<div dir="ltr"><div class="gmail_quote"><div dir="ltr">On Mon, Oct 15, 2018 at 1:48 PM Ian Romanick <<a href="mailto:idr@freedesktop.org">idr@freedesktop.org</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On 10/14/2018 03:11 PM, Matt Turner wrote:<br> > Unfortunately some int64 lowerings generate more int64 operations, so we<br> > need to call this function a few times. Also call<br> > nir_lower_alu_to_scalar() beforehand to make more int64 operations<br> > available for lowering.<br> > ---<br> >Â src/intel/compiler/brw_nir.c | 10 +++++++---<br> >Â 1 file changed, 7 insertions(+), 3 deletions(-)<br> > <br> > diff --git a/src/intel/compiler/brw_nir.c b/src/intel/compiler/brw_nir.c<br> > index f61baee230a..066724c58a6 100644<br> > --- a/src/intel/compiler/brw_nir.c<br> > +++ b/src/intel/compiler/brw_nir.c<br> > @@ -670,12 +670,16 @@ brw_preprocess_nir(const struct brw_compiler *compiler, nir_shader *nir)<br> >Â Â Â */<br> >Â Â Â OPT(nir_opt_algebraic);<br> >Â <br> > +Â Â if (is_scalar) {<br> > +Â Â Â OPT(nir_lower_alu_to_scalar);<br> > +Â Â }<br> > +<br> >Â Â Â /* Lower int64 instructions before nir_optimize so that loop unrolling<br> >Â Â Â * sees their actual cost.<br> >Â Â Â */<br> > -Â Â nir_lower_int64(nir, nir_lower_imul64 |<br> > -Â Â Â Â Â Â Â Â Â Â Â Â nir_lower_isign64 |<br> > -Â Â Â Â Â Â Â Â Â Â Â Â nir_lower_divmod64);<br> > +Â Â while (nir_lower_int64(nir, nir_lower_imul64 |<br> > +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â nir_lower_isign64 |<br> > +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â nir_lower_divmod64));<br> <br> I don't know that we have a specific coding standard about this, but I<br> always like<br> <br> Â Â Â Â while (foo)<br> Â Â Â Â Â Â Â Â /* empty */ ;<br> <br> just to make it obvious.<br></blockquote><div><br></div><div>I think I'd opt for "continue" but yeah, I think one or the other is better than silently putting the semicolon at the end.<br></div></div></div>