From patchwork Tue Nov 3 13:09:09 2015 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Jerin Jacob X-Patchwork-Id: 8589 Return-Path: X-Original-To: patchwork@dpdk.org Delivered-To: patchwork@dpdk.org Received: from [92.243.14.124] (localhost [IPv6:::1]) by dpdk.org (Postfix) with ESMTP id 5F51091D6; Tue, 3 Nov 2015 14:10:34 +0100 (CET) Received: from na01-bn1-obe.outbound.protection.outlook.com (mail-bn1bon0067.outbound.protection.outlook.com [157.56.111.67]) by dpdk.org (Postfix) with ESMTP id 6383A8F9D for ; Tue, 3 Nov 2015 14:10:32 +0100 (CET) Authentication-Results: spf=none (sender IP is ) smtp.mailfrom=Jerin.Jacob@caviumnetworks.com; Received: from localhost.caveonetworks.com (111.93.218.67) by BY2PR0701MB1974.namprd07.prod.outlook.com (10.163.155.20) with Microsoft SMTP Server (TLS) id 15.1.312.18; Tue, 3 Nov 2015 13:10:29 +0000 From: Jerin Jacob To: Date: Tue, 3 Nov 2015 18:39:09 +0530 Message-ID: <1446556153-18845-9-git-send-email-jerin.jacob@caviumnetworks.com> X-Mailer: git-send-email 2.1.0 In-Reply-To: <1446556153-18845-8-git-send-email-jerin.jacob@caviumnetworks.com> References: <1446556153-18845-1-git-send-email-jerin.jacob@caviumnetworks.com> <1446556153-18845-2-git-send-email-jerin.jacob@caviumnetworks.com> <1446556153-18845-3-git-send-email-jerin.jacob@caviumnetworks.com> <1446556153-18845-4-git-send-email-jerin.jacob@caviumnetworks.com> <1446556153-18845-5-git-send-email-jerin.jacob@caviumnetworks.com> <1446556153-18845-6-git-send-email-jerin.jacob@caviumnetworks.com> <1446556153-18845-7-git-send-email-jerin.jacob@caviumnetworks.com> <1446556153-18845-8-git-send-email-jerin.jacob@caviumnetworks.com> MIME-Version: 1.0 X-Originating-IP: [111.93.218.67] X-ClientProxiedBy: MAXPR01CA0048.INDPRD01.PROD.OUTLOOK.COM (25.164.146.148) To BY2PR0701MB1974.namprd07.prod.outlook.com (25.163.155.20) X-Microsoft-Exchange-Diagnostics: 1; BY2PR0701MB1974; 2:A+rzhi6vt0Y7GaAo2L73Mi7dlOIOfRH+r+N2+kWY0xkIzzWk1bp4vIyiJYj9wafYNjPnrdSwnHwZQCZoimntFw0pg7AXcHwkilYxgRAsVJ/TlHAykEuoDuFt5RJVuEJuHx+S9hWZqDOijjnlSKeeMIkx3h0EBn9tXz5nmVJ1kHk=; 3:LuUFMHNaj274VYEBuu4rVsJMi7NMelfLw9jY8zaZuIq0YaSOZnhWoK3hc7WCIYyPKs6QIrGlSlT/tgNVmZFrJ7bnLDikuH6nY0SG1HviClVXinQfBLVH9lp/UBQWs9CW3FUMpl9RAucKnbtk8/FTWQ==; 25:j+bMhdPKw/izRbFAcbRqp5uB2MtmDScz9XXS/YUFVQXxCyVpD5id7HRBEacwZ0eLieCb074A2jjtYAT2LbngrkEBsCyuUYHVhGB/EAlNxzUemNEbZMtCUqBobnEIjiTGNyxAUWjXWi/ho18v2Z+qKivRT+NWWeqjWnNZKbKYxyG2b0m1gaHfXKYwNvrvfeGo4KjMWSB4ro+iT7VmXilqRrI2tnwlSRSTGB0aYaSAbkXbVsujtNwptNZxE2kpTm4usjZjgRO8c44f7sG0ibY31g== X-Microsoft-Antispam: UriScan:;BCL:0;PCL:0;RULEID:;SRVR:BY2PR0701MB1974; X-Microsoft-Exchange-Diagnostics: 1; BY2PR0701MB1974; 20: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 X-Microsoft-Antispam-PRVS: X-Exchange-Antispam-Report-Test: UriScan:(236414709691187); X-Exchange-Antispam-Report-CFA-Test: BCL:0; PCL:0; RULEID:(601004)(2401047)(8121501046)(5005006)(520078)(10201501046)(3002001); SRVR:BY2PR0701MB1974; BCL:0; PCL:0; RULEID:; SRVR:BY2PR0701MB1974; X-Microsoft-Exchange-Diagnostics: 1; BY2PR0701MB1974; 4:Uw2sBAk4WdwEcU0X2MyUre0r4gsbawB0xJAHesdokLVUrgHkyufHAmNXr7eP69L94a/a12da/5CYKqCVK+lsZhBs3XHLmBIxGMvf3vlHBcvr816Q6/TsNaD6cOy0J6Hk8xQxS6euqEXoZeBfMQNssTkrN0ZiypUsVg3B4ZgquwXKHrrkjycwLOwda1Oe9s7yCLBaC2VOLAq/DwER7Ss1W4F6/sqMHV4MSSB2GHiMEhPxyM9Vy+yQszS3I6fo1uTuW2LuF0pB111K0uogLj5Ehf/24Joe8al/cKUiycZTJ5z41KkzebWAMmkIMFfp8jlLwYna0rek9rpNvyaRLKBALHYJnjcVfPLSo4tRFagu5p8dJXh64pSPOjpjuHOZWgu7 X-Forefront-PRVS: 0749DC2CE6 X-Forefront-Antispam-Report: SFV:NSPM; SFS:(10009020)(6009001)(6069001)(189002)(199003)(81156007)(5001960100002)(50226001)(50986999)(5004730100002)(2950100001)(110136002)(77096005)(107886002)(66066001)(47776003)(189998001)(5007970100001)(42186005)(19580395003)(86362001)(76506005)(87976001)(69596002)(97736004)(50466002)(101416001)(53416004)(48376002)(106356001)(105586002)(19580405001)(36756003)(5008740100001)(76176999)(40100003)(92566002)(4001430100002)(575784001)(5009440100003)(93886004)(33646002)(122386002)(5003940100001)(229853001)(2351001)(7099028); DIR:OUT; SFP:1101; SCL:1; SRVR:BY2PR0701MB1974; H:localhost.caveonetworks.com; FPR:; SPF:None; PTR:InfoNoRecords; MX:1; A:1; LANG:en; Received-SPF: None (protection.outlook.com: caviumnetworks.com does not designate permitted sender hosts) X-Microsoft-Exchange-Diagnostics: =?us-ascii?Q?1; BY2PR0701MB1974; 23:sUwRdFf4TsDBSfpsu5e04r9bQTuSuZfb/IiagMb?= =?us-ascii?Q?xzXgErcf/MVBINrwlKqdorguEqoLBo1sPmciSoNnxQqKIxypFN2WXjDn662J?= =?us-ascii?Q?pGiOg8lbalJkfS81GpEbfektOhgRSX7yfuaon44b7SeW00tFW3l0zrls3tkB?= =?us-ascii?Q?78Gg50iqL6KmzCie6K+Jjwbzj69LW+ma8G/TrE7YC0avGbqYAFU9WDVmzGx/?= =?us-ascii?Q?i5cfW7jXWbU9IueTVIpY6rFMd1o+wAlNTK3aTz+fLVjKx4U/DcfMXIGcMEmI?= =?us-ascii?Q?s5Rlfq1ZCNtWQC9bUBQammslumfNV4ahkWdmsXFG2lQxrjkikLxEg/grCdLt?= =?us-ascii?Q?BgH3/9TX7g6hTj1a8AQMT+0cGewdPVKjLluMKwUepEjrAZYCIVx7zD+nGk3V?= =?us-ascii?Q?oao37LfQ5DAwdzwQ0CE3vaaYc8M5zabH++ld2TXBW/kXQz7BZGTdAGeQnTiq?= =?us-ascii?Q?WMlMkmx0tHoA8zBcqeJQGFeQhlMEXXuZuGe3FT7YAoCHKW+g8bO5o/iWCw8b?= =?us-ascii?Q?Tl7472vH7JgkFoObe4pnd88Z4kDLAQUHHq0W+CYswEx+bn8xzl40cRWhp+2V?= =?us-ascii?Q?Vhfk7zkTmbLDTJ0HJ3Wch4frRwa5G1ThnS0OhRRjnCyTXquqhN3tw0OWNDPA?= =?us-ascii?Q?AoxLbfyP/SmP6yDzAb/GQsjbXT9hfhJj17KsjzXRREM+fHuipHr03cDBfean?= =?us-ascii?Q?fvopNnAcnNBUvZlOwTxgVZhR45qlHKS+swtUzr1h/KP0geR3X6L04ZajJlU0?= =?us-ascii?Q?akrP9WXgkLib1uUu+q2CDauiSuHLKs69jV5Y3kbQMrXZUp6JjtmOxUlvSgPm?= =?us-ascii?Q?D1kddC555lseDhAuq9lZ8xk/TPoTf1I3kogsxQW1Kg0b+XFMd4vw57pNzwmy?= =?us-ascii?Q?uxq89ua4v1+rQUElPTDZqBXL3ipU7Ju6/stkz3lrA+n9k0WKHqF881nCSNtZ?= =?us-ascii?Q?Jkf1gwLh6zpuEB5QDOxDTc4ZN0JKpiznZ13p9dyLrBvwA1fiUIRYjFvWn7Gs?= =?us-ascii?Q?HKBcddO0mzz4mEDfNEPXxhya74JH2lXcmbYAZ/iLhKmx3ChdanzAEzCBKY2N?= =?us-ascii?Q?cjmqQ4EEu2h6BuUqDvvD//RHI4ln6YQPK8iUBk1PLmIAvoSLCqGP0/cc6O2w?= =?us-ascii?Q?K9Tj+i87SpEqdNmngtBsog8zWIjLfAAxvslzpD3EPTFYXv3158rgDv4wS+D5?= =?us-ascii?Q?T6Jgvr/OlCshJ8hEKDVYQXUIWGRqR4UG6e5sA+eoYRtquk81RWf5sYJcZww?= =?us-ascii?Q?=3D=3D?= X-Microsoft-Exchange-Diagnostics: 1; BY2PR0701MB1974; 5:US33DhLuEGks0OTn7DeRF5Z0ydc/rZbtkWXa2PhlfjLnIThGGO9glfRW4slMChgZ6riTY5rzRQ+Oc/tdreoy/8TTOIze4W7DLlo7wWneNjH5Feba7fwNBiOvbMa2aj1x0M9pa9bdBmvE4rx54188hg==; 24:EgteGPWBW59RdjQAah7LQfK8tHI47Ad+KloPzWpY8817mBa943Gg16VP82+uqsWPNcCgqqZNLPE1IXxv6bxVV7iXuDltlxX2fWVV+TgF+WI=; 20:geblNKDQKt0BMc7iyOXjwTEMbR9r3mMSW9/zsOQ1QIyXuGZFyYzYI+81s0T3VOiQ4B//Nc/Q5KZTDXnG3SAn3A== SpamDiagnosticOutput: 1:23 SpamDiagnosticMetadata: NSPM X-OriginatorOrg: caviumnetworks.com X-MS-Exchange-CrossTenant-OriginalArrivalTime: 03 Nov 2015 13:10:29.0339 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-Transport-CrossTenantHeadersStamped: BY2PR0701MB1974 Subject: [dpdk-dev] [PATCH 08/12] acl: arm64: acl implementation using NEON gcc intrinsic X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" verified with testacl and acl_autotest applications on arm64 architecture. Signed-off-by: Jerin Jacob --- app/test-acl/main.c | 4 + lib/librte_acl/Makefile | 5 + lib/librte_acl/acl.h | 4 + lib/librte_acl/acl_run_neon.c | 46 +++++++ lib/librte_acl/acl_run_neon.h | 290 ++++++++++++++++++++++++++++++++++++++++++ lib/librte_acl/rte_acl.c | 25 ++++ lib/librte_acl/rte_acl.h | 1 + 7 files changed, 375 insertions(+) create mode 100644 lib/librte_acl/acl_run_neon.c create mode 100644 lib/librte_acl/acl_run_neon.h diff --git a/app/test-acl/main.c b/app/test-acl/main.c index 72ce83c..0b0c093 100644 --- a/app/test-acl/main.c +++ b/app/test-acl/main.c @@ -101,6 +101,10 @@ static const struct acl_alg acl_alg[] = { .name = "avx2", .alg = RTE_ACL_CLASSIFY_AVX2, }, + { + .name = "neon", + .alg = RTE_ACL_CLASSIFY_NEON, + }, }; static struct { diff --git a/lib/librte_acl/Makefile b/lib/librte_acl/Makefile index 7a1cf8a..27f91d5 100644 --- a/lib/librte_acl/Makefile +++ b/lib/librte_acl/Makefile @@ -48,9 +48,14 @@ SRCS-$(CONFIG_RTE_LIBRTE_ACL) += rte_acl.c SRCS-$(CONFIG_RTE_LIBRTE_ACL) += acl_bld.c SRCS-$(CONFIG_RTE_LIBRTE_ACL) += acl_gen.c SRCS-$(CONFIG_RTE_LIBRTE_ACL) += acl_run_scalar.c +ifeq ($(CONFIG_RTE_ARCH_ARM64),y) +SRCS-$(CONFIG_RTE_LIBRTE_ACL) += acl_run_neon.c +else SRCS-$(CONFIG_RTE_LIBRTE_ACL) += acl_run_sse.c +endif CFLAGS_acl_run_sse.o += -msse4.1 +CFLAGS_acl_run_neon.o += -flax-vector-conversions -Wno-maybe-uninitialized # # If the compiler supports AVX2 instructions, diff --git a/lib/librte_acl/acl.h b/lib/librte_acl/acl.h index eb4930c..09d6784 100644 --- a/lib/librte_acl/acl.h +++ b/lib/librte_acl/acl.h @@ -230,6 +230,10 @@ int rte_acl_classify_avx2(const struct rte_acl_ctx *ctx, const uint8_t **data, uint32_t *results, uint32_t num, uint32_t categories); +int +rte_acl_classify_neon(const struct rte_acl_ctx *ctx, const uint8_t **data, + uint32_t *results, uint32_t num, uint32_t categories); + #ifdef __cplusplus } #endif /* __cplusplus */ diff --git a/lib/librte_acl/acl_run_neon.c b/lib/librte_acl/acl_run_neon.c new file mode 100644 index 0000000..b014451 --- /dev/null +++ b/lib/librte_acl/acl_run_neon.c @@ -0,0 +1,46 @@ +/* + * BSD LICENSE + * + * Copyright (C) Cavium networks Ltd. 2015. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the + * distribution. + * * Neither the name of Cavium networks nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#include "acl_run_neon.h" + +int +rte_acl_classify_neon(const struct rte_acl_ctx *ctx, const uint8_t **data, + uint32_t *results, uint32_t num, uint32_t categories) +{ + if (likely(num >= 8)) + return search_neon_8(ctx, data, results, num, categories); + else if (num >= 4) + return search_neon_4(ctx, data, results, num, categories); + else + return rte_acl_classify_scalar(ctx, data, results, num, + categories); +} diff --git a/lib/librte_acl/acl_run_neon.h b/lib/librte_acl/acl_run_neon.h new file mode 100644 index 0000000..4579476 --- /dev/null +++ b/lib/librte_acl/acl_run_neon.h @@ -0,0 +1,290 @@ +/* + * BSD LICENSE + * + * Copyright (C) Cavium networks Ltd. 2015. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the + * distribution. + * * Neither the name of Cavium networks nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#include "acl_run.h" +#include "acl_vect.h" + +struct _neon_acl_const { + rte_xmm_t xmm_shuffle_input; + rte_xmm_t xmm_index_mask; + rte_xmm_t range_base; +} neon_acl_const __attribute__((aligned(RTE_CACHE_LINE_SIZE))) = { + { + .u32 = {0x00000000, 0x04040404, 0x08080808, 0x0c0c0c0c} + }, + { + .u32 = {RTE_ACL_NODE_INDEX, RTE_ACL_NODE_INDEX, + RTE_ACL_NODE_INDEX, RTE_ACL_NODE_INDEX} + }, + { + .u32 = {0xffffff00, 0xffffff04, 0xffffff08, 0xffffff0c} + }, +}; + +/* + * Resolve priority for multiple results (neon version). + * This consists comparing the priority of the current traversal with the + * running set of results for the packet. + * For each result, keep a running array of the result (rule number) and + * its priority for each category. + */ +static inline void +resolve_priority_neon(uint64_t transition, int n, const struct rte_acl_ctx *ctx, + struct parms *parms, + const struct rte_acl_match_results *p, + uint32_t categories) +{ + uint32_t x; + int32x4_t results, priority, results1, priority1; + uint32x4_t selector; + int32_t *saved_results, *saved_priority; + + for (x = 0; x < categories; x += RTE_ACL_RESULTS_MULTIPLIER) { + saved_results = (int32_t *)(&parms[n].cmplt->results[x]); + saved_priority = (int32_t *)(&parms[n].cmplt->priority[x]); + + /* get results and priorities for completed trie */ + results = vld1q_s32( + (const int32_t *)&p[transition].results[x]); + priority = vld1q_s32( + (const int32_t *)&p[transition].priority[x]); + + /* if this is not the first completed trie */ + if (parms[n].cmplt->count != ctx->num_tries) { + /* get running best results and their priorities */ + results1 = vld1q_s32(saved_results); + priority1 = vld1q_s32(saved_priority); + + /* select results that are highest priority */ + selector = vcgtq_s32(priority1, priority); + results = vbslq_s32(selector, results1, results); + priority = vbslq_s32(selector, priority1, priority); + } + + /* save running best results and their priorities */ + vst1q_s32(saved_results, results); + vst1q_s32(saved_priority, priority); + } +} + +/* + * Check for any match in 4 transitions + */ +static inline __attribute__((always_inline)) uint32_t +check_any_match_x4(uint64_t val[]) +{ + return ((val[0] | val[1] | val[2] | val[3]) & RTE_ACL_NODE_MATCH); +} + +static inline __attribute__((always_inline)) void +acl_match_check_x4(int slot, const struct rte_acl_ctx *ctx, struct parms *parms, + struct acl_flow_data *flows, uint64_t transitions[]) +{ + while (check_any_match_x4(transitions)) { + transitions[0] = acl_match_check(transitions[0], slot, ctx, + parms, flows, resolve_priority_neon); + transitions[1] = acl_match_check(transitions[1], slot + 1, ctx, + parms, flows, resolve_priority_neon); + transitions[2] = acl_match_check(transitions[2], slot + 2, ctx, + parms, flows, resolve_priority_neon); + transitions[3] = acl_match_check(transitions[3], slot + 3, ctx, + parms, flows, resolve_priority_neon); + } +} + +/* + * Process 4 transitions (in 2 NEON Q registers) in parallel + */ +static inline __attribute__((always_inline)) int32x4_t +transition4(int32x4_t next_input, const uint64_t *trans, uint64_t transitions[]) +{ + int32x4x2_t tr_hi_lo; + int32x4_t t, in, r; + uint32x4_t index_msk, node_type, addr; + uint32x4_t dfa_msk, mask, quad_ofs, dfa_ofs; + + /* Move low 32 into tr_hi_lo.val[0] and high 32 into tr_hi_lo.val[1] */ + tr_hi_lo = vld2q_s32((const int32_t *)transitions); + + /* Calculate the address (array index) for all 4 transitions. */ + + index_msk = vld1q_u32((const uint32_t *)&neon_acl_const.xmm_index_mask); + + /* Calc node type and node addr */ + node_type = vbicq_s32(tr_hi_lo.val[0], index_msk); + addr = vandq_s32(tr_hi_lo.val[0], index_msk); + + /* t = 0 */ + t = veorq_s32(node_type, node_type); + + /* mask for DFA type(0) nodes */ + dfa_msk = vceqq_u32(node_type, t); + + mask = vld1q_s32((const int32_t *)&neon_acl_const.xmm_shuffle_input); + in = vqtbl1q_u8((uint8x16_t)next_input, (uint8x16_t)mask); + + /* DFA calculations. */ + r = vshrq_n_u32(in, 30); /* div by 64 */ + mask = vld1q_s32((const int32_t *)&neon_acl_const.range_base); + r = vaddq_u8(r, mask); + t = vshrq_n_u32(in, 24); + r = vqtbl1q_u8((uint8x16_t)tr_hi_lo.val[1], (uint8x16_t)r); + dfa_ofs = vsubq_s32(t, r); + + /* QUAD/SINGLE calculations. */ + t = vcgtq_s8(in, tr_hi_lo.val[1]); + t = vabsq_s8(t); + t = vpaddlq_u8(t); + quad_ofs = vpaddlq_u16(t); + + /* blend DFA and QUAD/SINGLE. */ + t = vbslq_u8(dfa_msk, dfa_ofs, quad_ofs); + + /* calculate address for next transitions */ + addr = vaddq_u32(addr, t); + + /* Fill next transitions */ + transitions[0] = trans[vgetq_lane_u32(addr, 0)]; + transitions[1] = trans[vgetq_lane_u32(addr, 1)]; + transitions[2] = trans[vgetq_lane_u32(addr, 2)]; + transitions[3] = trans[vgetq_lane_u32(addr, 3)]; + + return vshrq_n_u32(next_input, CHAR_BIT); +} + +/* + * Execute trie traversal with 8 traversals in parallel + */ +static inline int +search_neon_8(const struct rte_acl_ctx *ctx, const uint8_t **data, + uint32_t *results, uint32_t total_packets, uint32_t categories) +{ + int n; + struct acl_flow_data flows; + uint64_t index_array[8]; + struct completion cmplt[8]; + struct parms parms[8]; + int32x4_t input0, input1; + + acl_set_flow(&flows, cmplt, RTE_DIM(cmplt), data, results, + total_packets, categories, ctx->trans_table); + + for (n = 0; n < 8; n++) { + cmplt[n].count = 0; + index_array[n] = acl_start_next_trie(&flows, parms, n, ctx); + } + + /* Check for any matches. */ + acl_match_check_x4(0, ctx, parms, &flows, &index_array[0]); + acl_match_check_x4(4, ctx, parms, &flows, &index_array[4]); + + while (flows.started > 0) { + /* Gather 4 bytes of input data for each stream. */ + input0 = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 0), input0, 0); + input1 = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 4), input1, 0); + + input0 = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 1), input0, 1); + input1 = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 5), input1, 1); + + input0 = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 2), input0, 2); + input1 = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 6), input1, 2); + + input0 = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 3), input0, 3); + input1 = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 7), input1, 3); + + /* Process the 4 bytes of input on each stream. */ + + input0 = transition4(input0, flows.trans, &index_array[0]); + input1 = transition4(input1, flows.trans, &index_array[4]); + + input0 = transition4(input0, flows.trans, &index_array[0]); + input1 = transition4(input1, flows.trans, &index_array[4]); + + input0 = transition4(input0, flows.trans, &index_array[0]); + input1 = transition4(input1, flows.trans, &index_array[4]); + + input0 = transition4(input0, flows.trans, &index_array[0]); + input1 = transition4(input1, flows.trans, &index_array[4]); + + /* Check for any matches. */ + acl_match_check_x4(0, ctx, parms, &flows, &index_array[0]); + acl_match_check_x4(4, ctx, parms, &flows, &index_array[4]); + } + + return 0; +} + +/* + * Execute trie traversal with 4 traversals in parallel + */ +static inline int +search_neon_4(const struct rte_acl_ctx *ctx, const uint8_t **data, + uint32_t *results, int total_packets, uint32_t categories) +{ + int n; + struct acl_flow_data flows; + uint64_t index_array[4]; + struct completion cmplt[4]; + struct parms parms[4]; + int32x4_t input; + + acl_set_flow(&flows, cmplt, RTE_DIM(cmplt), data, results, + total_packets, categories, ctx->trans_table); + + for (n = 0; n < 4; n++) { + cmplt[n].count = 0; + index_array[n] = acl_start_next_trie(&flows, parms, n, ctx); + } + + /* Check for any matches. */ + acl_match_check_x4(0, ctx, parms, &flows, index_array); + + while (flows.started > 0) { + /* Gather 4 bytes of input data for each stream. */ + input = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 0), input, 0); + input = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 1), input, 1); + input = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 2), input, 2); + input = vsetq_lane_s32(GET_NEXT_4BYTES(parms, 3), input, 3); + + /* Process the 4 bytes of input on each stream. */ + input = transition4(input, flows.trans, index_array); + input = transition4(input, flows.trans, index_array); + input = transition4(input, flows.trans, index_array); + input = transition4(input, flows.trans, index_array); + + /* Check for any matches. */ + acl_match_check_x4(0, ctx, parms, &flows, index_array); + } + + return 0; +} + diff --git a/lib/librte_acl/rte_acl.c b/lib/librte_acl/rte_acl.c index d60219f..e2fdebd 100644 --- a/lib/librte_acl/rte_acl.c +++ b/lib/librte_acl/rte_acl.c @@ -55,11 +55,32 @@ rte_acl_classify_avx2(__rte_unused const struct rte_acl_ctx *ctx, return -ENOTSUP; } +int __attribute__ ((weak)) +rte_acl_classify_sse(__rte_unused const struct rte_acl_ctx *ctx, + __rte_unused const uint8_t **data, + __rte_unused uint32_t *results, + __rte_unused uint32_t num, + __rte_unused uint32_t categories) +{ + return -ENOTSUP; +} + +int __attribute__ ((weak)) +rte_acl_classify_neon(__rte_unused const struct rte_acl_ctx *ctx, + __rte_unused const uint8_t **data, + __rte_unused uint32_t *results, + __rte_unused uint32_t num, + __rte_unused uint32_t categories) +{ + return -ENOTSUP; +} + static const rte_acl_classify_t classify_fns[] = { [RTE_ACL_CLASSIFY_DEFAULT] = rte_acl_classify_scalar, [RTE_ACL_CLASSIFY_SCALAR] = rte_acl_classify_scalar, [RTE_ACL_CLASSIFY_SSE] = rte_acl_classify_sse, [RTE_ACL_CLASSIFY_AVX2] = rte_acl_classify_avx2, + [RTE_ACL_CLASSIFY_NEON] = rte_acl_classify_neon, }; /* by default, use always available scalar code path. */ @@ -93,6 +114,9 @@ rte_acl_init(void) { enum rte_acl_classify_alg alg = RTE_ACL_CLASSIFY_DEFAULT; +#ifdef RTE_ARCH_ARM64 + alg = RTE_ACL_CLASSIFY_NEON; +#else #ifdef CC_AVX2_SUPPORT if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2)) alg = RTE_ACL_CLASSIFY_AVX2; @@ -102,6 +126,7 @@ rte_acl_init(void) #endif alg = RTE_ACL_CLASSIFY_SSE; +#endif rte_acl_set_default_classify(alg); } diff --git a/lib/librte_acl/rte_acl.h b/lib/librte_acl/rte_acl.h index 98ef2fc..0979a09 100644 --- a/lib/librte_acl/rte_acl.h +++ b/lib/librte_acl/rte_acl.h @@ -270,6 +270,7 @@ enum rte_acl_classify_alg { RTE_ACL_CLASSIFY_SCALAR = 1, /**< generic implementation. */ RTE_ACL_CLASSIFY_SSE = 2, /**< requires SSE4.1 support. */ RTE_ACL_CLASSIFY_AVX2 = 3, /**< requires AVX2 support. */ + RTE_ACL_CLASSIFY_NEON = 4, /**< requires NEON support. */ RTE_ACL_CLASSIFY_NUM /* should always be the last one. */ };