From patchwork Thu Jan 13 17:29:38 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Jakub Jelinek X-Patchwork-Id: 49997 Return-Path: X-Original-To: patchwork@sourceware.org Delivered-To: patchwork@sourceware.org Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id DE35C3857809 for ; Thu, 13 Jan 2022 17:31:52 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org DE35C3857809 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org; s=default; t=1642095112; bh=ZXI3q3TDPpGG3APMRi4v/RLbkIHeABjo+r5MVCcUUx0=; h=Date:To:Subject:References:In-Reply-To:List-Id:List-Unsubscribe: List-Archive:List-Post:List-Help:List-Subscribe:From:Reply-To:Cc: From; b=yUkVZz6cxhnQfHSVz0S+Tee89PRZyxjqChIatuQa7R/bvp64lh0f9qoDUskJ3xEd7 WIlW3AhyBNUDjjYOT7AIFN6PzT9u3P9hqWdTeB6NS4XPmnQw1Yow7168cb9xzdqh2N xqNa9/oxf+udHTj1WTf0zBWZljdwP190PkT+QdKo= X-Original-To: gcc-patches@gcc.gnu.org Delivered-To: gcc-patches@gcc.gnu.org Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.129.124]) by sourceware.org (Postfix) with ESMTPS id 24C6F385AC29 for ; Thu, 13 Jan 2022 17:30:05 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 24C6F385AC29 Received: from mimecast-mx01.redhat.com (mimecast-mx01.redhat.com [209.132.183.4]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id us-mta-529-NE_-7tG_PFiFYALnDFnAKQ-1; Thu, 13 Jan 2022 12:30:01 -0500 X-MC-Unique: NE_-7tG_PFiFYALnDFnAKQ-1 Received: from smtp.corp.redhat.com (int-mx03.intmail.prod.int.phx2.redhat.com [10.5.11.13]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mimecast-mx01.redhat.com (Postfix) with ESMTPS id 4ECF312CF00; Thu, 13 Jan 2022 17:29:45 +0000 (UTC) Received: from tucnak.zalov.cz (unknown [10.39.195.246]) by smtp.corp.redhat.com (Postfix) with ESMTPS id 5CC3F8588F; Thu, 13 Jan 2022 17:29:44 +0000 (UTC) Received: from tucnak.zalov.cz (localhost [127.0.0.1]) by tucnak.zalov.cz (8.16.1/8.16.1) with ESMTPS id 20DHTexq2643307 (version=TLSv1.3 cipher=TLS_AES_256_GCM_SHA384 bits=256 verify=NOT); Thu, 13 Jan 2022 18:29:40 +0100 Received: (from jakub@localhost) by tucnak.zalov.cz (8.16.1/8.16.1/Submit) id 20DHTd0n2643306; Thu, 13 Jan 2022 18:29:39 +0100 Date: Thu, 13 Jan 2022 18:29:38 +0100 To: Richard Biener Subject: [PATCH] forwprop, v2: Canonicalize atomic fetch_op op x to op_fetch or vice versa [PR98737] Message-ID: <20220113172938.GD2646553@tucnak> References: <20220113090351.GR2646553@tucnak> <4o21092-3s81-73q2-no26-s8n9p8r4qpq6@fhfr.qr> <20220113142752.GY2646553@tucnak> <0r9op46-83sq-67nr-q8pn-65o386s154s@fhfr.qr> MIME-Version: 1.0 In-Reply-To: <0r9op46-83sq-67nr-q8pn-65o386s154s@fhfr.qr> X-Scanned-By: MIMEDefang 2.79 on 10.5.11.13 X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com Content-Disposition: inline X-Spam-Status: No, score=-5.5 required=5.0 tests=BAYES_00, DKIMWL_WL_HIGH, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, RCVD_IN_DNSWL_LOW, RCVD_IN_MSPIKE_H3, RCVD_IN_MSPIKE_WL, SPF_HELO_NONE, SPF_NONE, 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: Jakub Jelinek via Gcc-patches From: Jakub Jelinek Reply-To: Jakub Jelinek Cc: gcc-patches@gcc.gnu.org Errors-To: gcc-patches-bounces+patchwork=sourceware.org@gcc.gnu.org Sender: "Gcc-patches" On Thu, Jan 13, 2022 at 04:07:20PM +0100, Richard Biener wrote: > I'm mostly concerned about the replace_uses_by use. forwprop > will go over newly emitted stmts and thus the hypothetical added > > lhs2 = d; > > record the copy and schedule the stmt for removal, substituting 'd' > in each use as it goes along the function and folding them. It's > a bit iffy (and maybe has unintended side-effects in odd cases) > to trample around and fold stuff behind that flows back. > > I'd always vote to simplify the folding code so it's easier to > maintain and not micro-optimize there since it's not going to be > a hot part of the compiler. Ok. So like this? 2022-01-13 Jakub Jelinek PR target/98737 * tree-ssa-forwprop.c (simplify_builtin_call): Canonicalize __atomic_fetch_op (p, x, y) op x into __atomic_op_fetch (p, x, y) and __atomic_op_fetch (p, x, y) iop x into __atomic_fetch_op (p, x, y). * gcc.dg/tree-ssa/pr98737-1.c: New test. * gcc.dg/tree-ssa/pr98737-2.c: New test. Jakub --- gcc/tree-ssa-forwprop.c.jj 2022-01-11 23:11:23.467275019 +0100 +++ gcc/tree-ssa-forwprop.c 2022-01-13 18:09:50.318625915 +0100 @@ -1241,12 +1241,19 @@ constant_pointer_difference (tree p1, tr memset (p + 4, ' ', 3); into memcpy (p, "abcd ", 7); - call if the latter can be stored by pieces during expansion. */ + call if the latter can be stored by pieces during expansion. + + Also canonicalize __atomic_fetch_op (p, x, y) op x + to __atomic_op_fetch (p, x, y) or + __atomic_op_fetch (p, x, y) iop x + to __atomic_fetch_op (p, x, y) when possible (also __sync). */ static bool simplify_builtin_call (gimple_stmt_iterator *gsi_p, tree callee2) { gimple *stmt1, *stmt2 = gsi_stmt (*gsi_p); + enum built_in_function other_atomic = END_BUILTINS; + enum tree_code atomic_op = ERROR_MARK; tree vuse = gimple_vuse (stmt2); if (vuse == NULL) return false; @@ -1448,6 +1455,290 @@ simplify_builtin_call (gimple_stmt_itera } } break; + + #define CASE_ATOMIC(NAME, OTHER, OP) \ + case BUILT_IN_##NAME##_1: \ + case BUILT_IN_##NAME##_2: \ + case BUILT_IN_##NAME##_4: \ + case BUILT_IN_##NAME##_8: \ + case BUILT_IN_##NAME##_16: \ + atomic_op = OP; \ + other_atomic \ + = (enum built_in_function) (BUILT_IN_##OTHER##_1 \ + + (DECL_FUNCTION_CODE (callee2) \ + - BUILT_IN_##NAME##_1)); \ + goto handle_atomic_fetch_op; + + CASE_ATOMIC (ATOMIC_FETCH_ADD, ATOMIC_ADD_FETCH, PLUS_EXPR) + CASE_ATOMIC (ATOMIC_FETCH_SUB, ATOMIC_SUB_FETCH, MINUS_EXPR) + CASE_ATOMIC (ATOMIC_FETCH_AND, ATOMIC_AND_FETCH, BIT_AND_EXPR) + CASE_ATOMIC (ATOMIC_FETCH_XOR, ATOMIC_XOR_FETCH, BIT_XOR_EXPR) + CASE_ATOMIC (ATOMIC_FETCH_OR, ATOMIC_OR_FETCH, BIT_IOR_EXPR) + + CASE_ATOMIC (SYNC_FETCH_AND_ADD, SYNC_ADD_AND_FETCH, PLUS_EXPR) + CASE_ATOMIC (SYNC_FETCH_AND_SUB, SYNC_SUB_AND_FETCH, MINUS_EXPR) + CASE_ATOMIC (SYNC_FETCH_AND_AND, SYNC_AND_AND_FETCH, BIT_AND_EXPR) + CASE_ATOMIC (SYNC_FETCH_AND_XOR, SYNC_XOR_AND_FETCH, BIT_XOR_EXPR) + CASE_ATOMIC (SYNC_FETCH_AND_OR, SYNC_OR_AND_FETCH, BIT_IOR_EXPR) + + CASE_ATOMIC (ATOMIC_ADD_FETCH, ATOMIC_FETCH_ADD, MINUS_EXPR) + CASE_ATOMIC (ATOMIC_SUB_FETCH, ATOMIC_FETCH_SUB, PLUS_EXPR) + CASE_ATOMIC (ATOMIC_XOR_FETCH, ATOMIC_FETCH_XOR, BIT_XOR_EXPR) + + CASE_ATOMIC (SYNC_ADD_AND_FETCH, SYNC_FETCH_AND_ADD, MINUS_EXPR) + CASE_ATOMIC (SYNC_SUB_AND_FETCH, SYNC_FETCH_AND_SUB, PLUS_EXPR) + CASE_ATOMIC (SYNC_XOR_AND_FETCH, SYNC_FETCH_AND_XOR, BIT_XOR_EXPR) + +#undef CASE_ATOMIC + + handle_atomic_fetch_op: + if (gimple_call_num_args (stmt2) >= 2 && gimple_call_lhs (stmt2)) + { + tree lhs2 = gimple_call_lhs (stmt2), lhsc = lhs2; + tree arg = gimple_call_arg (stmt2, 1); + gimple *use_stmt, *cast_stmt = NULL; + use_operand_p use_p; + tree ndecl = builtin_decl_explicit (other_atomic); + + if (ndecl == NULL_TREE || !single_imm_use (lhs2, &use_p, &use_stmt)) + break; + + if (gimple_assign_cast_p (use_stmt)) + { + cast_stmt = use_stmt; + lhsc = gimple_assign_lhs (cast_stmt); + if (lhsc == NULL_TREE + || !INTEGRAL_TYPE_P (TREE_TYPE (lhsc)) + || (TYPE_PRECISION (TREE_TYPE (lhsc)) + != TYPE_PRECISION (TREE_TYPE (lhs2))) + || !single_imm_use (lhsc, &use_p, &use_stmt)) + { + use_stmt = cast_stmt; + cast_stmt = NULL; + lhsc = lhs2; + } + } + + bool ok = false; + tree oarg = NULL_TREE; + enum tree_code ccode = ERROR_MARK; + tree crhs1 = NULL_TREE, crhs2 = NULL_TREE; + if (is_gimple_assign (use_stmt) + && gimple_assign_rhs_code (use_stmt) == atomic_op) + { + if (gimple_assign_rhs1 (use_stmt) == lhsc) + oarg = gimple_assign_rhs2 (use_stmt); + else if (atomic_op != MINUS_EXPR) + oarg = gimple_assign_rhs1 (use_stmt); + } + else if (atomic_op == MINUS_EXPR + && is_gimple_assign (use_stmt) + && gimple_assign_rhs_code (use_stmt) == PLUS_EXPR + && TREE_CODE (arg) == INTEGER_CST + && (TREE_CODE (gimple_assign_rhs2 (use_stmt)) + == INTEGER_CST)) + { + tree a = fold_convert (TREE_TYPE (lhs2), arg); + tree o = fold_convert (TREE_TYPE (lhs2), + gimple_assign_rhs2 (use_stmt)); + if (wi::to_wide (a) == wi::neg (wi::to_wide (o))) + ok = true; + } + else if (atomic_op == BIT_AND_EXPR || atomic_op == BIT_IOR_EXPR) + ; + else if (gimple_code (use_stmt) == GIMPLE_COND) + { + ccode = gimple_cond_code (use_stmt); + crhs1 = gimple_cond_lhs (use_stmt); + crhs2 = gimple_cond_rhs (use_stmt); + } + else if (is_gimple_assign (use_stmt)) + { + if (gimple_assign_rhs_class (use_stmt) == GIMPLE_BINARY_RHS) + { + ccode = gimple_assign_rhs_code (use_stmt); + crhs1 = gimple_assign_rhs1 (use_stmt); + crhs2 = gimple_assign_rhs2 (use_stmt); + } + else if (gimple_assign_rhs_code (use_stmt) == COND_EXPR) + { + tree cond = gimple_assign_rhs1 (use_stmt); + if (COMPARISON_CLASS_P (cond)) + { + ccode = TREE_CODE (cond); + crhs1 = TREE_OPERAND (cond, 0); + crhs2 = TREE_OPERAND (cond, 1); + } + } + } + if (ccode == EQ_EXPR || ccode == NE_EXPR) + { + /* Deal with x - y == 0 or x ^ y == 0 + being optimized into x == y and x + cst == 0 + into x == -cst. */ + tree o = NULL_TREE; + if (crhs1 == lhsc) + o = crhs2; + else if (crhs2 == lhsc) + o = crhs1; + if (o && atomic_op != PLUS_EXPR) + oarg = o; + else if (o + && TREE_CODE (o) == INTEGER_CST + && TREE_CODE (arg) == INTEGER_CST) + { + tree a = fold_convert (TREE_TYPE (lhs2), arg); + o = fold_convert (TREE_TYPE (lhs2), o); + if (wi::to_wide (a) == wi::neg (wi::to_wide (o))) + ok = true; + } + } + if (oarg && !ok) + { + if (operand_equal_p (arg, oarg, 0)) + ok = true; + else if (TREE_CODE (arg) == SSA_NAME + && TREE_CODE (oarg) == SSA_NAME) + { + tree oarg2 = oarg; + if (gimple_assign_cast_p (SSA_NAME_DEF_STMT (oarg))) + { + gimple *g = SSA_NAME_DEF_STMT (oarg); + oarg2 = gimple_assign_rhs1 (g); + if (TREE_CODE (oarg2) != SSA_NAME + || !INTEGRAL_TYPE_P (TREE_TYPE (oarg2)) + || (TYPE_PRECISION (TREE_TYPE (oarg2)) + != TYPE_PRECISION (TREE_TYPE (oarg)))) + oarg2 = oarg; + } + if (gimple_assign_cast_p (SSA_NAME_DEF_STMT (arg))) + { + gimple *g = SSA_NAME_DEF_STMT (arg); + tree rhs1 = gimple_assign_rhs1 (g); + /* Handle e.g. + x.0_1 = (long unsigned int) x_4(D); + _2 = __atomic_fetch_add_8 (&vlong, x.0_1, 0); + _3 = (long int) _2; + _7 = x_4(D) + _3; */ + if (rhs1 == oarg || rhs1 == oarg2) + ok = true; + /* Handle e.g. + x.18_1 = (short unsigned int) x_5(D); + _2 = (int) x.18_1; + _3 = __atomic_fetch_xor_2 (&vshort, _2, 0); + _4 = (short int) _3; + _8 = x_5(D) ^ _4; + This happens only for char/short. */ + else if (TREE_CODE (rhs1) == SSA_NAME + && INTEGRAL_TYPE_P (TREE_TYPE (rhs1)) + && (TYPE_PRECISION (TREE_TYPE (rhs1)) + == TYPE_PRECISION (TREE_TYPE (lhs2)))) + { + g = SSA_NAME_DEF_STMT (rhs1); + if (gimple_assign_cast_p (g) + && (gimple_assign_rhs1 (g) == oarg + || gimple_assign_rhs1 (g) == oarg2)) + ok = true; + } + } + if (!ok && arg == oarg2) + /* Handle e.g. + _1 = __sync_fetch_and_add_4 (&v, x_5(D)); + _2 = (int) _1; + x.0_3 = (int) x_5(D); + _7 = _2 + x.0_3; */ + ok = true; + } + } + + if (ok) + { + tree new_lhs = make_ssa_name (TREE_TYPE (lhs2)); + gimple_call_set_lhs (stmt2, new_lhs); + gimple_call_set_fndecl (stmt2, ndecl); + gimple_stmt_iterator gsi = gsi_for_stmt (use_stmt); + if (ccode == ERROR_MARK) + gimple_assign_set_rhs_with_ops (&gsi, cast_stmt + ? NOP_EXPR : SSA_NAME, + new_lhs); + else + { + crhs1 = new_lhs; + crhs2 = build_zero_cst (TREE_TYPE (lhs2)); + if (gimple_code (use_stmt) == GIMPLE_COND) + { + gcond *cond_stmt = as_a (use_stmt); + gimple_cond_set_lhs (cond_stmt, crhs1); + gimple_cond_set_rhs (cond_stmt, crhs2); + } + else if (gimple_assign_rhs_class (use_stmt) + == GIMPLE_BINARY_RHS) + { + gimple_assign_set_rhs1 (use_stmt, crhs1); + gimple_assign_set_rhs2 (use_stmt, crhs2); + } + else + { + gcc_checking_assert (gimple_assign_rhs_code (use_stmt) + == COND_EXPR); + tree cond = build2 (ccode, boolean_type_node, + crhs1, crhs2); + gimple_assign_set_rhs1 (use_stmt, cond); + } + } + update_stmt (use_stmt); + /* & and | aren't reversible. */ + if (atomic_op != BIT_AND_EXPR + && atomic_op != BIT_IOR_EXPR + /* With -fnon-call-exceptions if we can't + add stmts after the call easily. */ + && !stmt_ends_bb_p (stmt2)) + { + gsi = gsi_for_stmt (stmt2); + tree type = TREE_TYPE (lhs2); + if (TREE_CODE (arg) == INTEGER_CST) + arg = fold_convert (type, arg); + else if (!useless_type_conversion_p (type, TREE_TYPE (arg))) + { + tree narg = make_ssa_name (type); + gimple *g = gimple_build_assign (narg, NOP_EXPR, arg); + gsi_insert_after (&gsi, g, GSI_NEW_STMT); + arg = narg; + } + enum tree_code rcode; + switch (atomic_op) + { + case PLUS_EXPR: rcode = MINUS_EXPR; break; + case MINUS_EXPR: rcode = PLUS_EXPR; break; + case BIT_XOR_EXPR: rcode = atomic_op; break; + default: gcc_unreachable (); + } + gimple *g = gimple_build_assign (lhs2, rcode, new_lhs, arg); + gsi_insert_after (&gsi, g, GSI_NEW_STMT); + update_stmt (stmt2); + } + else + { + imm_use_iterator iter; + FOR_EACH_IMM_USE_STMT (use_stmt, iter, lhs2) + if (use_stmt != cast_stmt) + { + gcc_assert (is_gimple_debug (use_stmt)); + gimple_debug_bind_reset_value (use_stmt); + update_stmt (use_stmt); + } + if (cast_stmt) + { + gsi = gsi_for_stmt (cast_stmt); + gsi_remove (&gsi, true); + } + update_stmt (stmt2); + release_ssa_name (lhs2); + } + } + } + break; + default: break; } --- gcc/testsuite/gcc.dg/tree-ssa/pr98737-1.c.jj 2022-01-12 14:48:45.743941426 +0100 +++ gcc/testsuite/gcc.dg/tree-ssa/pr98737-1.c 2022-01-12 16:36:54.228346979 +0100 @@ -0,0 +1,148 @@ +/* PR target/98737 */ +/* { dg-do compile { target i?86-*-* x86_64-*-* powerpc*-*-* aarch64*-*-* } } */ +/* { dg-options "-O2 -fdump-tree-optimized -fcompare-debug" } */ +/* { dg-additional-options "-march=i686" { target ia32 } } */ +/* { dg-final { scan-tree-dump-not "__atomic_fetch_" "optimized" } } */ +/* { dg-final { scan-tree-dump-not "__sync_fetch_and_" "optimized" } } */ + +typedef signed char schar; +typedef unsigned long ulong; +typedef unsigned int uint; +typedef unsigned short ushort; +typedef unsigned char uchar; +long vlong; +int vint; +short vshort; +schar vschar; +ulong vulong; +uint vuint; +ushort vushort; +uchar vuchar; +#define A(n, t, ut, f, o, ...) \ +t fn##n (t x) \ +{ \ + ut z = f (&v##t, x, ##__VA_ARGS__); \ + t w = (t) z; \ + return w o x; \ +} +#define B(n, f, o, ...) \ + A(n##0, long, ulong, f, o, ##__VA_ARGS__) \ + A(n##1, int, uint, f, o, ##__VA_ARGS__) \ + A(n##2, short, ushort, f, o, ##__VA_ARGS__) \ + A(n##3, schar, uchar, f, o, ##__VA_ARGS__) \ + A(n##4, ulong, ulong, f, o, ##__VA_ARGS__) \ + A(n##5, uint, uint, f, o, ##__VA_ARGS__) \ + A(n##6, ushort, ushort, f, o, ##__VA_ARGS__) \ + A(n##7, uchar, uchar, f, o, ##__VA_ARGS__) + +B(00, __atomic_fetch_add, +, __ATOMIC_RELAXED) +B(01, __atomic_fetch_sub, -, __ATOMIC_RELAXED) +B(02, __atomic_fetch_and, &, __ATOMIC_RELAXED) +B(03, __atomic_fetch_xor, ^, __ATOMIC_RELAXED) +B(04, __atomic_fetch_or, |, __ATOMIC_RELAXED) +B(05, __sync_fetch_and_add, +) +B(06, __sync_fetch_and_sub, -) +B(07, __sync_fetch_and_and, &) +B(08, __sync_fetch_and_xor, ^) +B(09, __sync_fetch_and_or, |) + +#undef A +#define A(n, t, ut, f, o, ...) \ +t fn##n (void) \ +{ \ + ut z = f (&v##t, 42, ##__VA_ARGS__); \ + t w = (t) z; \ + return w o 42; \ +} + +B(10, __atomic_fetch_add, +, __ATOMIC_RELAXED) +B(11, __atomic_fetch_sub, -, __ATOMIC_RELAXED) +B(12, __atomic_fetch_and, &, __ATOMIC_RELAXED) +B(13, __atomic_fetch_xor, ^, __ATOMIC_RELAXED) +B(14, __atomic_fetch_or, |, __ATOMIC_RELAXED) +B(15, __sync_fetch_and_add, +) +B(16, __sync_fetch_and_sub, -) +B(17, __sync_fetch_and_and, &) +B(18, __sync_fetch_and_xor, ^) +B(19, __sync_fetch_and_or, |) + +#undef A +#define A(n, t, ut, f, o, ...) \ +t fn##n (t x) \ +{ \ + ut z = f (&v##t, x, ##__VA_ARGS__); \ + t w = (t) z; \ + t v = w o x; \ + return v == 0; \ +} + +B(20, __atomic_fetch_add, +, __ATOMIC_RELAXED) +B(21, __atomic_fetch_sub, -, __ATOMIC_RELAXED) +B(22, __atomic_fetch_and, &, __ATOMIC_RELAXED) +B(23, __atomic_fetch_xor, ^, __ATOMIC_RELAXED) +B(24, __atomic_fetch_or, |, __ATOMIC_RELAXED) +B(25, __sync_fetch_and_add, +) +B(26, __sync_fetch_and_sub, -) +B(27, __sync_fetch_and_and, &) +B(28, __sync_fetch_and_xor, ^) +B(29, __sync_fetch_and_or, |) + +#undef A +#define A(n, t, ut, f, o, ...) \ +t fn##n (void) \ +{ \ + ut z = f (&v##t, 42, ##__VA_ARGS__); \ + t w = (t) z; \ + t v = w o 42; \ + return v != 0; \ +} + +B(30, __atomic_fetch_add, +, __ATOMIC_RELAXED) +B(31, __atomic_fetch_sub, -, __ATOMIC_RELAXED) +B(32, __atomic_fetch_and, &, __ATOMIC_RELAXED) +B(33, __atomic_fetch_xor, ^, __ATOMIC_RELAXED) +B(34, __atomic_fetch_or, |, __ATOMIC_RELAXED) +B(35, __sync_fetch_and_add, +) +B(36, __sync_fetch_and_sub, -) +B(37, __sync_fetch_and_and, &) +B(38, __sync_fetch_and_xor, ^) +B(39, __sync_fetch_and_or, |) + +#undef A +#define A(n, t, ut, f, o, ...) \ +t fn##n (t x) \ +{ \ + return (t) (((t) f (&v##t, x, ##__VA_ARGS__)) \ + o x) != 0; \ +} + +B(40, __atomic_fetch_add, +, __ATOMIC_RELAXED) +B(41, __atomic_fetch_sub, -, __ATOMIC_RELAXED) +B(42, __atomic_fetch_and, &, __ATOMIC_RELAXED) +B(43, __atomic_fetch_xor, ^, __ATOMIC_RELAXED) +B(44, __atomic_fetch_or, |, __ATOMIC_RELAXED) +B(45, __sync_fetch_and_add, +) +B(46, __sync_fetch_and_sub, -) +B(47, __sync_fetch_and_and, &) +B(48, __sync_fetch_and_xor, ^) +B(49, __sync_fetch_and_or, |) + +#undef A +#define A(n, t, ut, f, o, ...) \ +t fn##n (void) \ +{ \ + return (t) (((t) f (&v##t, 42, ##__VA_ARGS__))\ + o 42) == 0; \ +} + +B(50, __atomic_fetch_add, +, __ATOMIC_RELAXED) +B(51, __atomic_fetch_sub, -, __ATOMIC_RELAXED) +B(52, __atomic_fetch_and, &, __ATOMIC_RELAXED) +B(53, __atomic_fetch_xor, ^, __ATOMIC_RELAXED) +/* (whatever | 42) == 0 is 0, so we can't test this. */ +/* B(54, __atomic_fetch_or, |, __ATOMIC_RELAXED) */ +B(55, __sync_fetch_and_add, +) +B(56, __sync_fetch_and_sub, -) +B(57, __sync_fetch_and_and, &) +B(58, __sync_fetch_and_xor, ^) +/* B(59, __sync_fetch_and_or, |) */ --- gcc/testsuite/gcc.dg/tree-ssa/pr98737-2.c.jj 2022-01-12 16:43:29.411766485 +0100 +++ gcc/testsuite/gcc.dg/tree-ssa/pr98737-2.c 2022-01-12 16:41:24.301534958 +0100 @@ -0,0 +1,123 @@ +/* PR target/98737 */ +/* { dg-do compile { target i?86-*-* x86_64-*-* powerpc*-*-* aarch64*-*-* } } */ +/* { dg-options "-O2 -fdump-tree-optimized -fcompare-debug" } */ +/* { dg-additional-options "-march=i686" { target ia32 } } */ +/* { dg-final { scan-tree-dump-not "__atomic_\[^f]" "optimized" } } */ +/* { dg-final { scan-tree-dump-not "__sync_\[^f]" "optimized" } } */ + +typedef signed char schar; +typedef unsigned long ulong; +typedef unsigned int uint; +typedef unsigned short ushort; +typedef unsigned char uchar; +long vlong; +int vint; +short vshort; +schar vschar; +ulong vulong; +uint vuint; +ushort vushort; +uchar vuchar; +#define A(n, t, ut, f, o, ...) \ +t fn##n (t x) \ +{ \ + ut z = f (&v##t, x, ##__VA_ARGS__); \ + t w = (t) z; \ + return w o x; \ +} +#define B(n, f, o, ...) \ + A(n##0, long, ulong, f, o, ##__VA_ARGS__) \ + A(n##1, int, uint, f, o, ##__VA_ARGS__) \ + A(n##2, short, ushort, f, o, ##__VA_ARGS__) \ + A(n##3, schar, uchar, f, o, ##__VA_ARGS__) \ + A(n##4, ulong, ulong, f, o, ##__VA_ARGS__) \ + A(n##5, uint, uint, f, o, ##__VA_ARGS__) \ + A(n##6, ushort, ushort, f, o, ##__VA_ARGS__) \ + A(n##7, uchar, uchar, f, o, ##__VA_ARGS__) + +B(00, __atomic_add_fetch, -, __ATOMIC_RELAXED) +B(01, __atomic_sub_fetch, +, __ATOMIC_RELAXED) +B(03, __atomic_xor_fetch, ^, __ATOMIC_RELAXED) +B(05, __sync_add_and_fetch, -) +B(06, __sync_sub_and_fetch, +) +B(08, __sync_xor_and_fetch, ^) + +#undef A +#define A(n, t, ut, f, o, ...) \ +t fn##n (void) \ +{ \ + ut z = f (&v##t, 42, ##__VA_ARGS__); \ + t w = (t) z; \ + return w o 42; \ +} + +B(10, __atomic_add_fetch, -, __ATOMIC_RELAXED) +B(11, __atomic_sub_fetch, +, __ATOMIC_RELAXED) +B(13, __atomic_xor_fetch, ^, __ATOMIC_RELAXED) +B(15, __sync_add_and_fetch, -) +B(16, __sync_sub_and_fetch, +) +B(18, __sync_xor_and_fetch, ^) + +#undef A +#define A(n, t, ut, f, o, ...) \ +t fn##n (t x) \ +{ \ + ut z = f (&v##t, x, ##__VA_ARGS__); \ + t w = (t) z; \ + t v = w o x; \ + return v == 0; \ +} + +B(20, __atomic_add_fetch, -, __ATOMIC_RELAXED) +B(21, __atomic_sub_fetch, +, __ATOMIC_RELAXED) +B(23, __atomic_xor_fetch, ^, __ATOMIC_RELAXED) +B(25, __sync_add_and_fetch, -) +B(26, __sync_sub_and_fetch, +) +B(28, __sync_xor_and_fetch, ^) + +#undef A +#define A(n, t, ut, f, o, ...) \ +t fn##n (void) \ +{ \ + ut z = f (&v##t, 42, ##__VA_ARGS__); \ + t w = (t) z; \ + t v = w o 42; \ + return v != 0; \ +} + +B(30, __atomic_add_fetch, -, __ATOMIC_RELAXED) +B(31, __atomic_sub_fetch, +, __ATOMIC_RELAXED) +B(33, __atomic_xor_fetch, ^, __ATOMIC_RELAXED) +B(35, __sync_add_and_fetch, -) +B(36, __sync_sub_and_fetch, +) +B(38, __sync_xor_and_fetch, ^) + +#undef A +#define A(n, t, ut, f, o, ...) \ +t fn##n (t x) \ +{ \ + return (t) (((t) f (&v##t, x, ##__VA_ARGS__)) \ + o x) != 0; \ +} + +B(40, __atomic_add_fetch, -, __ATOMIC_RELAXED) +B(41, __atomic_sub_fetch, +, __ATOMIC_RELAXED) +B(43, __atomic_xor_fetch, ^, __ATOMIC_RELAXED) +B(45, __sync_add_and_fetch, -) +B(46, __sync_sub_and_fetch, +) +B(48, __sync_xor_and_fetch, ^) + +#undef A +#define A(n, t, ut, f, o, ...) \ +t fn##n (void) \ +{ \ + return (t) (((t) f (&v##t, 42, ##__VA_ARGS__))\ + o 42) == 0; \ +} + +B(50, __atomic_add_fetch, -, __ATOMIC_RELAXED) +B(51, __atomic_sub_fetch, +, __ATOMIC_RELAXED) +B(53, __atomic_xor_fetch, ^, __ATOMIC_RELAXED) +B(55, __sync_add_and_fetch, -) +B(56, __sync_sub_and_fetch, +) +B(58, __sync_xor_and_fetch, ^)