From patchwork Wed Jan 5 07:34:41 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Kewen.Lin" X-Patchwork-Id: 1575502 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Authentication-Results: bilbo.ozlabs.org; dkim=pass (1024-bit key; unprotected) header.d=gcc.gnu.org header.i=@gcc.gnu.org header.a=rsa-sha256 header.s=default header.b=btAEHuOu; dkim-atps=neutral Authentication-Results: ozlabs.org; spf=pass (sender SPF authorized) smtp.mailfrom=gcc.gnu.org (client-ip=8.43.85.97; helo=sourceware.org; envelope-from=gcc-patches-bounces+incoming=patchwork.ozlabs.org@gcc.gnu.org; receiver=) Received: from sourceware.org (server2.sourceware.org [8.43.85.97]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (4096 bits) server-digest SHA256) (No client certificate requested) by bilbo.ozlabs.org (Postfix) with ESMTPS id 4JTLqy0JfHz9sXM for ; Wed, 5 Jan 2022 18:35:40 +1100 (AEDT) Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id B1ECA3858C2C for ; Wed, 5 Jan 2022 07:35:36 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org B1ECA3858C2C DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org; s=default; t=1641368136; bh=y9zksjkyJ5rBK2iTuyYbifBnlxS5+b7KtQLrN/uYb5c=; h=To:Subject:Date:List-Id:List-Unsubscribe:List-Archive:List-Post: List-Help:List-Subscribe:From:Reply-To:Cc:From; b=btAEHuOuIFqPVv2GHklkGpiLPcshCIqpv12e88V6rh7INqZuziMfslcQaLPCMkPPJ efbBzPkvli5geXL6/HN5eqhP9C7W4kSKKhNVkFwS0mjcobZg7UlUOqaMofm4OtQesX h92xiRiHuk40rS9n8wkNoD5t38YkgZMaO5BBERLI= X-Original-To: gcc-patches@gcc.gnu.org Delivered-To: gcc-patches@gcc.gnu.org Received: from mx0a-001b2d01.pphosted.com (mx0b-001b2d01.pphosted.com [148.163.158.5]) by sourceware.org (Postfix) with ESMTPS id 978EE3858C2C for ; Wed, 5 Jan 2022 07:34:51 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 978EE3858C2C Received: from pps.filterd (m0098413.ppops.net [127.0.0.1]) by mx0b-001b2d01.pphosted.com (8.16.1.2/8.16.1.2) with SMTP id 2055lfCJ028669; Wed, 5 Jan 2022 07:34:51 GMT Received: from pps.reinject (localhost [127.0.0.1]) by mx0b-001b2d01.pphosted.com with ESMTP id 3dcmawkbfg-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Wed, 05 Jan 2022 07:34:50 +0000 Received: from m0098413.ppops.net (m0098413.ppops.net [127.0.0.1]) by pps.reinject (8.16.0.43/8.16.0.43) with SMTP id 2057Yct9000450; Wed, 5 Jan 2022 07:34:50 GMT Received: from ppma04fra.de.ibm.com (6a.4a.5195.ip4.static.sl-reverse.com [149.81.74.106]) by mx0b-001b2d01.pphosted.com with ESMTP id 3dcmawkbf1-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Wed, 05 Jan 2022 07:34:50 +0000 Received: from pps.filterd (ppma04fra.de.ibm.com [127.0.0.1]) by ppma04fra.de.ibm.com (8.16.1.2/8.16.1.2) with SMTP id 2057X4PC024566; Wed, 5 Jan 2022 07:34:48 GMT Received: from b06cxnps4075.portsmouth.uk.ibm.com (d06relay12.portsmouth.uk.ibm.com [9.149.109.197]) by ppma04fra.de.ibm.com with ESMTP id 3daeka65nf-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Wed, 05 Jan 2022 07:34:48 +0000 Received: from d06av24.portsmouth.uk.ibm.com (mk.ibm.com [9.149.105.60]) by b06cxnps4075.portsmouth.uk.ibm.com (8.14.9/8.14.9/NCO v10.0) with ESMTP id 2057YivF45613328 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-GCM-SHA384 bits=256 verify=OK); Wed, 5 Jan 2022 07:34:44 GMT Received: from d06av24.portsmouth.uk.ibm.com (unknown [127.0.0.1]) by IMSVA (Postfix) with ESMTP id 868EE42041; Wed, 5 Jan 2022 07:34:44 +0000 (GMT) Received: from d06av24.portsmouth.uk.ibm.com (unknown [127.0.0.1]) by IMSVA (Postfix) with ESMTP id 846884203F; Wed, 5 Jan 2022 07:34:42 +0000 (GMT) Received: from kewenlins-mbp.cn.ibm.com (unknown [9.200.147.170]) by d06av24.portsmouth.uk.ibm.com (Postfix) with ESMTP; Wed, 5 Jan 2022 07:34:42 +0000 (GMT) To: GCC Patches Subject: [PATCH v3] rs6000: Fix some issues in rs6000_can_inline_p [PR102059] Message-ID: <85f65aa7-84d3-0e57-ccfb-b95aa24bbe8b@linux.ibm.com> Date: Wed, 5 Jan 2022 15:34:41 +0800 User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:78.0) Gecko/20100101 Thunderbird/78.10.0 Content-Language: en-US X-TM-AS-GCONF: 00 X-Proofpoint-GUID: ifIFjTb9Xhjt7dCROb4jZs4l247V_aUF X-Proofpoint-ORIG-GUID: xNUKmvw4_qDTn8SqRFQlAlnnBGz_orb1 X-Proofpoint-UnRewURL: 0 URL was un-rewritten MIME-Version: 1.0 X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.205,Aquarius:18.0.790,Hydra:6.0.425,FMLib:17.11.62.513 definitions=2022-01-05_02,2022-01-04_01,2021-12-02_01 X-Proofpoint-Spam-Details: rule=outbound_notspam policy=outbound score=0 clxscore=1015 malwarescore=0 impostorscore=0 priorityscore=1501 mlxscore=0 bulkscore=0 lowpriorityscore=0 suspectscore=0 spamscore=0 mlxlogscore=999 adultscore=0 phishscore=0 classifier=spam adjust=0 reason=mlx scancount=1 engine=8.12.0-2110150000 definitions=main-2201050049 X-Spam-Status: No, score=-11.7 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_EF, GIT_PATCH_0, KAM_SHORT, RCVD_IN_MSPIKE_H2, SPF_HELO_NONE, SPF_PASS, TXREP autolearn=ham autolearn_force=no version=3.4.4 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on server2.sourceware.org X-BeenThere: gcc-patches@gcc.gnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Gcc-patches mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-Patchwork-Original-From: "Kewen.Lin via Gcc-patches" From: "Kewen.Lin" Reply-To: "Kewen.Lin" Cc: Michael Meissner , Peter Bergner , Bill Schmidt , David Edelsohn , Segher Boessenkool Errors-To: gcc-patches-bounces+incoming=patchwork.ozlabs.org@gcc.gnu.org Sender: "Gcc-patches" Hi, This patch is to fix the inconsistent behaviors for non-LTO mode and LTO mode. As Martin pointed out, currently the function rs6000_can_inline_p simply makes it inlinable if callee_tree is NULL, but it's unexpected, we should use the command line options from target_option_default_node as default. It replaces rs6000_isa_flags with target_option_default_node when caller_tree is NULL since it's more straightforward and doesn't suffer from some bug not to keep rs6000_isa_flags as default. It also extends the scope of the check for the case that callee has explicit set options, inlining in test case pr102059-5.c can happen unexpectedly before, it's fixed accordingly. As Richi/Mike pointed out, some tuning flags like MASK_P8_FUSION can be neglected for always inlining, this patch also takes some flags when the callee is attributed by always_inline. v1: https://gcc.gnu.org/pipermail/gcc-patches/2021-September/578552.html v2: https://gcc.gnu.org/pipermail/gcc-patches/2021-December/586112.html This patch is one re-post of this updated version[1] and also rebased and adjusted on top of the related commit r12-6219. Bootstrapped and regtested on powerpc64-linux-gnu P8 and powerpc64le-linux-gnu P9 and P10. Is it ok for trunk? [1] https://gcc.gnu.org/pipermail/gcc-patches/2021-December/586296.html BR, Kewen ----- gcc/ChangeLog: PR target/102059 * config/rs6000/rs6000.c (rs6000_can_inline_p): Adjust with target_option_default_node and consider always_inline_safe flags. gcc/testsuite/ChangeLog: PR target/102059 * gcc.target/powerpc/pr102059-4.c: New test. * gcc.target/powerpc/pr102059-5.c: New test. * gcc.target/powerpc/pr102059-6.c: New test. * gcc.target/powerpc/pr102059-7.c: New test. * gcc.target/powerpc/pr102059-8.c: New test. * gcc.dg/lto/pr102059-1_0.c: Remove unneeded option. --- gcc/config/rs6000/rs6000.c | 110 +++++++++++------- gcc/testsuite/gcc.dg/lto/pr102059-1_0.c | 2 +- gcc/testsuite/gcc.target/powerpc/pr102059-4.c | 24 ++++ gcc/testsuite/gcc.target/powerpc/pr102059-5.c | 20 ++++ gcc/testsuite/gcc.target/powerpc/pr102059-6.c | 95 +++++++++++++++ gcc/testsuite/gcc.target/powerpc/pr102059-7.c | 22 ++++ gcc/testsuite/gcc.target/powerpc/pr102059-8.c | 22 ++++ 7 files changed, 255 insertions(+), 40 deletions(-) create mode 100644 gcc/testsuite/gcc.target/powerpc/pr102059-4.c create mode 100644 gcc/testsuite/gcc.target/powerpc/pr102059-5.c create mode 100644 gcc/testsuite/gcc.target/powerpc/pr102059-6.c create mode 100644 gcc/testsuite/gcc.target/powerpc/pr102059-7.c create mode 100644 gcc/testsuite/gcc.target/powerpc/pr102059-8.c diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c index 7d07b47d9e3..60e131f2191 100644 --- a/gcc/config/rs6000/rs6000.c +++ b/gcc/config/rs6000/rs6000.c @@ -25379,55 +25379,87 @@ rs6000_update_ipa_fn_target_info (unsigned int &info, const gimple *stmt) static bool rs6000_can_inline_p (tree caller, tree callee) { - bool ret = false; tree caller_tree = DECL_FUNCTION_SPECIFIC_TARGET (caller); tree callee_tree = DECL_FUNCTION_SPECIFIC_TARGET (callee); - /* If the callee has no option attributes, then it is ok to inline. */ + /* If the caller/callee has option attributes, then use them. + Otherwise, use the command line options. */ if (!callee_tree) - ret = true; - - else - { - HOST_WIDE_INT caller_isa; - struct cl_target_option *callee_opts = TREE_TARGET_OPTION (callee_tree); - HOST_WIDE_INT callee_isa = callee_opts->x_rs6000_isa_flags; - HOST_WIDE_INT explicit_isa = callee_opts->x_rs6000_isa_flags_explicit; + callee_tree = target_option_default_node; + if (!caller_tree) + caller_tree = target_option_default_node; + + struct cl_target_option *caller_opts = TREE_TARGET_OPTION (caller_tree); + struct cl_target_option *callee_opts = TREE_TARGET_OPTION (callee_tree); + HOST_WIDE_INT caller_isa = caller_opts->x_rs6000_isa_flags; + HOST_WIDE_INT callee_isa = callee_opts->x_rs6000_isa_flags; + + bool always_inline + = DECL_DISREGARD_INLINE_LIMITS (callee) + && lookup_attribute ("always_inline", DECL_ATTRIBUTES (callee)); + + /* Some features can be tolerated for always inlines. */ + unsigned HOST_WIDE_INT always_inline_safe_mask + /* Fusion option masks. */ + = OPTION_MASK_P8_FUSION | OPTION_MASK_P10_FUSION + | OPTION_MASK_P8_FUSION_SIGN | OPTION_MASK_P10_FUSION + | OPTION_MASK_P10_FUSION_LD_CMPI | OPTION_MASK_P10_FUSION_2LOGICAL + | OPTION_MASK_P10_FUSION_LOGADD | OPTION_MASK_P10_FUSION_ADDLOG + | OPTION_MASK_P10_FUSION_2ADD + /* Like fusion, some option masks which are just for optimization. */ + | OPTION_MASK_SAVE_TOC_INDIRECT | OPTION_MASK_PCREL_OPT; + + /* Some features are totally safe for inlining (or always inlines), + let's exclude them from the following checkings. */ + HOST_WIDE_INT safe_mask = always_inline ? always_inline_safe_mask : 0; + cgraph_node *callee_node = cgraph_node::get (callee); + if (ipa_fn_summaries && ipa_fn_summaries->get (callee_node) != NULL) + { + unsigned int info = ipa_fn_summaries->get (callee_node)->target_info; + if ((info & RS6000_FN_TARGET_INFO_HTM) == 0) + safe_mask |= OPTION_MASK_HTM; + } + + callee_isa &= ~safe_mask; + + /* The callee's options must be a subset of the caller's options, i.e. + a vsx function may inline an altivec function, but a no-vsx function + must not inline a vsx function. */ + if ((caller_isa & callee_isa) != callee_isa) + { + if (TARGET_DEBUG_TARGET) + fprintf (stderr, + "rs6000_can_inline_p:, caller %s, callee %s, cannot " + "inline since callee's options set isn't a subset of " + "caller's options set.\n", + get_decl_name (caller), get_decl_name (callee)); + return false; + } - /* If the caller has option attributes, then use them. - Otherwise, use the command line options. */ - if (caller_tree) - caller_isa = TREE_TARGET_OPTION (caller_tree)->x_rs6000_isa_flags; - else - caller_isa = rs6000_isa_flags; + /* For those options that the callee has explicitly enabled or disabled, + then we must enforce that the callee's and caller's options match + exactly; see PR70010. */ + HOST_WIDE_INT explicit_isa = callee_opts->x_rs6000_isa_flags_explicit; + explicit_isa &= ~safe_mask; - cgraph_node *callee_node = cgraph_node::get (callee); - if (ipa_fn_summaries && ipa_fn_summaries->get (callee_node) != NULL) - { - unsigned int info = ipa_fn_summaries->get (callee_node)->target_info; - if ((info & RS6000_FN_TARGET_INFO_HTM) == 0) - { - callee_isa &= ~OPTION_MASK_HTM; - explicit_isa &= ~OPTION_MASK_HTM; - } - } - - /* The callee's options must be a subset of the caller's options, i.e. - a vsx function may inline an altivec function, but a no-vsx function - must not inline a vsx function. However, for those options that the - callee has explicitly enabled or disabled, then we must enforce that - the callee's and caller's options match exactly; see PR70010. */ - if (((caller_isa & callee_isa) == callee_isa) - && (caller_isa & explicit_isa) == (callee_isa & explicit_isa)) - ret = true; + if ((caller_isa & explicit_isa) != (callee_isa & explicit_isa)) + { + if (TARGET_DEBUG_TARGET) + fprintf (stderr, + "rs6000_can_inline_p:, caller %s, callee %s, cannot " + "inline since callee's options set isn't a subset of " + "caller's options set by considering callee's " + "explicitly set options.\n", + get_decl_name (caller), get_decl_name (callee)); + return false; } if (TARGET_DEBUG_TARGET) - fprintf (stderr, "rs6000_can_inline_p:, caller %s, callee %s, %s inline\n", - get_decl_name (caller), get_decl_name (callee), - (ret ? "can" : "cannot")); + fprintf (stderr, + "rs6000_can_inline_p:, caller %s, callee %s, can inline.\n", + get_decl_name (caller), get_decl_name (callee)); - return ret; + return true; } /* Allocate a stack temp and fixup the address so it meets the particular diff --git a/gcc/testsuite/gcc.dg/lto/pr102059-1_0.c b/gcc/testsuite/gcc.dg/lto/pr102059-1_0.c index e1004f5cfbf..b215b701097 100644 --- a/gcc/testsuite/gcc.dg/lto/pr102059-1_0.c +++ b/gcc/testsuite/gcc.dg/lto/pr102059-1_0.c @@ -1,7 +1,7 @@ /* { dg-lto-do link } */ /* { dg-skip-if "power10 and above only" { ! { power10_ok } } } */ /* -Wno-attributes suppresses always_inline warnings. */ -/* { dg-lto-options { "-O2 -mdejagnu-cpu=power8 -flto -Wno-attributes -mno-power8-fusion" } } */ +/* { dg-lto-options { "-O2 -mdejagnu-cpu=power8 -flto -Wno-attributes" } } */ int __attribute__ ((always_inline)) foo1 (int *b) diff --git a/gcc/testsuite/gcc.target/powerpc/pr102059-4.c b/gcc/testsuite/gcc.target/powerpc/pr102059-4.c new file mode 100644 index 00000000000..d2a002cf141 --- /dev/null +++ b/gcc/testsuite/gcc.target/powerpc/pr102059-4.c @@ -0,0 +1,24 @@ +/* { dg-do compile } */ +/* { dg-require-effective-target powerpc_htm_ok } */ +/* { dg-options "-O2 -mdejagnu-cpu=power8" } */ + +/* Verify it emits inlining error msg at non-LTO mode. */ + +#include + +static inline int __attribute__ ((always_inline)) +foo (int *b) /* { dg-error "inlining failed in call to .* target specific option mismatch" } */ +{ + int res = _HTM_STATE(__builtin_ttest()); + *b += res; + return *b; +} + +#pragma GCC target "cpu=power10" +int +bar (int *a) +{ + *a = foo (a); /* { dg-message "called from here" } */ + return 0; +} + diff --git a/gcc/testsuite/gcc.target/powerpc/pr102059-5.c b/gcc/testsuite/gcc.target/powerpc/pr102059-5.c new file mode 100644 index 00000000000..1d5d6c38bf3 --- /dev/null +++ b/gcc/testsuite/gcc.target/powerpc/pr102059-5.c @@ -0,0 +1,20 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -mno-vsx" } */ + +/* Verify it emits inlining error msg when the callee has explicit + disabling option from command line. */ + +vector int c, a, b; + +static inline void __attribute__ ((__always_inline__)) +foo () /* { dg-error "inlining failed in call to .* target specific option mismatch" } */ +{ + c = a + b; +} + +__attribute__ ((target ("vsx"))) +int main () +{ + foo (); /* { dg-message "called from here" } */ + c = a + b; +} diff --git a/gcc/testsuite/gcc.target/powerpc/pr102059-6.c b/gcc/testsuite/gcc.target/powerpc/pr102059-6.c new file mode 100644 index 00000000000..9684cab986a --- /dev/null +++ b/gcc/testsuite/gcc.target/powerpc/pr102059-6.c @@ -0,0 +1,95 @@ +/* { dg-do compile } */ +/* -Wno-attributes suppresses always_inline warnings. */ +/* { dg-options "-O2 -mdejagnu-cpu=power8 -Wno-attributes" } */ + +/* Verify it doesn't emit inlining error msg since some mismatched + features are considered as safe for always_inline. */ + +/* 1. Callee enables Power8 fusion implicitly, while caller + with Power9 doesn't support power8 fusion at all. */ + +__attribute__ ((always_inline)) +int callee1 (int *b) +{ + *b += 1; + return *b; +} + +#pragma GCC target "cpu=power9" +int caller1 (int *a) +{ + *a = callee1 (a); + return 0; +} + +/* 2. Caller enables indirect toc save feature while callee + disables it explicitly. */ + +#pragma GCC target "save-toc-indirect" +__attribute__ ((always_inline)) +int callee2 (int *b) +{ + *b += 2; + return *b; +} + +#pragma GCC target "no-save-toc-indirect" +int caller2 (int *a) +{ + *a = callee2 (a); + return 0; +} + +/* 3. Caller disables Power10 fusion explicitly, while callee + still supports it as Power10 turns it on by default. */ + +#pragma GCC target "cpu=power10" +__attribute__ ((always_inline)) +int callee3 (int *b) +{ + *b += 3; + return *b; +} + +#pragma GCC target "cpu=power10,no-power10-fusion" +int caller3 (int *a) +{ + *a = callee3 (a); + return 0; +} + +/* 4. Caller enables Power10 fusion implicitly, while callee + disables it explicitly. */ + +#pragma GCC target "no-power10-fusion" +__attribute__ ((always_inline)) +int callee4 (int *b) +{ + *b += 4; + return *b; +} + +#pragma GCC target "cpu=power10" +int caller4 (int *a) +{ + *a = callee4 (a); + return 0; +} + +/* 5. Caller disables pcrel-opt while callee enables it explicitly. */ + +#pragma GCC target "cpu=power10,no-pcrel-opt" +__attribute__ ((always_inline)) +int callee5 (int *b) +{ + *b += 5; + return *b; +} + +#pragma GCC target "cpu=power10,pcrel-opt" +int caller5 (int *a) +{ + *a = callee5 (a); + return 0; +} + diff --git a/gcc/testsuite/gcc.target/powerpc/pr102059-7.c b/gcc/testsuite/gcc.target/powerpc/pr102059-7.c new file mode 100644 index 00000000000..0f27f2ce7d7 --- /dev/null +++ b/gcc/testsuite/gcc.target/powerpc/pr102059-7.c @@ -0,0 +1,22 @@ +/* { dg-do compile } */ +/* -Wno-attributes suppresses always_inline warnings. */ +/* { dg-options "-O2 -mdejagnu-cpu=power8 -mno-power8-fusion -Wno-attributes" } */ + +/* Verify it doesn't emit inlining error msg since the flag power8 + fusion is considered as safe for always_inline, it's still safe + even the flag is set explicitly. */ + +__attribute__ ((always_inline)) +int foo (int *b) +{ + *b += 10; + return *b; +} + +#pragma GCC target "power8-fusion" +int bar (int *a) +{ + *a = foo (a); + return 0; +} + diff --git a/gcc/testsuite/gcc.target/powerpc/pr102059-8.c b/gcc/testsuite/gcc.target/powerpc/pr102059-8.c new file mode 100644 index 00000000000..826ce88025f --- /dev/null +++ b/gcc/testsuite/gcc.target/powerpc/pr102059-8.c @@ -0,0 +1,22 @@ +/* { dg-do compile } */ +/* { dg-require-effective-target powerpc_htm_ok } */ +/* { dg-options "-O2 -mdejagnu-cpu=power8 -mhtm" } */ + +/* Verify the inlining won't perform when the callee requires + some target feature which isn't supported by caller, even + if the callee doesn't have any target attributes or pragmas. + If the inlining performs here, the compilation will fail. */ + +int +foo (int *b) +{ + *b += __builtin_ttest (); + return *b; +} + +__attribute__ ((target ("no-htm"))) int +bar (int *a) +{ + *a = foo (a); + return 0; +}