From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-lf1-x132.google.com (mail-lf1-x132.google.com [IPv6:2a00:1450:4864:20::132]) by sourceware.org (Postfix) with ESMTPS id 12F7C385840A for ; Wed, 30 Aug 2023 10:14:30 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org 12F7C385840A Authentication-Results: sourceware.org; dmarc=none (p=none dis=none) header.from=vrull.eu Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=vrull.eu Received: by mail-lf1-x132.google.com with SMTP id 2adb3069b0e04-500b66f8b27so5481848e87.3 for ; Wed, 30 Aug 2023 03:14:29 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=vrull.eu; s=google; t=1693390467; x=1693995267; darn=gcc.gnu.org; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:from:to:cc:subject:date :message-id:reply-to; bh=aGJ9I36+rOb41U1fBn0WlMvxiDnLPSAV/bsicCRMY8c=; b=T0qUc7CT3VvlyqNOYjV/V1GyTiqS2bGZ78E0PQhsFRoT5ODuxuOmEOv/ZFHYcaw0o3 5fHPiWczgiQQmx2wsb763T6hWsa5nhepS9qr70MkNrjnEnZ5JhGOmU4M5Y54BPSJifwm QiPrB2B+BF70kYsjFMjbB6JnX4Y3J9kfrEisLIgUp8ItoQ2YWKbBWMaYH9lRJODZQbAZ Us6mmtZP4zZHtdH3KrSr4h9QxxNitpOkOiiVB9f8LF2ohDpydenRRYsXGYmMLTKoercW 0g+bUjXEgIS2exxWqMY6qO24H1+gvdmt6oTL+8MLobrJ0a2lCWCRz0l28saqu3kDawK6 MH8Q== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20221208; t=1693390467; x=1693995267; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=aGJ9I36+rOb41U1fBn0WlMvxiDnLPSAV/bsicCRMY8c=; b=Ea6S8vnVFlKEtEZdU5IPs4WPkcpFscDsFihtDymS/U1AOrtM+QoELY+2suwL/7QRT/ Mc69rqNHOBbnXZDICkc/A9TosiyFqhXTxZCBLb9ERzHMo3MAtCeBXj5yvbCskKd5dpmN E7Yq8vOOFWIjQXsrJRetpE8gXSYYEwHccXBpC/pKz9j/HT6c0aM/RgxNUB047ICLfAk/ KpAKWyq7s8mTwFt3gLGdlc/iqvoHpad0yeUKejBzMbHoTjLuB4C0Z01fDDZQZqqAgogt pcoRMjC9unSTA6y9KLbiIyA/+hbx1vTYexZCbjBl1PbXXylh1i10INGT5B/w2+xrPm4k oyhg== X-Gm-Message-State: AOJu0YyqCX61AJhkVe6bNLf0ZXfpOjLhhg9xYoJB0vCtjVLuHiu6Bpij V1Q+qLy+BhMnLh5rK7t8FvpaIsYX6L8fRfUman5gJg== X-Google-Smtp-Source: AGHT+IFU+SEET5hRyV3t7CObIpVNfInryecXHShG48Fbe3IErZp5wNuQFEr9H/xwpLlyxtGbmaBTqA== X-Received: by 2002:a19:f604:0:b0:500:8fe8:7e80 with SMTP id x4-20020a19f604000000b005008fe87e80mr1161107lfe.26.1693390467088; Wed, 30 Aug 2023 03:14:27 -0700 (PDT) Received: from helsinki-03.engr ([2a01:4f9:6b:2a47::2]) by smtp.gmail.com with ESMTPSA id w7-20020a19c507000000b004fe36e673b8sm2327579lfe.178.2023.08.30.03.14.26 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Wed, 30 Aug 2023 03:14:26 -0700 (PDT) From: Manolis Tsamis To: gcc-patches@gcc.gnu.org Cc: Philipp Tomsich , Robin Dapp , Jakub Jelinek , Richard Sandiford , Manolis Tsamis Subject: [PATCH v3 3/4] ifcvt: Handle multiple rewired regs and refactor noce_convert_multiple_sets Date: Wed, 30 Aug 2023 12:13:59 +0200 Message-Id: <20230830101400.1539313-4-manolis.tsamis@vrull.eu> X-Mailer: git-send-email 2.34.1 In-Reply-To: <20230830101400.1539313-1-manolis.tsamis@vrull.eu> References: <20230830101400.1539313-1-manolis.tsamis@vrull.eu> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Spam-Status: No, score=-11.3 required=5.0 tests=BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,GIT_PATCH_0,JMQ_SPF_NEUTRAL,KAM_SHORT,RCVD_IN_DNSWL_NONE,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: The existing implementation of need_cmov_or_rewire and noce_convert_multiple_sets_1 assumes that sets are either REG or SUBREG. This commit enchances them so they can handle/rewire arbitrary set statements. To do that a new helper struct noce_multiple_sets_info is introduced which is used by noce_convert_multiple_sets and its helper functions. This results in cleaner function signatures, improved efficientcy (a number of vecs and hash set/map are replaced with a single vec of struct) and simplicity. gcc/ChangeLog: * ifcvt.cc (need_cmov_or_rewire): Renamed init_noce_multiple_sets_info. (init_noce_multiple_sets_info): Initialize noce_multiple_sets_info. (noce_convert_multiple_sets_1): Use noce_multiple_sets_info and handle rewiring of multiple registers. (noce_convert_multiple_sets): Updated to use noce_multiple_sets_info. * ifcvt.h (struct noce_multiple_sets_info): Introduce new struct noce_multiple_sets_info to store info for noce_convert_multiple_sets. gcc/testsuite/ChangeLog: * gcc.target/aarch64/ifcvt_multiple_sets_rewire.c: New test. Signed-off-by: Manolis Tsamis --- (no changes since v1) gcc/ifcvt.cc | 255 ++++++++---------- gcc/ifcvt.h | 16 ++ .../aarch64/ifcvt_multiple_sets_rewire.c | 20 ++ 3 files changed, 149 insertions(+), 142 deletions(-) create mode 100644 gcc/testsuite/gcc.target/aarch64/ifcvt_multiple_sets_rewire.c diff --git a/gcc/ifcvt.cc b/gcc/ifcvt.cc index efe8ab1577a..ecc0cbabef9 100644 --- a/gcc/ifcvt.cc +++ b/gcc/ifcvt.cc @@ -98,14 +98,10 @@ static bool dead_or_predicable (basic_block, basic_block, basic_block, edge, bool); static void noce_emit_move_insn (rtx, rtx); static rtx_insn *block_has_only_trap (basic_block); -static void need_cmov_or_rewire (basic_block, hash_set *, - hash_map *); +static void init_noce_multiple_sets_info (basic_block, + auto_delete_vec &); static bool noce_convert_multiple_sets_1 (struct noce_if_info *, - hash_set *, - hash_map *, - auto_vec *, - auto_vec *, - auto_vec *, int *); + auto_delete_vec &, int *); /* Count the number of non-jump active insns in BB. */ @@ -3270,24 +3266,13 @@ noce_convert_multiple_sets (struct noce_if_info *if_info) rtx x = XEXP (cond, 0); rtx y = XEXP (cond, 1); - /* The true targets for a conditional move. */ - auto_vec targets; - /* The temporaries introduced to allow us to not consider register - overlap. */ - auto_vec temporaries; - /* The insns we've emitted. */ - auto_vec unmodified_insns; - - hash_set need_no_cmov; - hash_map rewired_src; - - need_cmov_or_rewire (then_bb, &need_no_cmov, &rewired_src); + auto_delete_vec insn_info; + init_noce_multiple_sets_info (then_bb, insn_info); int last_needs_comparison = -1; bool ok = noce_convert_multiple_sets_1 - (if_info, &need_no_cmov, &rewired_src, &targets, &temporaries, - &unmodified_insns, &last_needs_comparison); + (if_info, insn_info, &last_needs_comparison); if (!ok) return false; @@ -3302,8 +3287,7 @@ noce_convert_multiple_sets (struct noce_if_info *if_info) end_sequence (); start_sequence (); ok = noce_convert_multiple_sets_1 - (if_info, &need_no_cmov, &rewired_src, &targets, &temporaries, - &unmodified_insns, &last_needs_comparison); + (if_info, insn_info, &last_needs_comparison); /* Actually we should not fail anymore if we reached here, but better still check. */ if (!ok) @@ -3312,12 +3296,12 @@ noce_convert_multiple_sets (struct noce_if_info *if_info) /* We must have seen some sort of insn to insert, otherwise we were given an empty BB to convert, and we can't handle that. */ - gcc_assert (!unmodified_insns.is_empty ()); + gcc_assert (!insn_info.is_empty ()); /* Now fixup the assignments. */ - for (unsigned i = 0; i < targets.length (); i++) - if (targets[i] != temporaries[i]) - noce_emit_move_insn (targets[i], temporaries[i]); + for (unsigned i = 0; i < insn_info.length (); i++) + if (insn_info[i]->target != insn_info[i]->temporary) + noce_emit_move_insn (insn_info[i]->target, insn_info[i]->temporary); /* Actually emit the sequence if it isn't too expensive. */ rtx_insn *seq = get_insns (); @@ -3332,10 +3316,10 @@ noce_convert_multiple_sets (struct noce_if_info *if_info) set_used_flags (insn); /* Mark all our temporaries and targets as used. */ - for (unsigned i = 0; i < targets.length (); i++) + for (unsigned i = 0; i < insn_info.length (); i++) { - set_used_flags (temporaries[i]); - set_used_flags (targets[i]); + set_used_flags (insn_info[i]->temporary); + set_used_flags (insn_info[i]->target); } set_used_flags (cond); @@ -3354,7 +3338,7 @@ noce_convert_multiple_sets (struct noce_if_info *if_info) return false; emit_insn_before_setloc (seq, if_info->jump, - INSN_LOCATION (unmodified_insns.last ())); + INSN_LOCATION (insn_info.last ()->unmodified_insn)); /* Clean up THEN_BB and the edges in and out of it. */ remove_edge (find_edge (test_bb, join_bb)); @@ -3375,20 +3359,12 @@ noce_convert_multiple_sets (struct noce_if_info *if_info) return true; } -/* This goes through all relevant insns of IF_INFO->then_bb and tries to - create conditional moves. In case a simple move sufficis the insn - should be listed in NEED_NO_CMOV. The rewired-src cases should be - specified via REWIRED_SRC. TARGETS, TEMPORARIES and UNMODIFIED_INSNS - are specified and used in noce_convert_multiple_sets and should be passed - to this function.. */ +/* This goes through all relevant insns of IF_INFO->then_bb and tries to create + conditional moves. Information for the insns is kept in INSN_INFO. */ static bool noce_convert_multiple_sets_1 (struct noce_if_info *if_info, - hash_set *need_no_cmov, - hash_map *rewired_src, - auto_vec *targets, - auto_vec *temporaries, - auto_vec *unmodified_insns, + auto_delete_vec &insn_info, int *last_needs_comparison) { basic_block then_bb = if_info->then_bb; @@ -3407,11 +3383,6 @@ noce_convert_multiple_sets_1 (struct noce_if_info *if_info, rtx_insn *insn; int count = 0; - - targets->truncate (0); - temporaries->truncate (0); - unmodified_insns->truncate (0); - bool second_try = *last_needs_comparison != -1; FOR_BB_INSNS (then_bb, insn) @@ -3420,6 +3391,8 @@ noce_convert_multiple_sets_1 (struct noce_if_info *if_info, if (!active_insn_p (insn)) continue; + noce_multiple_sets_info *info = insn_info[count]; + rtx set = single_set (insn); gcc_checking_assert (set); @@ -3427,9 +3400,12 @@ noce_convert_multiple_sets_1 (struct noce_if_info *if_info, rtx temp; rtx new_val = SET_SRC (set); - if (int *ii = rewired_src->get (insn)) - new_val = simplify_replace_rtx (new_val, (*targets)[*ii], - (*temporaries)[*ii]); + + int i, ii; + FOR_EACH_VEC_ELT (info->rewired_src, i, ii) + new_val = simplify_replace_rtx (new_val, insn_info[ii]->target, + insn_info[ii]->temporary); + rtx old_val = target; /* As we are transforming @@ -3467,11 +3443,6 @@ noce_convert_multiple_sets_1 (struct noce_if_info *if_info, else temp = target; - /* We have identified swap-style idioms before. A normal - set will need to be a cmov while the first instruction of a swap-style - idiom can be a regular move. This helps with costing. */ - bool need_cmov = !need_no_cmov->contains (insn); - /* If we had a non-canonical conditional jump (i.e. one where the fallthrough is to the "else" case) we need to reverse the conditional select. */ @@ -3516,6 +3487,11 @@ noce_convert_multiple_sets_1 (struct noce_if_info *if_info, old_val = lowpart_subreg (dst_mode, old_val, src_mode); } + /* We have identified swap-style idioms before. A normal + set will need to be a cmov while the first instruction of a swap-style + idiom can be a regular move. This helps with costing. */ + bool need_cmov = info->need_cmov; + /* Try emitting a conditional move passing the backend the canonicalized comparison. The backend is then able to recognize expressions like @@ -3621,9 +3597,10 @@ noce_convert_multiple_sets_1 (struct noce_if_info *if_info, /* Bookkeeping. */ count++; - targets->safe_push (target); - temporaries->safe_push (temp_dest); - unmodified_insns->safe_push (insn); + + info->target = target; + info->temporary = temp_dest; + info->unmodified_insn = insn; } /* Even if we did not actually need the comparison, we want to make sure @@ -3631,11 +3608,88 @@ noce_convert_multiple_sets_1 (struct noce_if_info *if_info, if (*last_needs_comparison == -1) *last_needs_comparison = 0; - return true; } +/* Find local swap-style idioms in BB and mark the first insn (1) + that is only a temporary as not needing a conditional move as + it is going to be dead afterwards anyway. + + (1) int tmp = a; + a = b; + b = tmp; + ifcvt + --> + + tmp = a; + a = cond ? b : a_old; + b = cond ? tmp : b_old; + + Additionally, store the index of insns like (2) when a subsequent + SET reads from their destination. + + (2) int c = a; + int d = c; + + ifcvt + --> + + c = cond ? a : c_old; + d = cond ? d : c; // Need to use c rather than c_old here. +*/ + +static void +init_noce_multiple_sets_info (basic_block bb, + auto_delete_vec &insn_info) +{ + rtx_insn *insn; + int count = 0; + auto_vec dests; + + /* Iterate over all SETs, storing the destinations + in DEST. + - If we hit a SET that reads from a destination + that we have seen before and the corresponding register + is dead afterwards, the register does not need to be + moved conditionally. + - If we encounter a previously changed register, + rewire the read to the original source. */ + FOR_BB_INSNS (bb, insn) + { + if (!active_insn_p (insn)) + continue; + + noce_multiple_sets_info *info = new noce_multiple_sets_info; + info->target = NULL_RTX; + info->temporary = NULL_RTX; + info->unmodified_insn = NULL; + info->need_cmov = true; + insn_info.safe_push (info); + + rtx set = single_set (insn); + gcc_checking_assert (set); + + rtx src = SET_SRC (set); + rtx dest = SET_DEST (set); + + /* Check if the current SET's source is the same + as any previously seen destination. + This is quadratic but the number of insns in BB + is bounded by PARAM_MAX_RTL_IF_CONVERSION_INSNS. */ + for (int i = count - 1; i >= 0; --i) + if (reg_mentioned_p (dests[i], src)) + { + if (find_reg_note (insn, REG_DEAD, src) != NULL_RTX) + insn_info[i]->need_cmov = false; + else + insn_info[count]->rewired_src.safe_push (i); + } + + dests.safe_push (dest); + count++; + } +} /* Return true iff basic block TEST_BB is suitable for conversion to a series of conditional moves. Also check that we have more than one @@ -4135,89 +4189,6 @@ check_cond_move_block (basic_block bb, return true; } -/* Find local swap-style idioms in BB and mark the first insn (1) - that is only a temporary as not needing a conditional move as - it is going to be dead afterwards anyway. - - (1) int tmp = a; - a = b; - b = tmp; - - ifcvt - --> - - tmp = a; - a = cond ? b : a_old; - b = cond ? tmp : b_old; - - Additionally, store the index of insns like (2) when a subsequent - SET reads from their destination. - - (2) int c = a; - int d = c; - - ifcvt - --> - - c = cond ? a : c_old; - d = cond ? d : c; // Need to use c rather than c_old here. -*/ - -static void -need_cmov_or_rewire (basic_block bb, - hash_set *need_no_cmov, - hash_map *rewired_src) -{ - rtx_insn *insn; - int count = 0; - auto_vec insns; - auto_vec dests; - - /* Iterate over all SETs, storing the destinations - in DEST. - - If we hit a SET that reads from a destination - that we have seen before and the corresponding register - is dead afterwards, the register does not need to be - moved conditionally. - - If we encounter a previously changed register, - rewire the read to the original source. */ - FOR_BB_INSNS (bb, insn) - { - rtx set, src, dest; - - if (!active_insn_p (insn)) - continue; - - set = single_set (insn); - if (set == NULL_RTX) - continue; - - src = SET_SRC (set); - if (SUBREG_P (src)) - src = SUBREG_REG (src); - dest = SET_DEST (set); - - /* Check if the current SET's source is the same - as any previously seen destination. - This is quadratic but the number of insns in BB - is bounded by PARAM_MAX_RTL_IF_CONVERSION_INSNS. */ - if (REG_P (src)) - for (int i = count - 1; i >= 0; --i) - if (reg_overlap_mentioned_p (src, dests[i])) - { - if (find_reg_note (insn, REG_DEAD, src) != NULL_RTX) - need_no_cmov->add (insns[i]); - else - rewired_src->put (insn, i); - } - - insns.safe_push (insn); - dests.safe_push (dest); - - count++; - } -} - /* Given a basic block BB suitable for conditional move conversion, a condition COND, and pointer maps THEN_VALS and ELSE_VALS containing the register values depending on COND, emit the insns in the block as diff --git a/gcc/ifcvt.h b/gcc/ifcvt.h index be1385aabe4..f953705f887 100644 --- a/gcc/ifcvt.h +++ b/gcc/ifcvt.h @@ -40,6 +40,22 @@ struct ce_if_block int pass; /* Pass number. */ }; +struct noce_multiple_sets_info +{ + /* A list of indices to instructions that we need to rewire into this + instruction when we replace them with temporary conditional moves. */ + auto_vec rewired_src; + /* The true targets for a conditional move. */ + rtx target; + /* The temporaries introduced to allow us to not consider register + overlap. */ + rtx temporary; + /* The insns we've emitted. */ + rtx_insn *unmodified_insn; + /* True if a simple move can be used instead of a conditional move. */ + bool need_cmov; +}; + /* Used by noce_process_if_block to communicate with its subroutines. The subroutines know that A and B may be evaluated freely. They diff --git a/gcc/testsuite/gcc.target/aarch64/ifcvt_multiple_sets_rewire.c b/gcc/testsuite/gcc.target/aarch64/ifcvt_multiple_sets_rewire.c new file mode 100644 index 00000000000..411874e96c2 --- /dev/null +++ b/gcc/testsuite/gcc.target/aarch64/ifcvt_multiple_sets_rewire.c @@ -0,0 +1,20 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -fdump-rtl-ce1" } */ + +void sink2(int, int); + +void cond1(int cond, int x, int y, int z) +{ + if (x) + { + x = y + z; + y = z + x; + } + + sink2(x, y); +} + +/* { dg-final { scan-assembler-times "csel\tw0, w0, w1" 1 } } */ +/* { dg-final { scan-assembler-times "csel\tw1, w3, w2" 1 } } */ + +/* { dg-final { scan-rtl-dump-times "if-conversion succeeded through noce_convert_multiple_sets" 1 "ce1" } } */ \ No newline at end of file -- 2.34.1