[Mesa-dev] [PATCH] i965/skl: Use 1 register for uniform pull constant payload

Kenneth Graunke kenneth at whitecape.org
Fri Feb 20 10:57:31 PST 2015


On Friday, February 20, 2015 10:33:18 AM Ben Widawsky wrote:
> On Fri, Feb 20, 2015 at 03:34:21AM -0800, Kenneth Graunke wrote:
> > On Thursday, February 19, 2015 10:48:08 PM Ben Widawsky wrote:
> > > When under dispatch_width=16 the previous code would allocate 2 registers for
> > > the payload when only one is needed. This manifested itself through bugs on SKL
> > > which needs to mess with this instruction.
> > > 
> > > Ken though this might impact shader-db, but apparently it doesn't
> > > 
> > > Cc: Kenneth Graunke <kenneth at whitecape.org>
> > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=89118
> > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88999
> > > Signed-off-by: Ben Widawsky <ben at bwidawsk.net>
> > > ---
> > >  src/mesa/drivers/dri/i965/brw_fs.cpp | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
> > > index c46e1d7..24125cc 100644
> > > --- a/src/mesa/drivers/dri/i965/brw_fs.cpp
> > > +++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
> > > @@ -3062,7 +3062,7 @@ fs_visitor::lower_uniform_pull_constant_loads()
> > >           assert(const_offset_reg.file == IMM &&
> > >                  const_offset_reg.type == BRW_REGISTER_TYPE_UD);
> > >           const_offset_reg.fixed_hw_reg.dw1.ud /= 4;
> > > -         fs_reg payload = vgrf(glsl_type::uint_type);
> > > +         fs_reg payload = fs_reg(GRF, alloc.allocate(1));
> > >  
> > >           /* We have to use a message header on Skylake to get SIMD4x2 mode.
> > >            * Reserve space for the register.
> > > 
> > 
> > Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>
> 
> Given that this seems to be required to get SKL stable, do I want to Cc 1.5 too?

That seems reasonable to me.  It likely won't apply (or build), since I
don't think 10.5 has alloc.allocate().  To be nice to Emil, I'm guessing
the right thing to do is probably to cherry-pick it to 10.5, change
alloc.allocate(1) to virtual_grf_alloc(1), and send that patch to
mesa-stable at lists.freedesktop.org marking it as a backport.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.freedesktop.org/archives/mesa-dev/attachments/20150220/beb0404e/attachment.sig>


More information about the mesa-dev mailing list