From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by sourceware.org (Postfix) with ESMTP id CA8C0385E00A for ; Thu, 20 Jul 2023 17:40:26 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org CA8C0385E00A Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=arm.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=arm.com Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.121.207.14]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id 557EC2F4; Thu, 20 Jul 2023 10:41:09 -0700 (PDT) Received: from localhost (e121540-lin.manchester.arm.com [10.32.110.72]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 90FD93F6C4; Thu, 20 Jul 2023 10:40:25 -0700 (PDT) From: Richard Sandiford To: Richard Biener Mail-Followup-To: Richard Biener ,Richard Biener via Gcc-patches , richard.sandiford@arm.com Cc: Richard Biener via Gcc-patches Subject: Re: [PATCH] tree-optimization/110742 - fix latent issue with permuting existing vectors References: <39395404-A479-4451-A2E8-2BF46AE766C0@suse.de> Date: Thu, 20 Jul 2023 18:40:24 +0100 In-Reply-To: <39395404-A479-4451-A2E8-2BF46AE766C0@suse.de> (Richard Biener's message of "Thu, 20 Jul 2023 19:10:49 +0200") Message-ID: User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/26.3 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable X-Spam-Status: No, score=-26.5 required=5.0 tests=BAYES_00,GIT_PATCH_0,KAM_DMARC_NONE,KAM_DMARC_STATUS,KAM_LAZY_DOMAIN_SECURITY,SPF_HELO_NONE,SPF_NONE,TXREP,T_SCC_BODY_TEXT_LINE autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org List-Id: Richard Biener writes: >> Am 20.07.2023 um 18:59 schrieb Richard Sandiford : >>=20 >> =EF=BB=BFRichard Biener writes: >>>>> Am 20.07.2023 um 16:09 schrieb Richard Sandiford : >>>>=20 >>>> =EF=BB=BFRichard Biener via Gcc-patches writ= es: >>>>> When we materialize a layout we push edge permutes to constant/extern= al >>>>> defs without checking we can actually do so. For externals defined >>>>> by vector stmts rather than scalar components we can't. >>>>>=20 >>>>> Bootstrapped and tested on x86_64-unknown-linux-gnu. >>>>>=20 >>>>> OK? >>>>>=20 >>>>> Thanks, >>>>> Richard. >>>>>=20 >>>>> PR tree-optimization/110742 >>>>> * tree-vect-slp.cc (vect_optimize_slp_pass::get_result_with_layout): >>>>> Do not materialize an edge permutation in an external node with >>>>> vector defs. >>>>> (vect_slp_analyze_node_operations_1): Guard purely internal >>>>> nodes better. >>>>>=20 >>>>> * g++.dg/torture/pr110742.C: New testcase. >>>>> --- >>>>> gcc/testsuite/g++.dg/torture/pr110742.C | 47 +++++++++++++++++++++++++ >>>>> gcc/tree-vect-slp.cc | 8 +++-- >>>>> 2 files changed, 53 insertions(+), 2 deletions(-) >>>>> create mode 100644 gcc/testsuite/g++.dg/torture/pr110742.C >>>>>=20 >>>>> diff --git a/gcc/testsuite/g++.dg/torture/pr110742.C b/gcc/testsuite/= g++.dg/torture/pr110742.C >>>>> new file mode 100644 >>>>> index 00000000000..d41ac0479d2 >>>>> --- /dev/null >>>>> +++ b/gcc/testsuite/g++.dg/torture/pr110742.C >>>>> @@ -0,0 +1,47 @@ >>>>> +// { dg-do compile } >>>>> + >>>>> +struct HARD_REG_SET { >>>>> + HARD_REG_SET operator~() const { >>>>> + HARD_REG_SET res; >>>>> + for (unsigned int i =3D 0; i < (sizeof(elts) / sizeof((elts)[0])= ); ++i) >>>>> + res.elts[i] =3D ~elts[i]; >>>>> + return res; >>>>> + } >>>>> + HARD_REG_SET operator&(const HARD_REG_SET &other) const { >>>>> + HARD_REG_SET res; >>>>> + for (unsigned int i =3D 0; i < (sizeof(elts) / sizeof((elts)[0])= ); ++i) >>>>> + res.elts[i] =3D elts[i] & other.elts[i]; >>>>> + return res; >>>>> + } >>>>> + unsigned long elts[4]; >>>>> +}; >>>>> +typedef const HARD_REG_SET &const_hard_reg_set; >>>>> +inline bool hard_reg_set_subset_p(const_hard_reg_set x, const_hard_r= eg_set y) { >>>>> + unsigned long bad =3D 0; >>>>> + for (unsigned int i =3D 0; i < (sizeof(x.elts) / sizeof((x.elts)[0= ])); ++i) >>>>> + bad |=3D (x.elts[i] & ~y.elts[i]); >>>>> + return bad =3D=3D 0; >>>>> +} >>>>> +inline bool hard_reg_set_empty_p(const_hard_reg_set x) { >>>>> + unsigned long bad =3D 0; >>>>> + for (unsigned int i =3D 0; i < (sizeof(x.elts) / sizeof((x.elts)[0= ])); ++i) >>>>> + bad |=3D x.elts[i]; >>>>> + return bad =3D=3D 0; >>>>> +} >>>>> +extern HARD_REG_SET rr[2]; >>>>> +extern int t[2]; >>>>> +extern HARD_REG_SET nn; >>>>> +static HARD_REG_SET mm; >>>>> +void setup_reg_class_relations(void) { >>>>> + HARD_REG_SET intersection_set, union_set, temp_set2; >>>>> + for (int cl2 =3D 0; cl2 < 2; cl2++) { >>>>> + temp_set2 =3D rr[cl2] & ~nn; >>>>> + if (hard_reg_set_empty_p(mm) && hard_reg_set_empty_p(temp_set2))= { >>>>> + mm =3D rr[0] & nn; >>>>> + if (hard_reg_set_subset_p(mm, intersection_set)) >>>>> + if (!hard_reg_set_subset_p(mm, temp_set2) || >>>>> + hard_reg_set_subset_p(rr[0], rr[t[cl2]])) >>>>> + t[cl2] =3D 0; >>>>> + } >>>>> + } >>>>> +} >>>>> diff --git a/gcc/tree-vect-slp.cc b/gcc/tree-vect-slp.cc >>>>> index 693621ca990..1d79c77e8ce 100644 >>>>> --- a/gcc/tree-vect-slp.cc >>>>> +++ b/gcc/tree-vect-slp.cc >>>>> @@ -5198,7 +5198,10 @@ vect_optimize_slp_pass::get_result_with_layout= (slp_tree node, >>>>> return result; >>>>>=20 >>>>> if (SLP_TREE_DEF_TYPE (node) =3D=3D vect_constant_def >>>>> - || SLP_TREE_DEF_TYPE (node) =3D=3D vect_external_def) >>>>> + || (SLP_TREE_DEF_TYPE (node) =3D=3D vect_external_def >>>>> + && (to_layout_i =3D=3D 0 >>>>> + /* We can't permute vector defs. */ >>>>> + || SLP_TREE_VEC_DEFS (node).is_empty ()))) >>>>=20 >>>> Guess it's personal preference, but IMO it's easier to follow without = the >>>> to_layout_i condition, so that it ties directly to the create_partitio= ns >>>> test. >>>=20 >>> I don=E2=80=99t understand- in the code guarding this we seem to expect= to_layout_i =3D=3D 0 and that=E2=80=99s the case we can handle as noop. I= didn=E2=80=99t understand why the function doesn=E2=80=99t always just do = nothing in this case though, so I must have missed something. >>=20 >> OK, so I guess that disproves that my way is easier to understand :) >>=20 >> I think logically, the code is doing the equivalent of: >>=20 >> int partition_i =3D m_vertices[node->vertex].partition; >> if (partition < 0) >> { >> /* If the vector is uniform or unchanged, there's nothing to do. */ >> ...=20=20=20=20=20=20 >> } >> else >> { >> ... Return node if to_layout_i matches this partition's chosen layo= ut... >> } >>=20 >> And I guess I should have written it that way. >>=20 >> So when there is no partition, we have a constant or external def >> built from individual scalars. We can use the node as-is if the >> caller wants an unpermuted node or if all elements are equal >> (so that the permutation doesn't matter). Otherwise we need >> to permute the scalars. >>=20 >> When there is a partition, we can use the node as-is if the caller >> wants the layout that was chosen for that partition. Otherwise we >> need a new VEC_PERM_EXPR node. >>=20 >> In the particular case of external defs built from vectors, we're >> guaranteed that the node's chosen layout is 0 (i.e. the original layout), >> and so both ways work. > > Hmm, but I arrived here with that not being the case =E2=80=A6 (the chose= n Lay-out not zero) By chosen layout for the partition, I meant: unsigned int from_layout_i =3D m_partitions[partition_i].layout; That should be 0 for these external vector nodes due to the start_choosing_layout code that you mentioned in the PR. to_layout_i (the layout that the caller wants for a use of the node) can be anything, so yeah, to_layout_i !=3D 0 has to go through the "else" arm and generate a VEC_PERM_EXPR. But because from_layout_i =3D=3D 0 for these nodes, the "else" arm should handle all cases correctly, not just the to_layout_i !=3D 0 case. Thanks, Richard