{"id":2226301,"url":"http://patchwork.ozlabs.org/api/1.2/patches/2226301/?format=json","web_url":"http://patchwork.ozlabs.org/project/gcc/patch/bmm.hhubrmqub2.gcc.gcc-TEST.redi.31.1.14@forge-stage.sourceware.org/","project":{"id":17,"url":"http://patchwork.ozlabs.org/api/1.2/projects/17/?format=json","name":"GNU Compiler Collection","link_name":"gcc","list_id":"gcc-patches.gcc.gnu.org","list_email":"gcc-patches@gcc.gnu.org","web_url":null,"scm_url":null,"webscm_url":null,"list_archive_url":"","list_archive_url_format":"","commit_url_format":""},"msgid":"","list_archive_url":null,"date":"2026-04-22T10:44:29","name":"[v1,14/16] libstdc++: Create new base class of std::barrier for non-dependent code","commit_ref":null,"pull_url":null,"state":"new","archived":false,"hash":"d762f4d0b29a2335e65c9ddf603df04b6b2e441f","submitter":{"id":93210,"url":"http://patchwork.ozlabs.org/api/1.2/people/93210/?format=json","name":"Jonathan Wakely via Sourceware Forge","email":"forge-bot+redi@forge-stage.sourceware.org"},"delegate":null,"mbox":"http://patchwork.ozlabs.org/project/gcc/patch/bmm.hhubrmqub2.gcc.gcc-TEST.redi.31.1.14@forge-stage.sourceware.org/mbox/","series":[{"id":500987,"url":"http://patchwork.ozlabs.org/api/1.2/series/500987/?format=json","web_url":"http://patchwork.ozlabs.org/project/gcc/list/?series=500987","date":"2026-04-22T10:44:17","name":"atomic wait/notify ABI stabilization","version":1,"mbox":"http://patchwork.ozlabs.org/series/500987/mbox/"}],"comments":"http://patchwork.ozlabs.org/api/patches/2226301/comments/","check":"pending","checks":"http://patchwork.ozlabs.org/api/patches/2226301/checks/","tags":{},"related":[],"headers":{"Return-Path":"","X-Original-To":["incoming@patchwork.ozlabs.org","gcc-patches@gcc.gnu.org"],"Delivered-To":["patchwork-incoming@legolas.ozlabs.org","gcc-patches@gcc.gnu.org"],"Authentication-Results":["legolas.ozlabs.org;\n spf=pass (sender SPF authorized) smtp.mailfrom=gcc.gnu.org\n (client-ip=2620:52:6:3111::32; helo=vm01.sourceware.org;\n envelope-from=gcc-patches-bounces~incoming=patchwork.ozlabs.org@gcc.gnu.org;\n receiver=patchwork.ozlabs.org)","sourceware.org; dmarc=none (p=none dis=none)\n header.from=forge-stage.sourceware.org","sourceware.org;\n spf=pass smtp.mailfrom=forge-stage.sourceware.org","server2.sourceware.org;\n arc=none smtp.remote-ip=38.145.34.39"],"Received":["from vm01.sourceware.org (vm01.sourceware.org\n [IPv6:2620:52:6:3111::32])\n\t(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)\n\t key-exchange x25519 server-signature ECDSA (secp384r1) server-digest SHA384)\n\t(No client certificate requested)\n\tby legolas.ozlabs.org (Postfix) with ESMTPS id 4g0xcY4BGLz1yD5\n\tfor ; Wed, 22 Apr 2026 21:21:17 +1000 (AEST)","from vm01.sourceware.org (localhost [127.0.0.1])\n\tby sourceware.org (Postfix) with ESMTP id B2AD742DE31D\n\tfor ; Wed, 22 Apr 2026 11:21:15 +0000 (GMT)","from forge-stage.sourceware.org (vm08.sourceware.org [38.145.34.39])\n by sourceware.org (Postfix) with ESMTPS id 737704015EB8\n for ; Wed, 22 Apr 2026 10:46:09 +0000 (GMT)","from forge-stage.sourceware.org (localhost [IPv6:::1])\n (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)\n key-exchange x25519 server-signature ECDSA (prime256v1) server-digest SHA256)\n (No client certificate requested)\n by forge-stage.sourceware.org (Postfix) with ESMTPS id 4229842BBD\n for ; Wed, 22 Apr 2026 10:46:02 +0000 (UTC)"],"DKIM-Filter":["OpenDKIM Filter v2.11.0 sourceware.org B2AD742DE31D","OpenDKIM Filter v2.11.0 sourceware.org 737704015EB8"],"DMARC-Filter":"OpenDMARC Filter v1.4.2 sourceware.org 737704015EB8","ARC-Filter":"OpenARC Filter v1.0.0 sourceware.org 737704015EB8","ARC-Seal":"i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1776854769; cv=none;\n b=wfV3F2FBaPNTC7bpX19X64w3qcNNrX1RUy26TuA40kUEr+fLjwwApcmi8DPEVpZI3RDWEkJHQAplc+CTS7TvCTTqnHsSnQ44UVculb42qNhusQKaKVE0npOkWAPdkgGHCoGkIh4gVSt79cvlSej/DJ5HyJo+n2RsrJkFyFu41j8=","ARC-Message-Signature":"i=1; a=rsa-sha256; d=sourceware.org; s=key;\n t=1776854769; c=relaxed/simple;\n bh=HhyP15LknaSfnZH4eQ3oWjdn5Xu0uABVzT56FUmGNSA=;\n h=From:Date:Subject:To:Message-ID;\n b=eEq99Tzozhl4hrnJe06lDvYwaQyl7Ku7Pj35D0IEjGSAldhD/KcINzhAdeL3zfFmn0Wcsr+p2MVnRoIdIjgGP8uS8B9PXbF56hCTBdnjC0TjnwJS+j86xtn5f7mieqFg6jmUikhi8aNMlaaocudUWRlWDegrdLUH2vJ1c5ncIOM=","ARC-Authentication-Results":"i=1; server2.sourceware.org","From":"Jonathan Wakely via Sourceware Forge\n ","Date":"Wed, 22 Apr 2026 10:44:29 +0000","Subject":"[PATCH v1 14/16] libstdc++: Create new base class of std::barrier\n for non-dependent code","To":"gcc-patches mailing list ","Message-ID":"\n ","X-Mailer":"batrachomyomachia","X-Pull-Request-Organization":"gcc","X-Pull-Request-Repository":"gcc-TEST","X-Pull-Request":"https://forge.sourceware.org/gcc/gcc-TEST/pulls/31","References":"\n ","In-Reply-To":"\n ","X-Patch-URL":"\n https://forge.sourceware.org/redi/gcc/commit/4bc727565965551ce34119ecc8ea66bc285e60dc","X-BeenThere":"gcc-patches@gcc.gnu.org","X-Mailman-Version":"2.1.30","Precedence":"list","List-Id":"Gcc-patches mailing list ","List-Unsubscribe":",\n ","List-Archive":"","List-Post":"","List-Help":"","List-Subscribe":",\n ","Reply-To":"gcc-patches mailing list , redi@gcc.gnu.org","Errors-To":"gcc-patches-bounces~incoming=patchwork.ozlabs.org@gcc.gnu.org"},"content":"From: Jonathan Wakely \n\nThis moves all non-dependent state and logic for std::barrier into a\nnew non-template base class, to avoid template bloat.\n\nThis would permit moving the _M_arrive function into the library instead\nof the header.\n\nlibstdc++-v3/ChangeLog:\n\n\t* include/std/barrier (__tree_barrier_base): New class.\n\t(__tree_barrier): Move non-dependent code into\n\t__tree_barrier_base and derive from it.\n---\n libstdc++-v3/include/std/barrier | 170 +++++++++++++++++--------------\n 1 file changed, 91 insertions(+), 79 deletions(-)","diff":"diff --git a/libstdc++-v3/include/std/barrier b/libstdc++-v3/include/std/barrier\nindex 9c1de411f9ce..56270c99e056 100644\n--- a/libstdc++-v3/include/std/barrier\n+++ b/libstdc++-v3/include/std/barrier\n@@ -81,76 +81,101 @@ It looks different from literature pseudocode for two main reasons:\n \n enum class __barrier_phase_t : unsigned char { };\n \n- template\n- class __tree_barrier\n+ struct __tree_barrier_base\n+ {\n+ static constexpr ptrdiff_t\n+ max() noexcept\n+ { return __PTRDIFF_MAX__ - 1; }\n+\n+ protected:\n+ using __atomic_phase_ref_t = std::__atomic_ref<__barrier_phase_t>;\n+ using __atomic_phase_const_ref_t = std::__atomic_ref;\n+ static constexpr auto __phase_alignment =\n+\t\t __atomic_phase_ref_t::required_alignment;\n+\n+ using __tickets_t = std::array<__barrier_phase_t, 64>;\n+ struct alignas(64) /* naturally-align the heap state */ __state_t\n {\n- using __atomic_phase_ref_t = std::__atomic_ref<__barrier_phase_t>;\n- using __atomic_phase_const_ref_t = std::__atomic_ref;\n- static constexpr auto __phase_alignment =\n-\t\t __atomic_phase_ref_t::required_alignment;\n+ alignas(__phase_alignment) __tickets_t __tickets;\n+ };\n \n- using __tickets_t = std::array<__barrier_phase_t, 64>;\n- struct alignas(64) /* naturally-align the heap state */ __state_t\n- {\n-\talignas(__phase_alignment) __tickets_t __tickets;\n- };\n+ ptrdiff_t _M_expected;\n+ __atomic_base<__state_t*> _M_state{nullptr};\n+ __atomic_base _M_expected_adjustment{0};\n+ alignas(__phase_alignment) __barrier_phase_t _M_phase{};\n \n- ptrdiff_t _M_expected;\n- __atomic_base<__state_t*> _M_state{nullptr};\n- __atomic_base _M_expected_adjustment{0};\n- [[no_unique_address]] _CompletionF _M_completion;\n+ explicit constexpr\n+ __tree_barrier_base(ptrdiff_t __expected)\n+ : _M_expected(__expected)\n+ {\n+ __glibcxx_assert(__expected >= 0 && __expected <= max());\n \n- alignas(__phase_alignment) __barrier_phase_t _M_phase{};\n+ if (!std::is_constant_evaluated())\n+\t_M_state.store(_M_alloc_state().release(), memory_order_release);\n+ }\n \n- bool\n- _M_arrive(__barrier_phase_t __old_phase, size_t __current)\n- {\n-\tconst auto __old_phase_val = static_cast(__old_phase);\n-\tconst auto __half_step =\n-\t\t\t static_cast<__barrier_phase_t>(__old_phase_val + 1);\n-\tconst auto __full_step =\n-\t\t\t static_cast<__barrier_phase_t>(__old_phase_val + 2);\n-\n-\tsize_t __current_expected = _M_expected;\n-\t__current %= ((_M_expected + 1) >> 1);\n+ unique_ptr<__state_t[]>\n+ _M_alloc_state()\n+ {\n+ size_t const __count = (_M_expected + 1) >> 1;\n+ return std::make_unique<__state_t[]>(__count);\n+ }\n \n-\t__state_t* const __state = _M_state.load(memory_order_relaxed);\n+ bool\n+ _M_arrive(__barrier_phase_t __old_phase, size_t __current)\n+ {\n+ const auto __old_phase_val = static_cast(__old_phase);\n+ const auto __half_step =\n+\t\t\t static_cast<__barrier_phase_t>(__old_phase_val + 1);\n+ const auto __full_step =\n+\t\t\t static_cast<__barrier_phase_t>(__old_phase_val + 2);\n+\n+ size_t __current_expected = _M_expected;\n+ __current %= ((_M_expected + 1) >> 1);\n+\n+ __state_t* const __state = _M_state.load(memory_order_relaxed);\n+\n+ for (int __round = 0; ; ++__round)\n+\t{\n+\t if (__current_expected <= 1)\n+\t return true;\n+\t size_t const __end_node = ((__current_expected + 1) >> 1),\n+\t\t __last_node = __end_node - 1;\n+\t for ( ; ; ++__current)\n+\t {\n+\t if (__current == __end_node)\n+\t\t__current = 0;\n+\t auto __expect = __old_phase;\n+\t __atomic_phase_ref_t __phase(__state[__current]\n+\t\t\t\t\t .__tickets[__round]);\n+\t if (__current == __last_node && (__current_expected & 1))\n+\t\t{\n+\t\t if (__phase.compare_exchange_strong(__expect, __full_step,\n+\t\t\t\t\t\t memory_order_acq_rel))\n+\t\t break; // I'm 1 in 1, go to next __round\n+\t\t}\n+\t else if (__phase.compare_exchange_strong(__expect, __half_step,\n+\t\t\t\t\t\t memory_order_acq_rel))\n+\t\t{\n+\t\t return false; // I'm 1 in 2, done with arrival\n+\t\t}\n+\t else if (__expect == __half_step)\n+\t\t{\n+\t\t if (__phase.compare_exchange_strong(__expect, __full_step,\n+\t\t\t\t\t\t memory_order_acq_rel))\n+\t\t break; // I'm 2 in 2, go to next __round\n+\t\t}\n+\t }\n+\t __current_expected = __last_node + 1;\n+\t __current >>= 1;\n+\t}\n+ }\n+ };\n \n-\tfor (int __round = 0; ; ++__round)\n-\t {\n-\t if (__current_expected <= 1)\n-\t\treturn true;\n-\t size_t const __end_node = ((__current_expected + 1) >> 1),\n-\t\t\t __last_node = __end_node - 1;\n-\t for ( ; ; ++__current)\n-\t {\n-\t\tif (__current == __end_node)\n-\t\t __current = 0;\n-\t\tauto __expect = __old_phase;\n-\t\t__atomic_phase_ref_t __phase(__state[__current]\n-\t\t\t\t\t\t.__tickets[__round]);\n-\t\tif (__current == __last_node && (__current_expected & 1))\n-\t\t {\n-\t\t if (__phase.compare_exchange_strong(__expect, __full_step,\n-\t\t\t\t\t\t memory_order_acq_rel))\n-\t\t break; // I'm 1 in 1, go to next __round\n-\t\t }\n-\t\telse if (__phase.compare_exchange_strong(__expect, __half_step,\n-\t\t\t\t\t\t memory_order_acq_rel))\n-\t\t {\n-\t\t return false; // I'm 1 in 2, done with arrival\n-\t\t }\n-\t\telse if (__expect == __half_step)\n-\t\t {\n-\t\t if (__phase.compare_exchange_strong(__expect, __full_step,\n-\t\t\t\t\t\t memory_order_acq_rel))\n-\t\t break; // I'm 2 in 2, go to next __round\n-\t\t }\n-\t }\n-\t __current_expected = __last_node + 1;\n-\t __current >>= 1;\n-\t }\n- }\n+ template\n+ class __tree_barrier : public __tree_barrier_base\n+ {\n+ [[no_unique_address]] _CompletionF _M_completion;\n \n // _GLIBCXX_RESOLVE_LIB_DEFECTS\n // 3898. Possibly unintended preconditions for completion functions\n@@ -159,22 +184,10 @@ It looks different from literature pseudocode for two main reasons:\n public:\n using arrival_token = __barrier_phase_t;\n \n- static constexpr ptrdiff_t\n- max() noexcept\n- { return __PTRDIFF_MAX__ - 1; }\n-\n constexpr\n __tree_barrier(ptrdiff_t __expected, _CompletionF __completion)\n- : _M_expected(__expected), _M_completion(std::move(__completion))\n- {\n-\t__glibcxx_assert(__expected >= 0 && __expected <= max());\n-\n-\tif (!std::is_constant_evaluated())\n-\t {\n-\t size_t const __count = (_M_expected + 1) >> 1;\n-\t _M_state.store(new __state_t[__count], memory_order_release);\n-\t }\n- }\n+ : __tree_barrier_base(__expected), _M_completion(std::move(__completion))\n+ { }\n \n [[nodiscard]] arrival_token\n arrive(ptrdiff_t __update)\n@@ -191,8 +204,7 @@ It looks different from literature pseudocode for two main reasons:\n \n \tif (__cur == 0 && !_M_state.load(memory_order_relaxed)) [[unlikely]]\n \t {\n-\t size_t const __count = (_M_expected + 1) >> 1;\n-\t auto __p = make_unique<__state_t[]>(__count);\n+\t auto __p = _M_alloc_state();\n \t __state_t* __val = nullptr;\n \t if (_M_state.compare_exchange_strong(__val, __p.get(),\n \t\t\t\t\t\t memory_order_seq_cst,\n","prefixes":["v1","14/16"]}