[Mesa-dev] [PATCH 17/18] i965/cs: Add max_cs_threads

Kenneth Graunke kenneth at whitecape.org
Sun Mar 22 21:34:19 PDT 2015


On Sunday, March 22, 2015 12:39:25 PM Ben Widawsky wrote:
> On Sat, Mar 14, 2015 at 09:54:27PM -0700, Jordan Justen wrote:
[snip]
> > @@ -225,7 +229,8 @@ static const struct brw_device_info 
brw_device_info_hsw_gt3 = {
> >     .max_hs_threads = 504,                           \
> >     .max_ds_threads = 504,                           \
> >     .max_gs_threads = 504,                           \
> > -   .max_wm_threads = 384                            \
> > +   .max_wm_threads = 384,                           \
> > +   .max_cs_threads = 56 /* threads/subslice */
> 
> I think this is wrong for GT1, if such a thing exists (should be 42).

HSW GT1 is a thing - several Haswell Chromebooks, for example, are GT1.

> >  static const struct brw_device_info brw_device_info_bdw_gt1 = {
> >     GEN8_FEATURES, .gt = 1,
> > diff --git a/src/mesa/drivers/dri/i965/brw_device_info.h 
b/src/mesa/drivers/dri/i965/brw_device_info.h
> > index 7c9f5d0..6b4e168 100644
> > --- a/src/mesa/drivers/dri/i965/brw_device_info.h
> > +++ b/src/mesa/drivers/dri/i965/brw_device_info.h
> > @@ -70,6 +70,7 @@ struct brw_device_info
> >     unsigned max_ds_threads;
> >     unsigned max_gs_threads;
> >     unsigned max_wm_threads;
> > +   unsigned max_cs_threads;
> >  
> >     struct {
> >        unsigned size;
> 
> You're missing BYT, BSW, and SKL. I think it's fine to not implement them 
until
> they're well tested, but maybe you could at least set a conservative default 
for
> them?

+1...proper BYT and BSW numbers, please.  A conservative default seems
appropriate for Skylake, or I'd be fine with omitting it (I didn't put
in SKL numbers for tessellation either) - I think we're still in the
"use bogus numbers" stage.
-------------- 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/20150322/34bca162/attachment.sig>


More information about the mesa-dev mailing list