0700, Jason Ekstrand wrote:
> > The only things that adjust fs_visitor::max_dispatch_width are render
> > target writes which don't happen in compute shaders so they're
> > pointless.
> > ---
> >  src/intel/compiler/brw_fs.cpp | 6 ++----
> >  1 file changed, 2 insertions(+), 4 deletions(-)
> >
> > diff --git a/src/intel/compiler/brw_fs.cpp
> > b/src/intel/compiler/brw_fs.cpp
> > index a23366b..4c362ba 100644
> > --- a/src/intel/compiler/brw_fs.cpp
> > +++ b/src/intel/compiler/brw_fs.cpp
>
> Maybe add an assert before this to check that max_dispatch_width is >=
> 32 as expected here?
>

Done.


> > @@ -6818,8 +6818,7 @@ brw_compile_cs(const struct brw_compiler
> > *compiler, void *log_data,
> >                   NULL, /* Never used in core profile */
> >                   shader, 16, shader_time_index);
> >     if (likely(!(INTEL_DEBUG & DEBUG_NO16)) &&
> > -       !fail_msg && v8.max_dispatch_width >= 16 &&
> > -       min_dispatch_width <= 16) {
> > +       !fail_msg && min_dispatch_width <= 16) {
> >        /* Try a SIMD16 compile */
> >        if (min_dispatch_width <= 8)
> >           v16.import_uniforms(&v8);
> > @@ -6843,8 +6842,7 @@ brw_compile_cs(const struct brw_compiler
> > *compiler, void *log_data,
> >     fs_visitor v32(compiler, log_data, mem_ctx, key, &prog_data-
> > >base,
> >                   NULL, /* Never used in core profile */
> >                   shader, 32, shader_time_index);
> > -   if (!fail_msg && v8.max_dispatch_width >= 32 &&
> > -       (min_dispatch_width > 16 || (INTEL_DEBUG & DEBUG_DO32))) {
> > +   if (!fail_msg && (min_dispatch_width > 16 || (
>
> Maybe use unlikely() with (INTEL_DEBUG & DEBUG_DO32)?
>

That doesn't really go along with this change.  I've been meaning to rework
DO32 and I'm happy to fix it as part of that.
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev

Reply via email to