From patchwork Fri Feb 23 11:11:24 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Ajit Agarwal X-Patchwork-Id: 1903232 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@legolas.ozlabs.org Authentication-Results: legolas.ozlabs.org; dkim=pass (2048-bit key; unprotected) header.d=ibm.com header.i=@ibm.com header.a=rsa-sha256 header.s=pp1 header.b=LHfgxDeX; dkim-atps=neutral Authentication-Results: legolas.ozlabs.org; spf=pass (sender SPF authorized) smtp.mailfrom=gcc.gnu.org (client-ip=8.43.85.97; helo=server2.sourceware.org; envelope-from=gcc-patches-bounces+incoming=patchwork.ozlabs.org@gcc.gnu.org; receiver=patchwork.ozlabs.org) Received: from server2.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 ECDSA (secp384r1) server-digest SHA384) (No client certificate requested) by legolas.ozlabs.org (Postfix) with ESMTPS id 4Th6xY6nhnz23pm for ; Fri, 23 Feb 2024 22:20:17 +1100 (AEDT) Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id E9165385840C for ; Fri, 23 Feb 2024 11:20:15 +0000 (GMT) X-Original-To: gcc-patches@gcc.gnu.org Delivered-To: gcc-patches@gcc.gnu.org Received: from mx0a-001b2d01.pphosted.com (mx0a-001b2d01.pphosted.com [148.163.156.1]) by sourceware.org (Postfix) with ESMTPS id 203473858289 for ; Fri, 23 Feb 2024 11:19:48 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org 203473858289 Authentication-Results: sourceware.org; dmarc=none (p=none dis=none) header.from=linux.ibm.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=linux.ibm.com ARC-Filter: OpenARC Filter v1.0.0 sourceware.org 203473858289 Authentication-Results: server2.sourceware.org; arc=none smtp.remote-ip=148.163.156.1 ARC-Seal: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1708687195; cv=none; b=ppjUpDaYcuu9twIueD+rFgm9LPpWuGjf8OK6YDlqXlE33MuKK7I7sfEJ0DzWML1vRvjhWFfttuwXtL/LHErDPSFpfEPCKolDiA9IWNLXKLTFTztJpgZSUTtmJ58t5R8hfvQx+1tVZM5qm2g6tMDUcoaDI2ZjtySDlWUA4O3HWaU= ARC-Message-Signature: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1708687195; c=relaxed/simple; bh=zEUsnO2Uvov/oe1SteisHRKnLqzJfL2h8T2i983ZILU=; h=DKIM-Signature:Message-ID:Date:MIME-Version:To:From:Subject; b=blZJYkX4nqoSc3Wm1QkO4Wf05Pa3Vuy969t9TbTDSyYWD39P1Kbvr1pEasgDlgdN0gvYjqW6PsuAZ6R/asCjgfZb63c+JpbB5EZqehY8JlsE8OOGj+bpmShBbFOZYDkby7FMCPIGDc6M31z42NmhD8CfIb8Z6L36olxEIG72EEY= ARC-Authentication-Results: i=1; server2.sourceware.org Received: from pps.filterd (m0356517.ppops.net [127.0.0.1]) by mx0a-001b2d01.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 41NBHa4O018110; Fri, 23 Feb 2024 11:19:42 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=ibm.com; h=message-id : date : mime-version : to : from : subject : content-type : content-transfer-encoding; s=pp1; bh=ObPx2I4eYrEc54nqlEzlpwgUhwnUo8vgm0M8AReDzy0=; b=LHfgxDeXLgfnCjCCas6/J2msse0ZdyMgoeWM1mKQ/qSHwyB1Csi9fKrcPp0acQaypDnz feRmmVNIFpv5VxS3IIZIxowa87LOFLncs0cp9nkuY2U43n+nmMhJ+qi5z8AWnE5lJty1 3y/pHp/TYirF8msIkKb4/O8Kqum3xojiHuWNG6+Ww2LRR3Bywj313OxJ1pnbP3WIY+sz oCSZFa/Ne2u4HHUlor1Przv03qef9rLg5QEIWxS0AnDoac3QDiWZC+C/xfm6lXzK4mdQ cHNWIDWIzPOoNqlo3X5vRT5YHrqnBTMHAZ6DLaNUKIsoF2olXQcwBXIVmsjsW46OVbKm JQ== Received: from pps.reinject (localhost [127.0.0.1]) by mx0a-001b2d01.pphosted.com (PPS) with ESMTPS id 3wetbs81am-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Fri, 23 Feb 2024 11:19:42 +0000 Received: from m0356517.ppops.net (m0356517.ppops.net [127.0.0.1]) by pps.reinject (8.17.1.5/8.17.1.5) with ESMTP id 41NBIZXf020526; Fri, 23 Feb 2024 11:19:42 GMT Received: from ppma22.wdc07v.mail.ibm.com (5c.69.3da9.ip4.static.sl-reverse.com [169.61.105.92]) by mx0a-001b2d01.pphosted.com (PPS) with ESMTPS id 3wetbs818c-4 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Fri, 23 Feb 2024 11:19:41 +0000 Received: from pps.filterd (ppma22.wdc07v.mail.ibm.com [127.0.0.1]) by ppma22.wdc07v.mail.ibm.com (8.17.1.19/8.17.1.19) with ESMTP id 41NAKYkQ013532; Fri, 23 Feb 2024 11:11:35 GMT Received: from smtprelay01.wdc07v.mail.ibm.com ([172.16.1.68]) by ppma22.wdc07v.mail.ibm.com (PPS) with ESMTPS id 3wb7h0w5m5-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Fri, 23 Feb 2024 11:11:35 +0000 Received: from smtpav05.wdc07v.mail.ibm.com (smtpav05.wdc07v.mail.ibm.com [10.39.53.232]) by smtprelay01.wdc07v.mail.ibm.com (8.14.9/8.14.9/NCO v10.0) with ESMTP id 41NBBWrk23527982 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-GCM-SHA384 bits=256 verify=OK); Fri, 23 Feb 2024 11:11:34 GMT Received: from smtpav05.wdc07v.mail.ibm.com (unknown [127.0.0.1]) by IMSVA (Postfix) with ESMTP id 5F99058070; Fri, 23 Feb 2024 11:11:30 +0000 (GMT) Received: from smtpav05.wdc07v.mail.ibm.com (unknown [127.0.0.1]) by IMSVA (Postfix) with ESMTP id 6FC3F58066; Fri, 23 Feb 2024 11:11:26 +0000 (GMT) Received: from [9.43.83.218] (unknown [9.43.83.218]) by smtpav05.wdc07v.mail.ibm.com (Postfix) with ESMTP; Fri, 23 Feb 2024 11:11:26 +0000 (GMT) Message-ID: <6a013cf5-a5f2-44df-bdb3-a4985fbae06a@linux.ibm.com> Date: Fri, 23 Feb 2024 16:41:24 +0530 MIME-Version: 1.0 User-Agent: Mozilla Thunderbird Content-Language: en-US To: Richard Sandiford , Alex Coplan , "Kewen.Lin" , Michael Meissner , Segher Boessenkool , Peter Bergner , David Edelsohn , gcc-patches From: Ajit Agarwal Subject: [PATCH V3 0/2] aarch64: Place target independent and dependent changed code in one file. X-TM-AS-GCONF: 00 X-Proofpoint-GUID: oZMUdxZHo_jU4Nvhec52PZnJxNRXzn8G X-Proofpoint-ORIG-GUID: XymVtv3F2X38XyYlflKHXow_iC6CQIk3 X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.272,Aquarius:18.0.1011,Hydra:6.0.619,FMLib:17.11.176.26 definitions=2024-02-22_15,2024-02-23_01,2023-05-22_02 X-Proofpoint-Spam-Details: rule=outbound_notspam policy=outbound score=0 clxscore=1015 phishscore=0 lowpriorityscore=0 mlxlogscore=999 suspectscore=0 adultscore=0 malwarescore=0 priorityscore=1501 spamscore=0 bulkscore=0 impostorscore=0 mlxscore=0 classifier=spam adjust=0 reason=mlx scancount=1 engine=8.12.0-2311290000 definitions=main-2402230080 X-Spam-Status: No, score=-12.6 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_EF, GIT_PATCH_0, KAM_MANYTO, RCVD_IN_MSPIKE_H3, RCVD_IN_MSPIKE_WL, SPF_HELO_NONE, SPF_PASS, 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 X-BeenThere: gcc-patches@gcc.gnu.org X-Mailman-Version: 2.1.30 Precedence: list List-Id: Gcc-patches mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: gcc-patches-bounces+incoming=patchwork.ozlabs.org@gcc.gnu.org Hello Richard/Alex/Segher: This patch adds the changed code for target independent and dependent code for load store fusion. Common infrastructure of load store pair fusion is divided into target independent and target dependent changed code. Target independent code is the Generic code with pure virtual function to interface betwwen target independent and dependent code. Target dependent code is the implementation of pure virtual function for aarch64 target and the call to target independent code. Bootstrapped for aarch64-linux-gnu. Thanks & Regards Ajit aarch64: Place target independent and dependent changed code in one file. Common infrastructure of load store pair fusion is divided into target independent and target dependent changed code. Target independent code is the Generic code with pure virtual function to interface betwwen target independent and dependent code. Target dependent code is the implementation of pure virtual function for aarch64 target and the call to target independent code. 2024-02-23 Ajit Kumar Agarwal gcc/ChangeLog: * config/aarch64/aarch64-ldp-fusion.cc: Place target independent and dependent changed code. --- gcc/config/aarch64/aarch64-ldp-fusion.cc | 437 ++++++++++++++++------- 1 file changed, 305 insertions(+), 132 deletions(-) diff --git a/gcc/config/aarch64/aarch64-ldp-fusion.cc b/gcc/config/aarch64/aarch64-ldp-fusion.cc index 22ed95eb743..2ef22ff1e96 100644 --- a/gcc/config/aarch64/aarch64-ldp-fusion.cc +++ b/gcc/config/aarch64/aarch64-ldp-fusion.cc @@ -40,10 +40,10 @@ using namespace rtl_ssa; -static constexpr HOST_WIDE_INT LDP_IMM_BITS = 7; -static constexpr HOST_WIDE_INT LDP_IMM_SIGN_BIT = (1 << (LDP_IMM_BITS - 1)); -static constexpr HOST_WIDE_INT LDP_MAX_IMM = LDP_IMM_SIGN_BIT - 1; -static constexpr HOST_WIDE_INT LDP_MIN_IMM = -LDP_MAX_IMM - 1; +static constexpr HOST_WIDE_INT PAIR_MEM_IMM_BITS = 7; +static constexpr HOST_WIDE_INT PAIR_MEM_IMM_SIGN_BIT = (1 << (PAIR_MEM_IMM_BITS - 1)); +static constexpr HOST_WIDE_INT PAIR_MEM_MAX_IMM = PAIR_MEM_IMM_SIGN_BIT - 1; +static constexpr HOST_WIDE_INT PAIR_MEM_MIN_IMM = -PAIR_MEM_MAX_IMM - 1; // We pack these fields (load_p, fpsimd_p, and size) into an integer // (LFS) which we use as part of the key into the main hash tables. @@ -138,8 +138,18 @@ struct alt_base poly_int64 offset; }; +// Virtual base class for load/store walkers used in alias analysis. +struct alias_walker +{ + virtual bool conflict_p (int &budget) const = 0; + virtual insn_info *insn () const = 0; + virtual bool valid () const = 0; + virtual void advance () = 0; +}; + + // State used by the pass for a given basic block. -struct ldp_bb_info +struct pair_fusion { using def_hash = nofree_ptr_hash; using expr_key_t = pair_hash>; @@ -161,13 +171,13 @@ struct ldp_bb_info static const size_t obstack_alignment = sizeof (void *); bb_info *m_bb; - ldp_bb_info (bb_info *bb) : m_bb (bb), m_emitted_tombstone (false) + pair_fusion (bb_info *bb) : m_bb (bb), m_emitted_tombstone (false) { obstack_specify_allocation (&m_obstack, OBSTACK_CHUNK_SIZE, obstack_alignment, obstack_chunk_alloc, obstack_chunk_free); } - ~ldp_bb_info () + ~pair_fusion () { obstack_free (&m_obstack, nullptr); @@ -177,10 +187,50 @@ struct ldp_bb_info bitmap_obstack_release (&m_bitmap_obstack); } } + void track_access (insn_info *, bool load, rtx mem); + void transform (); + void cleanup_tombstones (); + virtual void set_multiword_subreg (insn_info *i1, insn_info *i2, + bool load_p) = 0; + virtual rtx gen_load_store_pair (rtx *pats, rtx writeback, + bool load_p) = 0; + void merge_pairs (insn_list_t &, insn_list_t &, + bool load_p, unsigned access_size); + virtual void transform_for_base (int load_size, access_group &group) = 0; + + bool try_fuse_pair (bool load_p, unsigned access_size, + insn_info *i1, insn_info *i2); + + bool fuse_pair (bool load_p, unsigned access_size, + int writeback, + insn_info *i1, insn_info *i2, + base_cand &base, + const insn_range_info &move_range); + + void do_alias_analysis (insn_info *alias_hazards[4], + alias_walker *walkers[4], + bool load_p); + + void track_tombstone (int uid); + + bool track_via_mem_expr (insn_info *, rtx mem, lfs_fields lfs); + + virtual bool is_fpsimd_op_p (rtx reg_op, machine_mode mem_mode, + bool load_p) = 0; + + virtual bool pair_operand_mode_ok_p (machine_mode mode) = 0; + virtual bool pair_trailing_writeback_p () = 0; + virtual bool pair_check_register_operand (bool load_p, rtx reg_op, + machine_mode mem_mode) = 0; + virtual int pair_mem_alias_check_limit () = 0; + virtual bool pair_is_writeback () = 0 ; + virtual bool pair_mem_ok_policy (rtx first_mem, bool load_p, + machine_mode mode) = 0; + virtual bool fuseable_store_p (insn_info *i1, insn_info *i2) = 0; + virtual bool fuseable_load_p (insn_info *info) = 0; - inline void track_access (insn_info *, bool load, rtx mem); - inline void transform (); - inline void cleanup_tombstones (); + template + void traverse_base_map (Map &map); private: obstack m_obstack; @@ -191,30 +241,157 @@ private: bool m_emitted_tombstone; inline splay_tree_node *node_alloc (access_record *); +}; - template - inline void traverse_base_map (Map &map); - inline void transform_for_base (int load_size, access_group &group); - - inline void merge_pairs (insn_list_t &, insn_list_t &, - bool load_p, unsigned access_size); +struct aarch64_pair_fusion : public pair_fusion +{ +public: + aarch64_pair_fusion (bb_info *bb) : pair_fusion (bb) {}; + bool is_fpsimd_op_p (rtx reg_op, machine_mode mem_mode, bool load_p) + { + const bool fpsimd_op_p + = reload_completed + ? (REG_P (reg_op) && FP_REGNUM_P (REGNO (reg_op))) + : (GET_MODE_CLASS (mem_mode) != MODE_INT + && (load_p || !aarch64_const_zero_rtx_p (reg_op))); + return fpsimd_op_p; + } - inline bool try_fuse_pair (bool load_p, unsigned access_size, - insn_info *i1, insn_info *i2); + bool pair_mem_ok_policy (rtx first_mem, bool load_p, machine_mode mode) + { + return aarch64_mem_ok_with_ldpstp_policy_model (first_mem, + load_p, + mode); + } + bool pair_operand_mode_ok_p (machine_mode mode); - inline bool fuse_pair (bool load_p, unsigned access_size, - int writeback, - insn_info *i1, insn_info *i2, - base_cand &base, - const insn_range_info &move_range); + void transform_for_base (int encoded_lfs, + access_group &group); + rtx gen_load_store_pair (rtx *pats, + rtx writeback, + bool load_p) + { + rtx pair_pat; - inline void track_tombstone (int uid); + if (writeback) + { + auto patvec = gen_rtvec (3, writeback, pats[0], pats[1]); + pair_pat = gen_rtx_PARALLEL (VOIDmode, patvec); + } + else if (load_p) + pair_pat = aarch64_gen_load_pair (XEXP (pats[0], 0), + XEXP (pats[1], 0), + XEXP (pats[0], 1)); + else + pair_pat = aarch64_gen_store_pair (XEXP (pats[0], 0), + XEXP (pats[0], 1), + XEXP (pats[1], 1)); + return pair_pat; + } - inline bool track_via_mem_expr (insn_info *, rtx mem, lfs_fields lfs); + void set_multiword_subreg (insn_info *i1, insn_info *i2, bool load_p) + { + if (i1 || i2 || load_p) + return; + return; + } + bool pair_trailing_writeback_p () + { + return aarch64_ldp_writeback > 1; + } + bool pair_check_register_operand (bool load_p, rtx reg_op, machine_mode mem_mode) + { + return (load_p + ? aarch64_ldp_reg_operand (reg_op, mem_mode) + : aarch64_stp_reg_operand (reg_op, mem_mode)); + } + int pair_mem_alias_check_limit () + { + return aarch64_ldp_alias_check_limit; + } + bool fuseable_store_p (insn_info *i1, insn_info *i2) { return i1 || i2;} + bool fuseable_load_p (insn_info *insn) { return insn;} + bool pair_is_writeback () + { + return !aarch64_ldp_writeback; + } }; +bool +store_modifies_mem_p (rtx mem, insn_info *store_insn, int &budget); +bool load_modified_by_store_p (insn_info *load, + insn_info *store, + int &budget); +extern insn_info * +try_repurpose_store (insn_info *first, + insn_info *second, + const insn_range_info &move_range); + +void reset_debug_use (use_info *use); + +extern void +fixup_debug_uses (obstack_watermark &attempt, + insn_info *insns[2], + rtx orig_rtl[2], + insn_info *pair_dst, + insn_info *trailing_add, + bool load_p, + int writeback, + rtx writeback_effect, + unsigned base_regno); + +void +fixup_debug_uses_trailing_add (obstack_watermark &attempt, + insn_info *pair_dst, + insn_info *trailing_add, + rtx writeback_effect); + + +extern void +fixup_debug_use (obstack_watermark &attempt, + use_info *use, + def_info *def, + rtx base, + poly_int64 wb_offset); + +extern insn_info * +find_trailing_add (insn_info *insns[2], + const insn_range_info &pair_range, + int initial_writeback, + rtx *writeback_effect, + def_info **add_def, + def_info *base_def, + poly_int64 initial_offset, + unsigned access_size); + +rtx drop_writeback (rtx mem); +rtx pair_mem_strip_offset (rtx mem, poly_int64 *offset); +bool any_pre_modify_p (rtx x); +bool any_post_modify_p (rtx x); +int encode_lfs (lfs_fields fields); +extern insn_info * latest_hazard_before (insn_info *insn, rtx *ignore, + insn_info *ignore_insn = nullptr); +insn_info * first_hazard_after (insn_info *insn, rtx *ignore); +bool ranges_overlap_p (const insn_range_info &r1, const insn_range_info &r2); +insn_range_info get_def_range (def_info *def); +insn_range_info def_downwards_move_range (def_info *def); +insn_range_info def_upwards_move_range (def_info *def); +rtx gen_tombstone (void); +rtx filter_notes (rtx note, rtx result, bool *eh_region, rtx *fr_expr); +rtx combine_reg_notes (insn_info *i1, insn_info *i2, bool load_p); +rtx extract_writebacks (bool load_p, rtx pats[2], int changed); +void do_alias_analysis (insn_info *alias_hazards[4], + alias_walker *walkers[4], + bool load_p); +int get_viable_bases (insn_info *insns[2], + vec &base_cands, + rtx cand_mems[2], + unsigned access_size, + bool reversed); +void dump_insn_list (FILE *f, const insn_list_t &l); + splay_tree_node * -ldp_bb_info::node_alloc (access_record *access) +pair_fusion::node_alloc (access_record *access) { using T = splay_tree_node; void *addr = obstack_alloc (&m_obstack, sizeof (T)); @@ -224,7 +401,7 @@ ldp_bb_info::node_alloc (access_record *access) // Given a mem MEM, if the address has side effects, return a MEM that accesses // the same address but without the side effects. Otherwise, return // MEM unchanged. -static rtx +rtx drop_writeback (rtx mem) { rtx addr = XEXP (mem, 0); @@ -261,8 +438,8 @@ drop_writeback (rtx mem) // Convenience wrapper around strip_offset that can also look through // RTX_AUTOINC addresses. The interface is like strip_offset except we take a // MEM so that we know the mode of the access. -static rtx -ldp_strip_offset (rtx mem, poly_int64 *offset) +rtx +pair_mem_strip_offset (rtx mem, poly_int64 *offset) { rtx addr = XEXP (mem, 0); @@ -295,7 +472,7 @@ ldp_strip_offset (rtx mem, poly_int64 *offset) } // Return true if X is a PRE_{INC,DEC,MODIFY} rtx. -static bool +bool any_pre_modify_p (rtx x) { const auto code = GET_CODE (x); @@ -303,7 +480,7 @@ any_pre_modify_p (rtx x) } // Return true if X is a POST_{INC,DEC,MODIFY} rtx. -static bool +bool any_post_modify_p (rtx x) { const auto code = GET_CODE (x); @@ -332,9 +509,15 @@ ldp_operand_mode_ok_p (machine_mode mode) return reload_completed || mode != TImode; } +bool +aarch64_pair_fusion::pair_operand_mode_ok_p (machine_mode mode) +{ + return (ldp_operand_mode_ok_p (mode)); +} + // Given LFS (load_p, fpsimd_p, size) fields in FIELDS, encode these // into an integer for use as a hash table key. -static int +int encode_lfs (lfs_fields fields) { int size_log2 = exact_log2 (fields.size); @@ -396,7 +579,7 @@ access_group::track (Alloc alloc_node, poly_int64 offset, insn_info *insn) // MEM_EXPR base (i.e. a tree decl) relative to which we can track the access. // LFS is used as part of the key to the hash table, see track_access. bool -ldp_bb_info::track_via_mem_expr (insn_info *insn, rtx mem, lfs_fields lfs) +pair_fusion::track_via_mem_expr (insn_info *insn, rtx mem, lfs_fields lfs) { if (!MEM_EXPR (mem) || !MEM_OFFSET_KNOWN_P (mem)) return false; @@ -412,7 +595,7 @@ ldp_bb_info::track_via_mem_expr (insn_info *insn, rtx mem, lfs_fields lfs) const machine_mode mem_mode = GET_MODE (mem); const HOST_WIDE_INT mem_size = GET_MODE_SIZE (mem_mode).to_constant (); - // Punt on misaligned offsets. LDP/STP instructions require offsets to be a + // Punt on misaligned offsets. PAIR MEM instructions require offsets to be a // multiple of the access size, and we believe that misaligned offsets on // MEM_EXPR bases are likely to lead to misaligned offsets w.r.t. RTL bases. if (!multiple_p (offset, mem_size)) @@ -438,46 +621,38 @@ ldp_bb_info::track_via_mem_expr (insn_info *insn, rtx mem, lfs_fields lfs) } // Main function to begin pair discovery. Given a memory access INSN, -// determine whether it could be a candidate for fusing into an ldp/stp, +// determine whether it could be a candidate for fusing into an pair mem, // and if so, track it in the appropriate data structure for this basic // block. LOAD_P is true if the access is a load, and MEM is the mem // rtx that occurs in INSN. void -ldp_bb_info::track_access (insn_info *insn, bool load_p, rtx mem) +pair_fusion::track_access (insn_info *insn, bool load_p, rtx mem) { // We can't combine volatile MEMs, so punt on these. if (MEM_VOLATILE_P (mem)) return; - // Ignore writeback accesses if the param says to do so. - if (!aarch64_ldp_writeback + // Ignore writeback accesses if the param says to do so + if (pair_is_writeback () && GET_RTX_CLASS (GET_CODE (XEXP (mem, 0))) == RTX_AUTOINC) return; const machine_mode mem_mode = GET_MODE (mem); - if (!ldp_operand_mode_ok_p (mem_mode)) + + if (!pair_operand_mode_ok_p (mem_mode)) return; rtx reg_op = XEXP (PATTERN (insn->rtl ()), !load_p); - // Ignore the access if the register operand isn't suitable for ldp/stp. - if (load_p - ? !aarch64_ldp_reg_operand (reg_op, mem_mode) - : !aarch64_stp_reg_operand (reg_op, mem_mode)) + if (!pair_check_register_operand (load_p, reg_op, mem_mode)) return; - // We want to segregate FP/SIMD accesses from GPR accesses. // // Before RA, we use the modes, noting that stores of constant zero // operands use GPRs (even in non-integer modes). After RA, we use // the hard register numbers. - const bool fpsimd_op_p - = reload_completed - ? (REG_P (reg_op) && FP_REGNUM_P (REGNO (reg_op))) - : (GET_MODE_CLASS (mem_mode) != MODE_INT - && (load_p || !aarch64_const_zero_rtx_p (reg_op))); - - // Note ldp_operand_mode_ok_p already rejected VL modes. + const bool fpsimd_op_p = is_fpsimd_op_p (reg_op, mem_mode, load_p); + // Note pair_operand_mode_ok_p already rejected VL modes. const HOST_WIDE_INT mem_size = GET_MODE_SIZE (mem_mode).to_constant (); const lfs_fields lfs = { load_p, fpsimd_op_p, mem_size }; @@ -487,7 +662,7 @@ ldp_bb_info::track_access (insn_info *insn, bool load_p, rtx mem) poly_int64 mem_off; rtx addr = XEXP (mem, 0); const bool autoinc_p = GET_RTX_CLASS (GET_CODE (addr)) == RTX_AUTOINC; - rtx base = ldp_strip_offset (mem, &mem_off); + rtx base = pair_mem_strip_offset (mem, &mem_off); if (!REG_P (base)) return; @@ -507,7 +682,7 @@ ldp_bb_info::track_access (insn_info *insn, bool load_p, rtx mem) // accesses until after RA. // // As it stands, addresses with offsets in range for LDR but not - // in range for LDP/STP are currently reloaded inefficiently, + // in range for PAIR MEM LOAD STORE are currently reloaded inefficiently, // ending up with a separate base register for each pair. // // In theory LRA should make use of @@ -519,8 +694,8 @@ ldp_bb_info::track_access (insn_info *insn, bool load_p, rtx mem) // that calls targetm.legitimize_address_displacement. // // So for now, it's better to punt when we can't be sure that the - // offset is in range for LDP/STP. Out-of-range cases can then be - // handled after RA by the out-of-range LDP/STP peepholes. Eventually, it + // offset is in range for PAIR MEM LOAD STORE. Out-of-range cases can then be + // handled after RA by the out-of-range PAIR MEM peepholes. Eventually, it // would be nice to handle known out-of-range opportunities in the // pass itself (for stack accesses, this would be in the post-RA pass). if (!reload_completed @@ -573,7 +748,7 @@ ldp_bb_info::track_access (insn_info *insn, bool load_p, rtx mem) gcc_unreachable (); // Base defs should be unique. } - // Punt on misaligned offsets. LDP/STP require offsets to be a multiple of + // Punt on misaligned offsets. PAIR MEM require offsets to be a multiple of // the access size. if (!multiple_p (mem_off, mem_size)) return; @@ -612,9 +787,9 @@ static bool no_ignore (insn_info *) { return false; } // // N.B. we ignore any defs/uses of memory here as we deal with that separately, // making use of alias disambiguation. -static insn_info * +insn_info * latest_hazard_before (insn_info *insn, rtx *ignore, - insn_info *ignore_insn = nullptr) + insn_info *ignore_insn) { insn_info *result = nullptr; @@ -698,7 +873,7 @@ latest_hazard_before (insn_info *insn, rtx *ignore, // // N.B. we ignore any defs/uses of memory here as we deal with that separately, // making use of alias disambiguation. -static insn_info * +insn_info * first_hazard_after (insn_info *insn, rtx *ignore) { insn_info *result = nullptr; @@ -787,7 +962,7 @@ first_hazard_after (insn_info *insn, rtx *ignore) } // Return true iff R1 and R2 overlap. -static bool +bool ranges_overlap_p (const insn_range_info &r1, const insn_range_info &r2) { // If either range is empty, then their intersection is empty. @@ -801,7 +976,7 @@ ranges_overlap_p (const insn_range_info &r1, const insn_range_info &r2) } // Get the range of insns that def feeds. -static insn_range_info get_def_range (def_info *def) +insn_range_info get_def_range (def_info *def) { insn_info *last = def->next_def ()->insn ()->prev_nondebug_insn (); return { def->insn (), last }; @@ -809,7 +984,7 @@ static insn_range_info get_def_range (def_info *def) // Given a def (of memory), return the downwards range within which we // can safely move this def. -static insn_range_info +insn_range_info def_downwards_move_range (def_info *def) { auto range = get_def_range (def); @@ -827,7 +1002,7 @@ def_downwards_move_range (def_info *def) // Given a def (of memory), return the upwards range within which we can // safely move this def. -static insn_range_info +insn_range_info def_upwards_move_range (def_info *def) { def_info *prev = def->prev_def (); @@ -974,7 +1149,7 @@ private: // Given candidate store insns FIRST and SECOND, see if we can re-purpose one // of them (together with its def of memory) for the stp insn. If so, return // that insn. Otherwise, return null. -static insn_info * +insn_info * try_repurpose_store (insn_info *first, insn_info *second, const insn_range_info &move_range) @@ -1001,7 +1176,7 @@ try_repurpose_store (insn_info *first, // // These are deleted at the end of the pass and uses re-parented appropriately // at this point. -static rtx +rtx gen_tombstone (void) { return gen_rtx_CLOBBER (VOIDmode, @@ -1034,7 +1209,7 @@ aarch64_operand_mode_for_pair_mode (machine_mode mode) // REG_EH_REGION note in the resulting list. FR_EXPR is used to return any // REG_FRAME_RELATED_EXPR note we find, as these can need special handling in // combine_reg_notes. -static rtx +rtx filter_notes (rtx note, rtx result, bool *eh_region, rtx *fr_expr) { for (; note; note = XEXP (note, 1)) @@ -1084,7 +1259,7 @@ filter_notes (rtx note, rtx result, bool *eh_region, rtx *fr_expr) // Return the notes that should be attached to a combination of I1 and I2, where // *I1 < *I2. LOAD_P is true for loads. -static rtx +rtx combine_reg_notes (insn_info *i1, insn_info *i2, bool load_p) { // Temporary storage for REG_FRAME_RELATED_EXPR notes. @@ -1133,7 +1308,7 @@ combine_reg_notes (insn_info *i1, insn_info *i2, bool load_p) // relative to the initial value of the base register, and output these // in PATS. Return an rtx that represents the overall change to the // base register. -static rtx +rtx extract_writebacks (bool load_p, rtx pats[2], int changed) { rtx base_reg = NULL_RTX; @@ -1150,7 +1325,7 @@ extract_writebacks (bool load_p, rtx pats[2], int changed) const bool autoinc_p = GET_RTX_CLASS (GET_CODE (addr)) == RTX_AUTOINC; poly_int64 offset; - rtx this_base = ldp_strip_offset (mem, &offset); + rtx this_base = pair_mem_strip_offset (mem, &offset); gcc_assert (REG_P (this_base)); if (base_reg) gcc_assert (rtx_equal_p (base_reg, this_base)); @@ -1207,7 +1382,7 @@ extract_writebacks (bool load_p, rtx pats[2], int changed) // base register. If there is one, we choose the first such update after // PAIR_DST that is still in the same BB as our pair. We return the new def in // *ADD_DEF and the resulting writeback effect in *WRITEBACK_EFFECT. -static insn_info * +insn_info * find_trailing_add (insn_info *insns[2], const insn_range_info &pair_range, int initial_writeback, @@ -1286,7 +1461,7 @@ find_trailing_add (insn_info *insns[2], off_hwi /= access_size; - if (off_hwi < LDP_MIN_IMM || off_hwi > LDP_MAX_IMM) + if (off_hwi < PAIR_MEM_MIN_IMM || off_hwi > PAIR_MEM_MAX_IMM) return nullptr; auto dump_prefix = [&]() @@ -1328,7 +1503,7 @@ find_trailing_add (insn_info *insns[2], // We just emitted a tombstone with uid UID, track it in a bitmap for // this BB so we can easily identify it later when cleaning up tombstones. void -ldp_bb_info::track_tombstone (int uid) +pair_fusion::track_tombstone (int uid) { if (!m_emitted_tombstone) { @@ -1344,7 +1519,7 @@ ldp_bb_info::track_tombstone (int uid) // Reset the debug insn containing USE (the debug insn has been // optimized away). -static void +void reset_debug_use (use_info *use) { auto use_insn = use->insn (); @@ -1360,7 +1535,7 @@ reset_debug_use (use_info *use) // is an update of the register BASE by a constant, given by WB_OFFSET, // and we can preserve debug info by accounting for the change in side // effects. -static void +void fixup_debug_use (obstack_watermark &attempt, use_info *use, def_info *def, @@ -1463,7 +1638,7 @@ fixup_debug_use (obstack_watermark &attempt, // is a trailing add insn which is being folded into the pair to make it // use writeback addressing, and WRITEBACK_EFFECT is the pattern for // TRAILING_ADD. -static void +void fixup_debug_uses_trailing_add (obstack_watermark &attempt, insn_info *pair_dst, insn_info *trailing_add, @@ -1500,7 +1675,7 @@ fixup_debug_uses_trailing_add (obstack_watermark &attempt, // writeback, and WRITEBACK_EFFECT is an rtx describing the overall update to // the base register in the final pair (if any). BASE_REGNO gives the register // number of the base register used in the final pair. -static void +void fixup_debug_uses (obstack_watermark &attempt, insn_info *insns[2], rtx orig_rtl[2], @@ -1528,7 +1703,7 @@ fixup_debug_uses (obstack_watermark &attempt, gcc_checking_assert (GET_RTX_CLASS (GET_CODE (XEXP (mem, 0))) == RTX_AUTOINC); - base = ldp_strip_offset (mem, &offset); + base = pair_mem_strip_offset (mem, &offset); gcc_checking_assert (REG_P (base) && REGNO (base) == base_regno); } fixup_debug_use (attempt, use, def, base, offset); @@ -1664,7 +1839,7 @@ fixup_debug_uses (obstack_watermark &attempt, // BASE gives the chosen base candidate for the pair and MOVE_RANGE is // a singleton range which says where to place the pair. bool -ldp_bb_info::fuse_pair (bool load_p, +pair_fusion::fuse_pair (bool load_p, unsigned access_size, int writeback, insn_info *i1, insn_info *i2, @@ -1684,6 +1859,9 @@ ldp_bb_info::fuse_pair (bool load_p, insn_change::DELETE); }; + if (*i1 > *i2) + return false; + insn_info *first = (*i1 < *i2) ? i1 : i2; insn_info *second = (first == i1) ? i2 : i1; @@ -1800,7 +1978,7 @@ ldp_bb_info::fuse_pair (bool load_p, { if (dump_file) fprintf (dump_file, - " ldp: i%d has wb but subsequent i%d has non-wb " + " pair_mem: i%d has wb but subsequent i%d has non-wb " "update of base (r%d), dropping wb\n", insns[0]->uid (), insns[1]->uid (), base_regno); gcc_assert (writeback_effect); @@ -1823,7 +2001,7 @@ ldp_bb_info::fuse_pair (bool load_p, } // If either of the original insns had writeback, but the resulting pair insn - // does not (can happen e.g. in the ldp edge case above, or if the writeback + // does not (can happen e.g. in the pair mem edge case above, or if the writeback // effects cancel out), then drop the def(s) of the base register as // appropriate. // @@ -1842,7 +2020,7 @@ ldp_bb_info::fuse_pair (bool load_p, // update of the base register and try and fold it in to make this into a // writeback pair. insn_info *trailing_add = nullptr; - if (aarch64_ldp_writeback > 1 + if (pair_trailing_writeback_p () && !writeback_effect && (!load_p || (!refers_to_regno_p (base_regno, base_regno + 1, XEXP (pats[0], 0), nullptr) @@ -1863,14 +2041,14 @@ ldp_bb_info::fuse_pair (bool load_p, } // Now that we know what base mem we're going to use, check if it's OK - // with the ldp/stp policy. + // with the pair mem policy. rtx first_mem = XEXP (pats[0], load_p); - if (!aarch64_mem_ok_with_ldpstp_policy_model (first_mem, - load_p, - GET_MODE (first_mem))) + if (!pair_mem_ok_policy (first_mem, + load_p, + GET_MODE (first_mem))) { if (dump_file) - fprintf (dump_file, "punting on pair (%d,%d), ldp/stp policy says no\n", + fprintf (dump_file, "punting on pair (%d,%d), pair mem policy says no\n", i1->uid (), i2->uid ()); return false; } @@ -1878,20 +2056,12 @@ ldp_bb_info::fuse_pair (bool load_p, rtx reg_notes = combine_reg_notes (first, second, load_p); rtx pair_pat; - if (writeback_effect) - { - auto patvec = gen_rtvec (3, writeback_effect, pats[0], pats[1]); - pair_pat = gen_rtx_PARALLEL (VOIDmode, patvec); - } - else if (load_p) - pair_pat = aarch64_gen_load_pair (XEXP (pats[0], 0), - XEXP (pats[1], 0), - XEXP (pats[0], 1)); - else - pair_pat = aarch64_gen_store_pair (XEXP (pats[0], 0), - XEXP (pats[0], 1), - XEXP (pats[1], 1)); + set_multiword_subreg (first, second, load_p); + + pair_pat = gen_load_store_pair (pats, writeback_effect, load_p); + if (pair_pat == NULL_RTX) + return false; insn_change *pair_change = nullptr; auto set_pair_pat = [pair_pat,reg_notes](insn_change *change) { rtx_insn *rti = change->insn ()->rtl (); @@ -2075,7 +2245,7 @@ ldp_bb_info::fuse_pair (bool load_p, // Return true if STORE_INSN may modify mem rtx MEM. Make sure we keep // within our BUDGET for alias analysis. -static bool +bool store_modifies_mem_p (rtx mem, insn_info *store_insn, int &budget) { if (!budget) @@ -2098,7 +2268,7 @@ store_modifies_mem_p (rtx mem, insn_info *store_insn, int &budget) // Return true if LOAD may be modified by STORE. Make sure we keep // within our BUDGET for alias analysis. -static bool +bool load_modified_by_store_p (insn_info *load, insn_info *store, int &budget) @@ -2133,15 +2303,6 @@ load_modified_by_store_p (insn_info *load, return false; } -// Virtual base class for load/store walkers used in alias analysis. -struct alias_walker -{ - virtual bool conflict_p (int &budget) const = 0; - virtual insn_info *insn () const = 0; - virtual bool valid () const = 0; - virtual void advance () = 0; -}; - // Implement some common functionality used by both store_walker // and load_walker. template @@ -2259,13 +2420,13 @@ public: // // We try to maintain the invariant that if a walker becomes invalid, we // set its pointer to null. -static void -do_alias_analysis (insn_info *alias_hazards[4], +void +pair_fusion::do_alias_analysis (insn_info *alias_hazards[4], alias_walker *walkers[4], bool load_p) { const int n_walkers = 2 + (2 * !load_p); - int budget = aarch64_ldp_alias_check_limit; + int budget = pair_mem_alias_check_limit(); auto next_walker = [walkers,n_walkers](int current) -> int { for (int j = 1; j <= n_walkers; j++) @@ -2350,7 +2511,7 @@ do_alias_analysis (insn_info *alias_hazards[4], // // Returns an integer where bit (1 << i) is set if INSNS[i] uses writeback // addressing. -static int +int get_viable_bases (insn_info *insns[2], vec &base_cands, rtx cand_mems[2], @@ -2365,7 +2526,7 @@ get_viable_bases (insn_info *insns[2], { const bool is_lower = (i == reversed); poly_int64 poly_off; - rtx base = ldp_strip_offset (cand_mems[i], &poly_off); + rtx base = pair_mem_strip_offset (cand_mems[i], &poly_off); if (GET_RTX_CLASS (GET_CODE (XEXP (cand_mems[i], 0))) == RTX_AUTOINC) writeback |= (1 << i); @@ -2373,7 +2534,7 @@ get_viable_bases (insn_info *insns[2], continue; // Punt on accesses relative to eliminable regs. See the comment in - // ldp_bb_info::track_access for a detailed explanation of this. + // pair_fusion::track_access for a detailed explanation of this. if (!reload_completed && (REGNO (base) == FRAME_POINTER_REGNUM || REGNO (base) == ARG_POINTER_REGNUM)) @@ -2397,7 +2558,7 @@ get_viable_bases (insn_info *insns[2], if (!is_lower) base_off--; - if (base_off < LDP_MIN_IMM || base_off > LDP_MAX_IMM) + if (base_off < PAIR_MEM_MIN_IMM || base_off > PAIR_MEM_MAX_IMM) continue; use_info *use = find_access (insns[i]->uses (), REGNO (base)); @@ -2454,12 +2615,12 @@ get_viable_bases (insn_info *insns[2], } // Given two adjacent memory accesses of the same size, I1 and I2, try -// and see if we can merge them into a ldp or stp. +// and see if we can merge them into a pair mem load and store. // // ACCESS_SIZE gives the (common) size of a single access, LOAD_P is true // if the accesses are both loads, otherwise they are both stores. bool -ldp_bb_info::try_fuse_pair (bool load_p, unsigned access_size, +pair_fusion::try_fuse_pair (bool load_p, unsigned access_size, insn_info *i1, insn_info *i2) { if (dump_file) @@ -2490,11 +2651,21 @@ ldp_bb_info::try_fuse_pair (bool load_p, unsigned access_size, reg_ops[i] = XEXP (pats[i], !load_p); } + if (!load_p && !fuseable_store_p (i1, i2)) + { + if (dump_file) + fprintf (dump_file, + "punting on store-mem-pairs due to non fuseable cand (%d,%d)\n", + insns[0]->uid (), insns[1]->uid ()); + return false; + } + + if (load_p && reg_overlap_mentioned_p (reg_ops[0], reg_ops[1])) { if (dump_file) fprintf (dump_file, - "punting on ldp due to reg conflcits (%d,%d)\n", + "punting on pair mem load due to reg conflcits (%d,%d)\n", insns[0]->uid (), insns[1]->uid ()); return false; } @@ -2787,7 +2958,7 @@ ldp_bb_info::try_fuse_pair (bool load_p, unsigned access_size, i1, i2, *base, range); } -static void +void dump_insn_list (FILE *f, const insn_list_t &l) { fprintf (f, "("); @@ -2843,7 +3014,7 @@ debug (const insn_list_t &l) // we can't re-order them anyway, so provided earlier passes have cleaned up // redundant loads, we shouldn't miss opportunities by doing this. void -ldp_bb_info::merge_pairs (insn_list_t &left_list, +pair_fusion::merge_pairs (insn_list_t &left_list, insn_list_t &right_list, bool load_p, unsigned access_size) @@ -2890,8 +3061,8 @@ ldp_bb_info::merge_pairs (insn_list_t &left_list, // of accesses. If we find two sets of adjacent accesses, call // merge_pairs. void -ldp_bb_info::transform_for_base (int encoded_lfs, - access_group &group) +aarch64_pair_fusion::transform_for_base (int encoded_lfs, + access_group &group) { const auto lfs = decode_lfs (encoded_lfs); const unsigned access_size = lfs.size; @@ -2919,7 +3090,7 @@ ldp_bb_info::transform_for_base (int encoded_lfs, // and remove all the tombstone insns, being sure to reparent any uses // of mem to previous defs when we do this. void -ldp_bb_info::cleanup_tombstones () +pair_fusion::cleanup_tombstones () { // No need to do anything if we didn't emit a tombstone insn for this BB. if (!m_emitted_tombstone) @@ -2947,7 +3118,7 @@ ldp_bb_info::cleanup_tombstones () template void -ldp_bb_info::traverse_base_map (Map &map) +pair_fusion::traverse_base_map (Map &map) { for (auto kv : map) { @@ -2958,7 +3129,7 @@ ldp_bb_info::traverse_base_map (Map &map) } void -ldp_bb_info::transform () +pair_fusion::transform () { traverse_base_map (expr_map); traverse_base_map (def_map); @@ -3174,7 +3345,9 @@ void ldp_fusion_bb (bb_info *bb) const bool track_stores = aarch64_tune_params.stp_policy_model != AARCH64_LDP_STP_POLICY_NEVER; - ldp_bb_info bb_state (bb); + pair_fusion *bb_state; + aarch64_pair_fusion derived (bb); + bb_state = &derived; for (auto insn : bb->nondebug_insns ()) { @@ -3194,13 +3367,13 @@ void ldp_fusion_bb (bb_info *bb) continue; if (track_stores && MEM_P (XEXP (pat, 0))) - bb_state.track_access (insn, false, XEXP (pat, 0)); + bb_state->track_access (insn, false, XEXP (pat, 0)); else if (track_loads && MEM_P (XEXP (pat, 1))) - bb_state.track_access (insn, true, XEXP (pat, 1)); + bb_state->track_access (insn, true, XEXP (pat, 1)); } - bb_state.transform (); - bb_state.cleanup_tombstones (); + bb_state->transform (); + bb_state->cleanup_tombstones (); } void ldp_fusion ()