From mboxrd@z Thu Jan 1 00:00:00 1970
From: bugzilla-daemon@freedesktop.org
Subject: [Bug 106928] When starting a match Rocket League crashes on "Go"
Date: Wed, 04 Jul 2018 07:50:56 +0000
Message-ID:
References:
Mime-Version: 1.0
Content-Type: multipart/mixed; boundary="===============1695097426=="
Return-path:
Received: from culpepper.freedesktop.org (culpepper.freedesktop.org
[IPv6:2610:10:20:722:a800:ff:fe98:4b55])
by gabe.freedesktop.org (Postfix) with ESMTP id EEA146EAC4
for ; Wed, 4 Jul 2018 07:50:55 +0000 (UTC)
In-Reply-To:
List-Unsubscribe: ,
List-Archive:
List-Post:
List-Help:
List-Subscribe: ,
Errors-To: dri-devel-bounces@lists.freedesktop.org
Sender: "dri-devel"
To: dri-devel@lists.freedesktop.org
List-Id: dri-devel@lists.freedesktop.org
--===============1695097426==
Content-Type: multipart/alternative; boundary="15306906550.BDD0B2.17111"
Content-Transfer-Encoding: 7bit
--15306906550.BDD0B2.17111
Date: Wed, 4 Jul 2018 07:50:55 +0000
MIME-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://bugs.freedesktop.org/
Auto-Submitted: auto-generated
https://bugs.freedesktop.org/show_bug.cgi?id=3D106928
--- Comment #18 from Miroslav =C5=A0ustek ---
(In reply to Roland Scheidegger from comment #12)
> (In reply to ubizjak from comment #11)
> > The (effectively the same patch as yours) proposed patch would be:
> >=20
> > diff --git a/src/gallium/drivers/r600/sb/sb_expr.cpp
> > b/src/gallium/drivers/r600/sb/sb_expr.cpp
> > index 7a5d62c8e8..a609d1377f 100644
> > --- a/src/gallium/drivers/r600/sb/sb_expr.cpp
> > +++ b/src/gallium/drivers/r600/sb/sb_expr.cpp
> > @@ -714,6 +714,8 @@ bool expr_handler::fold_assoc(alu_node *n) {
> >=20=20
> > n->src.resize(2);
> > n->bc.set_op(ALU_OP2_ADD);
> > + fold_alu_op2(*n);
> > + return true;
> > }
> > } else if (last_arg >=3D 0) {
> > n->src[0] =3D a->src[last_arg];
> >=20
> > WDYT?
>=20
> I am not quite convinced it's ok to return true (in fold_alu_op3) if the
> expression hasn't really been folded. You are quite right that just above=
it
> looks similar, but all other places always return the return value of
> fold_alu_op2 when calling into it from fold_alu_op3.
> (Not saying it isn't correct, just saying I can't tell...)
I guess it is OK. There already is the same pattern in fold_mul_add where it
tries to fold:
> ADD x, MUL(y, z)
into:
> MULADD y, z, x
and when it succeeds, it does:
>
> fold_alu_op3(*n);
> return true;
That means that no matter if fold_alu_op3 folds the instruction any further,
fold_mul_add returns true because it converted ADD and MUL into MULADD.
In fold_alu_op2 where fold_mul_add is called, the return value is used in an
early return condition:
> if (n.bc.op =3D=3D ALU_OP2_ADD) {
> if (fold_mul_add(&n))
> return true;
> }
It seems OK to me to return true after calling fold_alu_op2 in fold_alu_op3=
as
a sign that the node should not be folded as op3 anymore.
Also, the return value of fold_alu_op* is thrown away anyway here:
> bool expr_handler::try_fold(value* v) {
> assert(!v->gvn_source);
>=20
> if (v->def)
Here expr_handler::try_fold(node* n) calls the fold_alu_op* methods:
> try_fold(v->def);
>=20
> if (v->gvn_source)
> return true;
>=20
> return false;
> }
I like the patch that ubizjak@gmail.com posted:
> diff --git a/src/gallium/drivers/r600/sb/sb_expr.cpp b/src/gallium/driver=
s/r600/sb/sb_expr.cpp
> index 1df78da660..8cbff6f577 100644
> --- a/src/gallium/drivers/r600/sb/sb_expr.cpp
> +++ b/src/gallium/drivers/r600/sb/sb_expr.cpp
> @@ -723,6 +723,8 @@ bool expr_handler::fold_assoc(alu_node *n) {
>=20=20
> n->src.resize(2);
> n->bc.set_op(ALU_OP2_ADD);
> + fold_alu_op2(n);
> + return true;
> }
> } else if (last_arg >=3D 0) {
> n->src[0] =3D a->src[last_arg];
--=20
You are receiving this mail because:
You are the assignee for the bug.=
--15306906550.BDD0B2.17111
Date: Wed, 4 Jul 2018 07:50:55 +0000
MIME-Version: 1.0
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://bugs.freedesktop.org/
Auto-Submitted: auto-generated
Comme=
nt # 18
on bug 10692=
8
from Miroslav =C5=A0ustek
(In reply to Roland Scheidegger from comment #12)
> (In reply to ubizjak from comment #11)
> > The (effectively the same patch as yours) proposed patch would be:
> >=20
> > diff --git a/src/gallium/drivers/r600/sb/sb_expr.cpp
> > b/src/gallium/drivers/r600/sb/sb_expr.cpp
> > index 7a5d62c8e8..a609d1377f 100644
> > --- a/src/gallium/drivers/r600/sb/sb_expr.cpp
> > +++ b/src/gallium/drivers/r600/sb/sb_expr.cpp
> > @@ -714,6 +714,8 @@ bool expr_handler::fold_assoc=
(alu_node *n) {
> >=20=20
> > n->src.resize(2);
> > n->bc.set_op(ALU_OP2_ADD);
> > + fold_alu_op2(*n);
> > + return true;
> > }
> > } else if (last_arg >=3D 0) {
> > n->src[0] =3D a->src[last_arg];
> >=20
> > WDYT?
>=20
> I am not quite convinced it's ok to return true (in fold_alu_op3) if t=
he
> expression hasn't really been folded. You are quite right that just ab=
ove it
> looks similar, but all other places always return the return value of
> fold_alu_op2 when calling into it from fold_alu_op3.
> (Not saying it isn't correct, just saying I can't tell...)
I guess it is OK. There already is the same pattern in fold_mul_add where it
tries to fold:
> ADD x, MUL(y, z)
into:
> MULADD y, z, x
and when it succeeds, it does:
>
> fold_alu_op3(*n);
> return true;
That means that no matter if fold_alu_op3 folds the instruction any further,
fold_mul_add returns true because it converted ADD and MUL into MULADD.
In fold_alu_op2 where fold_mul_add is called, the return value is used in an
early return condition:
> if (n.bc.op =3D=3D ALU_OP2_ADD) {
> if (fold_mul_add(&n))
> return true;
> }
It seems OK to me to return true after calling fold_alu_op2 in fold_alu_op3=
as
a sign that the node should not be folded as op3 anymore.
Also, the return value of fold_alu_op* is thrown away anyway here:
> bool expr_handler::try_fold(value* v) {
> assert(!v->gvn_source);
>=20
> if (v->def)
Here expr_handler::try_fold(node* n) calls the fold_alu_op* methods:
> try_fold(v->def);
>=20
> if (v->gvn_source)
> return true;
>=20
> return false;
> }
I like the patch that ubizjak@=
gmail.com posted:
> diff --git a/src/gallium/drivers/r600/sb/sb_expr=
.cpp b/src/gallium/drivers/r600/sb/sb_expr.cpp
> index 1df78da660..8cbff6f577 100644
> --- a/src/gallium/drivers/r600/sb/sb_expr.cpp
> +++ b/src/gallium/drivers/r600/sb/sb_expr.cpp
> @@ -723,6 +723,8 @@ bool expr_handler::fold_assoc(alu_=
node *n) {
>=20=20
> n->src.resize(2);
> n->bc.set_op(ALU_OP2_ADD);
> + fold_alu_op2(n);
> + return true;
> }
> } else if (last_arg >=3D 0) {
> n->src[0] =3D a->src[last_arg];
You are receiving this mail because:
- You are the assignee for the bug.
=
--15306906550.BDD0B2.17111--
--===============1695097426==
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: base64
Content-Disposition: inline
X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KZHJpLWRldmVs
IG1haWxpbmcgbGlzdApkcmktZGV2ZWxAbGlzdHMuZnJlZWRlc2t0b3Aub3JnCmh0dHBzOi8vbGlz
dHMuZnJlZWRlc2t0b3Aub3JnL21haWxtYW4vbGlzdGluZm8vZHJpLWRldmVsCg==
--===============1695097426==--