From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from smtp-out1.suse.de (smtp-out1.suse.de [IPv6:2001:67c:2178:6::1c]) by sourceware.org (Postfix) with ESMTPS id 3E7A63858D33 for ; Wed, 1 Mar 2023 07:37:20 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org 3E7A63858D33 Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=suse.de Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=suse.de Received: from relay2.suse.de (relay2.suse.de [149.44.160.134]) by smtp-out1.suse.de (Postfix) with ESMTP id EE5B521A81 for ; Wed, 1 Mar 2023 07:37:18 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=suse.de; s=susede2_rsa; t=1677656238; h=from:from:reply-to:date:date:to:to:cc:mime-version:mime-version: content-type:content-type; bh=xiDamskb3KpGDxi8qs5ZJdgGN1nP3ZQgq4uSFfJRApo=; b=T0w/Po5pz29oxswUwBrcLmpPueWuThpdA45ZfC5Ps6HPLniwDQYbgsgjBR+WhSMgvqLqx0 etu6fa8G3DB0CtmNmgPQ0O+09ry70l7d7J9vxppOd76aKFWwyeyeiIlVQRgwbZuHLteIj9 onkTedZ+tGTfV4SY7kZtxIDG3ERe40g= DKIM-Signature: v=1; a=ed25519-sha256; c=relaxed/relaxed; d=suse.de; s=susede2_ed25519; t=1677656238; h=from:from:reply-to:date:date:to:to:cc:mime-version:mime-version: content-type:content-type; bh=xiDamskb3KpGDxi8qs5ZJdgGN1nP3ZQgq4uSFfJRApo=; b=9Ob+gAnq4whPhLZd077A8SKbOz2hUn1JzfAj7uD3kXXtOzfMtZ3R22GipqYWjAcV6epbpo 4vmJJSaWVGaXUqBQ== Received: from wotan.suse.de (wotan.suse.de [10.160.0.1]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by relay2.suse.de (Postfix) with ESMTPS id E50002C141 for ; Wed, 1 Mar 2023 07:37:18 +0000 (UTC) Date: Wed, 1 Mar 2023 07:37:18 +0000 (UTC) From: Richard Biener To: gcc-patches@gcc.gnu.org Subject: [PATCH] tree-optimization/108950 - widen-sum reduction ICE User-Agent: Alpine 2.22 (LSU 394 2020-01-19) MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII X-Spam-Status: No, score=-10.6 required=5.0 tests=BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,GIT_PATCH_0,MISSING_MID,SPF_HELO_NONE,SPF_PASS,TXREP 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: Message-ID: <20230301073718.4r2QnpccZ6mchaBvIIpzq3QUnE_uZVmmxxqpT8MF2ds@z> When we end up with a widen-sum with an invariant smaller operand the reduction code uses a wrong vector type for it, causing IL checking ICEs. The following fixes that and the inefficiency of using a widen-sum with a widenend invariant operand as well by actually performing the check the following comment wants. Bootstrapped and tested on x86_64-unknown-linux-gnu, pushed. PR tree-optimization/108950 * tree-vect-patterns.cc (vect_recog_widen_sum_pattern): Check oprnd0 is defined in the loop. * tree-vect-loop.cc (vectorizable_reduction): Record all operands vector types, compute that of invariants and properly update their SLP nodes. * gcc.dg/vect/pr108950.c: New testcase. --- gcc/testsuite/gcc.dg/vect/pr108950.c | 13 +++++++++++++ gcc/tree-vect-loop.cc | 17 +++++++++++------ gcc/tree-vect-patterns.cc | 4 +++- 3 files changed, 27 insertions(+), 7 deletions(-) create mode 100644 gcc/testsuite/gcc.dg/vect/pr108950.c diff --git a/gcc/testsuite/gcc.dg/vect/pr108950.c b/gcc/testsuite/gcc.dg/vect/pr108950.c new file mode 100644 index 00000000000..2163866dfa7 --- /dev/null +++ b/gcc/testsuite/gcc.dg/vect/pr108950.c @@ -0,0 +1,13 @@ +/* { dg-do compile } */ + +int m; +short int n; + +__attribute__ ((simd)) int +foo (void) +{ + m += n; + m += n; +} + +/* { dg-final { scan-tree-dump-not "widen_sum" "vect" } } */ diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc index ab7af0ea3b8..b17e8745d3f 100644 --- a/gcc/tree-vect-loop.cc +++ b/gcc/tree-vect-loop.cc @@ -6790,6 +6790,7 @@ vectorizable_reduction (loop_vec_info loop_vinfo, stmt_vector_for_cost *cost_vec) { tree vectype_in = NULL_TREE; + tree vectype_op[3] = { NULL_TREE, NULL_TREE, NULL_TREE }; class loop *loop = LOOP_VINFO_LOOP (loop_vinfo); enum vect_def_type cond_reduc_dt = vect_unknown_def_type; stmt_vec_info cond_stmt_vinfo = NULL; @@ -6799,7 +6800,6 @@ vectorizable_reduction (loop_vec_info loop_vinfo, bool nested_cycle = false; bool double_reduc = false; int vec_num; - tree tem; tree cr_index_scalar_type = NULL_TREE, cr_index_vector_type = NULL_TREE; tree cond_reduc_val = NULL_TREE; @@ -7037,7 +7037,7 @@ vectorizable_reduction (loop_vec_info loop_vinfo, enum vect_def_type dt; if (!vect_is_simple_use (loop_vinfo, stmt_info, slp_for_stmt_info, i + opno_adjust, &op.ops[i], &slp_op[i], &dt, - &tem, &def_stmt_info)) + &vectype_op[i], &def_stmt_info)) { if (dump_enabled_p ()) dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location, @@ -7052,15 +7052,20 @@ vectorizable_reduction (loop_vec_info loop_vinfo, if (VECTORIZABLE_CYCLE_DEF (dt)) return false; + if (!vectype_op[i]) + vectype_op[i] + = get_vectype_for_scalar_type (loop_vinfo, + TREE_TYPE (op.ops[i]), slp_op[i]); + /* To properly compute ncopies we are interested in the widest non-reduction input type in case we're looking at a widening accumulation that we later handle in vect_transform_reduction. */ if (lane_reduc_code_p - && tem + && vectype_op[i] && (!vectype_in || (GET_MODE_SIZE (SCALAR_TYPE_MODE (TREE_TYPE (vectype_in))) - < GET_MODE_SIZE (SCALAR_TYPE_MODE (TREE_TYPE (tem)))))) - vectype_in = tem; + < GET_MODE_SIZE (SCALAR_TYPE_MODE (TREE_TYPE (vectype_op[i])))))) + vectype_in = vectype_op[i]; if (op.code == COND_EXPR) { @@ -7581,7 +7586,7 @@ vectorizable_reduction (loop_vec_info loop_vinfo, && !lane_reduc_code_p && reduction_type != FOLD_LEFT_REDUCTION)) for (i = 0; i < (int) op.num_ops; i++) - if (!vect_maybe_update_slp_op_vectype (slp_op[i], vectype_in)) + if (!vect_maybe_update_slp_op_vectype (slp_op[i], vectype_op[i])) { if (dump_enabled_p ()) dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location, diff --git a/gcc/tree-vect-patterns.cc b/gcc/tree-vect-patterns.cc index cefe331620f..dd585e59bf7 100644 --- a/gcc/tree-vect-patterns.cc +++ b/gcc/tree-vect-patterns.cc @@ -1821,7 +1821,9 @@ vect_recog_widen_sum_pattern (vec_info *vinfo, of the above pattern. */ if (!vect_reassociating_reduction_p (vinfo, stmt_vinfo, PLUS_EXPR, - &oprnd0, &oprnd1)) + &oprnd0, &oprnd1) + || TREE_CODE (oprnd0) != SSA_NAME + || !vinfo->lookup_def (oprnd0)) return NULL; type = TREE_TYPE (gimple_get_lhs (last_stmt)); -- 2.35.3