From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org [79.124.17.100]) by master.gitmailbox.com (Postfix) with ESMTP id DF79543F73 for ; Tue, 23 Aug 2022 06:54:21 +0000 (UTC) Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 3724068BA26; Tue, 23 Aug 2022 09:54:18 +0300 (EEST) Received: from mga14.intel.com (mga14.intel.com [192.55.52.115]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 46A5968B357 for ; Tue, 23 Aug 2022 09:54:10 +0300 (EEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1661237656; x=1692773656; h=from:to:cc:subject:date:message-id:references: in-reply-to:content-id:content-transfer-encoding: mime-version; bh=tM+7unDw4oj1Br1kCLufTrernRHfHuwlzxzYZaqqb74=; b=hnjw86xGMzY52L3wnpPh+TYo4cSkZ4NOVdwPmbi3IADJ02jbNz8/Ke0v b8KZXkcOdq52WQ9JsRCROuWix22zOhqEhdc4FXNQ/UUmDGHaQG+da5Dt7 6P9sQk4CfJa8PcXV2aTBdAOB3Yi14EbPcV2H132gHum0iO5N3w6e3x+gE nK4LDn0goF0Q+ey8683lOLBjbgoy4IJv4jFTJ0CHzUz5O6aitOjexniIb eOWNgMNT+ddtAp7X7Ni/VqMMf9pHCc41inaZo4WhzQmS/iRDkQjlj4KgS 4ipAPjCsQF7vM6lypwYkd+XSQutWjAyOM+THEshBcq7DNkyc99ptZsfUU w==; X-IronPort-AV: E=McAfee;i="6500,9779,10447"; a="293607440" X-IronPort-AV: E=Sophos;i="5.93,256,1654585200"; d="scan'208";a="293607440" Received: from fmsmga008.fm.intel.com ([10.253.24.58]) by fmsmga103.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 22 Aug 2022 23:54:08 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.93,256,1654585200"; d="scan'208";a="669893098" Received: from orsmsx603.amr.corp.intel.com ([10.22.229.16]) by fmsmga008.fm.intel.com with ESMTP; 22 Aug 2022 23:54:08 -0700 Received: from orsmsx611.amr.corp.intel.com (10.22.229.24) by ORSMSX603.amr.corp.intel.com (10.22.229.16) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2375.31; Mon, 22 Aug 2022 23:54:07 -0700 Received: from ORSEDG601.ED.cps.intel.com (10.7.248.6) by orsmsx611.amr.corp.intel.com (10.22.229.24) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2375.31 via Frontend Transport; Mon, 22 Aug 2022 23:54:07 -0700 Received: from NAM11-DM6-obe.outbound.protection.outlook.com (104.47.57.168) by edgegateway.intel.com (134.134.137.102) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.1.2375.31; Mon, 22 Aug 2022 23:54:07 -0700 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=kH2bogKditcFB5QNXztL3QJUAVSF4RvwXk9bbBzK4TLM6iXfb05DGkaMTP/pfIHJtqNd+iAU7G/UiIvPOVSFLgBzo6cdtXXn8JeggW5hS8ZM34c9GCW9zLAlLnS2t85+3+JBjdiX1Vzo8pBYWlBIk/5ddn5A4F0AV5uJAaOQM0x6s0yaeBNmXcyB//e2PrzflIzZy0/DUqBmqIbe/t71CQvDxMexbC25rtgNaWx8AXUX68/7ejpdizCageDty0JEc7F62l66hdiK3UzMkv4Ld2Ii07i+gX4PkHIIM52dcHBclv3QgCmfysYeERS8ofW9o0NWJgiXyHgdJbp0PUhRNg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=tM+7unDw4oj1Br1kCLufTrernRHfHuwlzxzYZaqqb74=; b=RL2AZP0WKRtMFJ/BwOoiMIOWGGC6vvaTC8UNmPhASKQews5+hUsJWMdYKah9NjquJIGorI4sXqLhbDQcgwrfjVAhLGL/xtfM5O9C9T6GuQLozWFG3uubjTnUjPviO2y2lX4jQzn7Yn/DgHP3SFf30CCFhOqyA90aVYsK7mYu+jgVYKa7sxzN9W25oMABqonEfmjrgYUonFKMfM3svv4+KtA30RGwcdgPAeTWDoM5z4AC432Kue7fkWgZK0VP4SGGYpyCkOX/T08k1B1HPXATS4mqfYCI5YEEhqYSKuZirWIxJaZelZ6SBhPmix5r9C5CD/FBNdBYZ7BcrSr3h3T7SA== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=intel.com; dmarc=pass action=none header.from=intel.com; dkim=pass header.d=intel.com; arc=none Received: from BN9PR11MB5515.namprd11.prod.outlook.com (2603:10b6:408:104::8) by SN6PR11MB2717.namprd11.prod.outlook.com (2603:10b6:805:60::32) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.5546.21; Tue, 23 Aug 2022 06:54:05 +0000 Received: from BN9PR11MB5515.namprd11.prod.outlook.com ([fe80::8444:e015:794f:441b]) by BN9PR11MB5515.namprd11.prod.outlook.com ([fe80::8444:e015:794f:441b%2]) with mapi id 15.20.5546.023; Tue, 23 Aug 2022 06:54:05 +0000 From: "Xiang, Haihao" To: "ffmpeg-devel@ffmpeg.org" Thread-Topic: [FFmpeg-devel] [PATCH v2] libavfilter/x86/vf_convolution: add sobel filter optimization and unit test with intel AVX512 VNNI Thread-Index: AQHYsGwrrX6TB663qk69TueYmrG0i628GWsA Date: Tue, 23 Aug 2022 06:54:05 +0000 Message-ID: <2acd4fc77aaf2a43d9a444f082bf5a23aec8f333.camel@intel.com> References: <20220815053844.30298-1-bin.wang@intel.com> In-Reply-To: <20220815053844.30298-1-bin.wang@intel.com> Accept-Language: en-AS, zh-CN, en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-mailer: Evolution 3.28.5-0ubuntu0.18.04.2 authentication-results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=intel.com; x-ms-publictraffictype: Email x-ms-office365-filtering-correlation-id: f871d867-aae5-41e8-57fa-08da84d44501 x-ms-traffictypediagnostic: SN6PR11MB2717:EE_ x-ms-exchange-senderadcheck: 1 x-ms-exchange-antispam-relay: 0 x-microsoft-antispam: BCL:0; x-microsoft-antispam-message-info: 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 x-forefront-antispam-report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:BN9PR11MB5515.namprd11.prod.outlook.com; PTR:; CAT:NONE; SFS:(13230016)(346002)(136003)(39860400002)(376002)(396003)(366004)(2616005)(5660300002)(64756008)(966005)(186003)(76116006)(6486002)(4326008)(8936002)(8676002)(91956017)(66446008)(66946007)(66556008)(66476007)(30864003)(6506007)(107886003)(26005)(478600001)(36756003)(6512007)(71200400001)(41300700001)(86362001)(82960400001)(122000001)(316002)(6916009)(38070700005)(2906002)(45080400002)(38100700002)(99106002)(2004002); DIR:OUT; SFP:1102; x-ms-exchange-antispam-messagedata-chunkcount: 1 x-ms-exchange-antispam-messagedata-0: =?utf-8?B?b3hRaEpDdUx4RS9Fellrcm5LNFBadkFpN1h5STFLV0tYNEhDRERHd3VFazlX?= =?utf-8?B?K1Jsa05iUnNoNTk1a1FsTlFFR0EwYUVJRHBkWldIM2dnLzkwdHp3RENIclVC?= =?utf-8?B?V2U3UmhGTkJPMEt1Rms5U2N3bWhXem9acnZ1MFIrOWtIbWlra2NRanFIQUFL?= =?utf-8?B?VkxDVnZPODdIa1d4bzVtWXNMMjlwZDlLQS8wR0d5Y0xEak9SZ1ZSNGFRTjVD?= =?utf-8?B?di9CUDN1d1RGMHZ2NE9GS2x1M1pJRFkvb1dpb2RKWi9WZ2IwTHk1dS81SjFl?= =?utf-8?B?VFZYOUN6MnpqZ1oxTkZiemRYOGkra1prbE80UTh5TzkzcHFTMy8xM01rSkt1?= =?utf-8?B?Q3hEWExMY3NqTzRySVJoZFN2bUFvdjcyOGdWeVgvYjNLV21EaWcva29leGI4?= =?utf-8?B?M3pFSFFDOFZTd3o4Z3dBczRlZXRMd1hETnhVVFJtRENuQnZGUWNLb1k5aVlN?= =?utf-8?B?ZkNSNVZTWDFBV2ltcEl2ZWM3dmNqTzl5RG9kcE1mOWxWZXFNN3JrNHNkQURH?= =?utf-8?B?cWVqSnJwU1NUVXZKVi9nZGx3VURQOG52WTNpWUZCTEJKRzl2UUZlNHZOR0tL?= =?utf-8?B?MjRYRjlUa3J6K25kdW5MM3Z3M1FZMmp4MjlUMGdBelRHbXBuaDBaeldjdENj?= =?utf-8?B?VWYrSkVjTjVJbjcwTkRCaHZTSmJiWFEyeGlQL2l2eXZvalZzZm9VWktiRGtU?= =?utf-8?B?VHRaY1RGV0ZuY3VlQ1JOcXA2ZG4yY01NUC9KMVdoczRtOXcwdXovRFZ3K3g4?= =?utf-8?B?bitEL1ZpUTlVL1pRenA4V09FYlNTTG5WWWg0R2lVSG5wSWVhbyt6ejU4Mnl0?= =?utf-8?B?QnBGSFFhTUVBTFY2eElBcnJPYzBuZkdQcldnczZHZUdobXg2WENsWmtsd2VR?= =?utf-8?B?NEZramhQRWo5cWJYZlpsTEYzc0tyazRxZVNxblpEbHFXOHBvbEpaTTRQYkM2?= =?utf-8?B?L1lIN2dSVzd6ZjRMTFZVVkp2N3p1NThHVUtmK1ZocXpJZnlSSnRIdXNXUDdT?= =?utf-8?B?Y0IxYUUzZXZ5OGdCa3RKcTVYQWtkOVo4RzkxNEVtNEFWMFhSdVcwK0FJOG9X?= =?utf-8?B?eG1VWXh3TThOTkNjZXN1bWladHdIME11N2ZUdDJheEVUeEdQZnFrcWZ5UTRG?= =?utf-8?B?c3EvWU1YU2FDMkN2Wmw3VVJUR3VyZEc5OGxRaEdXQnJ4ajRKZTJ4ZjBFR2VD?= =?utf-8?B?QlpDMHdxbW1vclgxKzNXQkZZQmlsdnhYNjRLSkVjQTQwSm1MQkQzaHFIaGZJ?= =?utf-8?B?L0JGa21sNW02K3dLQ1dXVFlTY3JnQk5MdzNVaUdJRnZUT1dDWnFmSy8zNUVi?= =?utf-8?B?amp3b3BTWWhtRk1QSTlieHBMV0JSRStYcTRCRGhqSUFlcVBZRVVoQnRMbCtj?= =?utf-8?B?OUs3a0hZWk8rMVlXYlFwMjAzTWJXbmlkelh5RTVKRFpmR04zVUlaVFFTanpU?= =?utf-8?B?VEptaEN3azU4ZDZOOGhXMEtObnM1ZmMrMWpyYU11eEplbE5BWUVoYXRDTkxa?= =?utf-8?B?RDAvSnd4ZWRqUnVVTmtHSzVIb0Y2ZktWdy9aMkRza1kzblg5cmpyVnJsZTdC?= =?utf-8?B?S2Ftb1EwYzBRaU50dUhqQ2ZvUm9CU29hQmIxUk5wVXpoVG5BZ0I0MlpFSGh4?= =?utf-8?B?eEhSbzdZdlIwNzBlM1dyK1BQZDhpV3dSM0pqc1Y3ZGpRZHp4a3Jiejg5bERY?= =?utf-8?B?STFNUHdEV3BsWUdOdFl3elBWaGpxUUgxK2orZWRUVm9SSi93N0gxY3hWTyt0?= =?utf-8?B?TWE1NnlkaWxicTZ3TlR4R1lVUkNpYVkvTUI2Q1NiN0hTekJnRkJIckhFN2VN?= =?utf-8?B?Y1JQVmp1UWFzMlBvN2ExZGw0Si9qV1RWOGVNK0xicGNIdk1sWVN4bER0elhQ?= =?utf-8?B?UG8xdmluazFUZmlraTVXc0kzVmR5anhRQkJoR29nUXRtME5MQWtUNkF3bkRE?= =?utf-8?B?TVBFY2lrbzhNTHg4K3p6SXFRLzR1NElRNWYrbWxMZGc4Y2drK1gveC83Q0k0?= =?utf-8?B?ZUgrUjNhUmhnQlBnRksyYVBTNDBWLzVVMDdPaUV0aC9MSHhRZmtab1RoRkNu?= =?utf-8?B?OWhqWG5xRjNrYlFzcldianJEMC9RL1N3NGFCRnBZL3AySExYNkI0OTBhd29H?= =?utf-8?B?YTNXSm56QVI4dGNVbG8xYzBCNFZiZWZEc3BqSnJ6dzc4NS9yakVXaFFsODg3?= =?utf-8?B?N2c9PQ==?= Content-ID: <3CF4F2FD133E51448E3F4E072B173B93@namprd11.prod.outlook.com> MIME-Version: 1.0 X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-AuthSource: BN9PR11MB5515.namprd11.prod.outlook.com X-MS-Exchange-CrossTenant-Network-Message-Id: f871d867-aae5-41e8-57fa-08da84d44501 X-MS-Exchange-CrossTenant-originalarrivaltime: 23 Aug 2022 06:54:05.1371 (UTC) X-MS-Exchange-CrossTenant-fromentityheader: Hosted X-MS-Exchange-CrossTenant-id: 46c98d88-e344-4ed4-8496-4ed7712e255d X-MS-Exchange-CrossTenant-mailboxtype: HOSTED X-MS-Exchange-CrossTenant-userprincipalname: uEk+/cy61ON8Ht9I6Wv1bcWPuYCf3JqhHq/5gxV//+dnqXRl4FmZet/GiTweVwehxmXeXsXUHBjhJ/mukxkqeA== X-MS-Exchange-Transport-CrossTenantHeadersStamped: SN6PR11MB2717 X-OriginatorOrg: intel.com Subject: Re: [FFmpeg-devel] [PATCH v2] libavfilter/x86/vf_convolution: add sobel filter optimization and unit test with intel AVX512 VNNI X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: FFmpeg development discussions and patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Reply-To: FFmpeg development discussions and patches Cc: "Wang, Bin" Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" Archived-At: List-Archive: List-Post: On Mon, 2022-08-15 at 13:38 +0800, bin.wang-at-intel.com@ffmpeg.org wrote: > From: bwang30 > > This commit enabled assembly code with intel AVX512 VNNI and added unit test > for sobel filter > > sobel_c: 4537 > sobel_avx512icl 2470 > > Signed-off-by: bwang30 > --- > libavfilter/convolution.h | 2 + > libavfilter/vf_convolution.c | 23 ++++ > libavfilter/x86/vf_convolution.asm | 162 ++++++++++++++++++++++++++ > libavfilter/x86/vf_convolution_init.c | 18 +++ > tests/checkasm/Makefile | 1 + > tests/checkasm/checkasm.c | 3 + > tests/checkasm/checkasm.h | 1 + > tests/checkasm/vf_convolution.c | 103 ++++++++++++++++ > 8 files changed, 313 insertions(+) > create mode 100644 tests/checkasm/vf_convolution.c > > diff --git a/libavfilter/convolution.h b/libavfilter/convolution.h > index 88aabe9a20..4520ad13e0 100644 > --- a/libavfilter/convolution.h > +++ b/libavfilter/convolution.h > @@ -61,4 +61,6 @@ typedef struct ConvolutionContext { > } ConvolutionContext; > > void ff_convolution_init_x86(ConvolutionContext *s); > +void ff_sobel_init_x86(ConvolutionContext *s); > +void ff_convolution_init(ConvolutionContext *s, const char *filter_name); > #endif > diff --git a/libavfilter/vf_convolution.c b/libavfilter/vf_convolution.c > index 9a9c099e6d..f27385a7fb 100644 > --- a/libavfilter/vf_convolution.c > +++ b/libavfilter/vf_convolution.c > @@ -874,6 +874,9 @@ static int param_init(AVFilterContext *ctx) > if (s->depth > 8) > for (p = 0; p < s->nb_planes; p++) > s->filter[p] = filter16_sobel; > +#if CONFIG_CONVOLUTION_FILTER && ARCH_X86_64 > + ff_sobel_init_x86(s); > +#endif > } else if (!strcmp(ctx->filter->name, "kirsch")) { > if (s->depth > 8) > for (p = 0; p < s->nb_planes; p++) > @@ -887,6 +890,26 @@ static int param_init(AVFilterContext *ctx) > return 0; > } > > +void ff_convolution_init(ConvolutionContext *s, const char *filter_name) > +{ > + if (!strcmp(filter_name, "sobel")) { > + for (int i = 0; i < 4; i++) { > + s->filter[i] = filter_sobel; > + s->copy[i] = !((1 << i) & s->planes); > + s->size[i] = 3; > + s->setup[i] = setup_3x3; > + s->rdiv[i] = s->scale; > + s->bias[i] = s->delta; > + } > + if (s->depth > 8) > + for (int i = 0; i < 4; i++) > + s->filter[i] = filter16_sobel; > +#if CONFIG_CONVOLUTION_FILTER && ARCH_X86_64 > + ff_sobel_init_x86(s); > +#endif > + } > +} > + > static int config_input(AVFilterLink *inlink) > { > AVFilterContext *ctx = inlink->dst; > diff --git a/libavfilter/x86/vf_convolution.asm > b/libavfilter/x86/vf_convolution.asm > index 754d4d1064..59c807b218 100644 > --- a/libavfilter/x86/vf_convolution.asm > +++ b/libavfilter/x86/vf_convolution.asm > @@ -22,6 +22,10 @@ > > SECTION_RODATA > half: dd 0.5 > +data_p1: dd 1 > +data_n1: dd -1 > +data_p2: dd 2 > +data_n2: dd -2 > > SECTION .text > > @@ -154,3 +158,161 @@ cglobal filter_3x3, 4, 15, 7, dst, width, rdiv, bias, > matrix, ptr, c0, c1, c2, c > INIT_XMM sse4 > FILTER_3X3 > %endif > + > + > +%macro SOBEL_MUL_16 3 > + movd xmm2, [%2] > + VPBROADCASTD m2, xmm2 > + movdqu xmm3, [c%1q + xq] > + vpmovzxbd m3, xmm3 > + vpdpbusd m%3, m3, m2 > +%endmacro > + > +%macro SOBEL_ADD_16 2 > + movdqu xmm3, [c%1q + xq] > + vpmovzxbd m3, xmm3 > + vpaddd m%2, m3 > +%endmacro > + > + > +%macro SOBEL_MUL 2 > + movzx ptrd, byte [c%1q + xq] > + imul ptrd, [%2] > + add rd, ptrd > +%endmacro > + > +%macro SOBEL_ADD 1 > + movzx ptrd, byte [c%1q + xq] > + add rd, ptrd > +%endmacro > + > +; void filter_sobel_avx512(uint8_t *dst, int width, > +; float scale, float delta, const int *const matrix, > +; const uint8_t *c[], int peak, int radius, > +; int dstride, int stride) > +%macro FILTER_SOBEL 0 > +%if UNIX64 > +cglobal filter_sobel, 4, 15, 7, dst, width, matrix, ptr, c0, c1, c2, c3, c4, > c5, c6, c7, c8, r, x > +%else > +cglobal filter_sobel, 4, 15, 7, dst, width, rdiv, bias, matrix, ptr, c0, c1, > c2, c3, c4, c5, c6, c7, c8, r, x > +%endif > +%if WIN64 > + SWAP xmm0, xmm2 > + SWAP xmm1, xmm3 > + mov r2q, matrixmp > + mov r3q, ptrmp > + DEFINE_ARGS dst, width, matrix, ptr, c0, c1, c2, c3, c4, c5, c6, c7, c8, > r, x > +%endif > + movsxdifnidn widthq, widthd > + VBROADCASTSS m0, xmm0 > + VBROADCASTSS m1, xmm1 > + pxor m6, m6 > + mov c0q, [ptrq + 0*gprsize] > + mov c1q, [ptrq + 1*gprsize] > + mov c2q, [ptrq + 2*gprsize] > + mov c3q, [ptrq + 3*gprsize] > + mov c4q, [ptrq + 4*gprsize] > + mov c5q, [ptrq + 5*gprsize] > + mov c6q, [ptrq + 6*gprsize] > + mov c7q, [ptrq + 7*gprsize] > + mov c8q, [ptrq + 8*gprsize] > + > + xor xq, xq > + cmp widthq, mmsize/4 > + jl .loop2 > + > + mov rq, widthq > + and rq, mmsize/4-1 > + sub widthq, rq > + > +.loop1: > + pxor m4, m4 > + pxor m5, m5 > + > + ;Gx > + SOBEL_MUL_16 0, data_n1, 4 > + SOBEL_MUL_16 1, data_n2, 4 > + SOBEL_MUL_16 2, data_n1, 4 > + SOBEL_ADD_16 6, 4 > + SOBEL_MUL_16 7, data_p2, 4 > + SOBEL_ADD_16 8, 4 > + > + cvtdq2ps m4, m4 > + mulps m4, m4 > + > + ;Gy > + SOBEL_MUL_16 0, data_n1, 5 > + SOBEL_ADD_16 2, 5 > + SOBEL_MUL_16 3, data_n2, 5 > + SOBEL_MUL_16 5, data_p2, 5 > + SOBEL_MUL_16 6, data_n1, 5 > + SOBEL_ADD_16 8, 5 > + > + cvtdq2ps m5, m5 > + VFMADD231PS m4, m5, m5 > + > + sqrtps m4, m4 > + mulps m4, m0 ; sum *= scale > + addps m4, m1 ; sum += delta > + cvttps2dq m4, m4 > + vpmovusdb xmm4, m4 > + movdqu [dstq + xq], xmm4 > + > + add xq, mmsize/4 > + cmp xq, widthq > + jl .loop1 > + > + add widthq, rq > + cmp xq, widthq > + jge .end > + > +.loop2: > + xor rd, rd > + pxor m4, m4 > + > + ;Gx > + SOBEL_MUL 0, data_n1 > + SOBEL_MUL 1, data_n2 > + SOBEL_MUL 2, data_n1 > + SOBEL_ADD 6 > + SOBEL_MUL 7, data_p2 > + SOBEL_ADD 8 > + > + cvtsi2ss xmm4, rd > + mulss xmm4, xmm4 > + > + xor rd, rd > + ;Gy > + SOBEL_MUL 0, data_n1 > + SOBEL_ADD 2 > + SOBEL_MUL 3, data_n2 > + SOBEL_MUL 5, data_p2 > + SOBEL_MUL 6, data_n1 > + SOBEL_ADD 8 > + > + cvtsi2ss xmm5, rd > + mulss xmm5, xmm5 ; b1 * b1 > + addss xmm4, xmm5 > + > + sqrtps xmm4, xmm4 > + mulss xmm4, xmm0 ; sum *= rdiv > + addss xmm4, xmm1 ; sum += bias > + cvttps2dq xmm4, xmm4 ; trunc to integer > + packssdw xmm4, xmm4 > + packuswb xmm4, xmm4 > + movd rd, xmm4 > + mov [dstq + xq], rb > + > + add xq, 1 > + cmp xq, widthq > + jl .loop2 > +.end: > + RET > +%endmacro > + > +%if ARCH_X86_64 > +%if HAVE_AVX512ICL_EXTERNAL > +INIT_ZMM avx512icl > +FILTER_SOBEL > +%endif > +%endif > diff --git a/libavfilter/x86/vf_convolution_init.c > b/libavfilter/x86/vf_convolution_init.c > index b78a47d02b..52a3d28991 100644 > --- a/libavfilter/x86/vf_convolution_init.c > +++ b/libavfilter/x86/vf_convolution_init.c > @@ -29,6 +29,11 @@ void ff_filter_3x3_sse4(uint8_t *dst, int width, > const uint8_t *c[], int peak, int radius, > int dstride, int stride, int size); > > +void ff_filter_sobel_avx512icl(uint8_t *dst, int width, > + float scale, float delta, const int *const matrix, > + const uint8_t *c[], int peak, int radius, > + int dstride, int stride, int size); > + > av_cold void ff_convolution_init_x86(ConvolutionContext *s) > { > #if ARCH_X86_64 > @@ -44,3 +49,16 @@ av_cold void ff_convolution_init_x86(ConvolutionContext *s) > } > #endif > } > + > +av_cold void ff_sobel_init_x86(ConvolutionContext *s) > +{ > +#if ARCH_X86_64 > + int cpu_flags = av_get_cpu_flags(); > + for (int i = 0; i < s->nb_planes; i++) { > + if (s->depth == 8) { > + if (EXTERNAL_AVX512ICL(cpu_flags)) > + s->filter[i] = ff_filter_sobel_avx512icl; > + } > + } > +#endif > +} > diff --git a/tests/checkasm/Makefile b/tests/checkasm/Makefile > index 1ac170491b..4e91547fde 100644 > --- a/tests/checkasm/Makefile > +++ b/tests/checkasm/Makefile > @@ -44,6 +44,7 @@ AVFILTEROBJS-$(CONFIG_GBLUR_FILTER) += vf_gblur.o > AVFILTEROBJS-$(CONFIG_HFLIP_FILTER) += vf_hflip.o > AVFILTEROBJS-$(CONFIG_THRESHOLD_FILTER) += vf_threshold.o > AVFILTEROBJS-$(CONFIG_NLMEANS_FILTER) += vf_nlmeans.o > +AVFILTEROBJS-$(CONFIG_CONVOLUTION_FILTER) += vf_convolution.o > > CHECKASMOBJS-$(CONFIG_AVFILTER) += $(AVFILTEROBJS-yes) > > diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c > index e56fd3850e..ae5e9d1143 100644 > --- a/tests/checkasm/checkasm.c > +++ b/tests/checkasm/checkasm.c > @@ -191,6 +191,9 @@ static const struct { > #if CONFIG_THRESHOLD_FILTER > { "vf_threshold", checkasm_check_vf_threshold }, > #endif > + #if CONFIG_CONVOLUTION_FILTER > + { "vf_convolution", checkasm_check_vf_convolution }, > + #endif > #endif > #if CONFIG_SWSCALE > { "sw_gbrp", checkasm_check_sw_gbrp }, > diff --git a/tests/checkasm/checkasm.h b/tests/checkasm/checkasm.h > index d7645d3730..71c722cf77 100644 > --- a/tests/checkasm/checkasm.h > +++ b/tests/checkasm/checkasm.h > @@ -85,6 +85,7 @@ void checkasm_check_vf_eq(void); > void checkasm_check_vf_gblur(void); > void checkasm_check_vf_hflip(void); > void checkasm_check_vf_threshold(void); > +void checkasm_check_vf_convolution(void); > void checkasm_check_vp8dsp(void); > void checkasm_check_vp9dsp(void); > void checkasm_check_videodsp(void); > diff --git a/tests/checkasm/vf_convolution.c b/tests/checkasm/vf_convolution.c > new file mode 100644 > index 0000000000..a10da8b45e > --- /dev/null > +++ b/tests/checkasm/vf_convolution.c > @@ -0,0 +1,103 @@ > +/* > + * This file is part of FFmpeg. > + * > + * FFmpeg is free software; you can redistribute it and/or modify > + * it under the terms of the GNU General Public License as published by > + * the Free Software Foundation; either version 2 of the License, or > + * (at your option) any later version. > + * > + * FFmpeg is distributed in the hope that it will be useful, > + * but WITHOUT ANY WARRANTY; without even the implied warranty of > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the > + * GNU General Public License for more details. > + * > + * You should have received a copy of the GNU General Public License along > + * with FFmpeg; if not, write to the Free Software Foundation, Inc., > + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. > + */ > + > +#include > +#include "checkasm.h" > +#include "libavfilter/avfilter.h" > +#include "libavfilter/convolution.h" > +#include "libavutil/intreadwrite.h" > +#include "libavutil/mem_internal.h" > + > +#define WIDTH 512 > +#define HEIGHT 512 > +#define SRC_STRIDE 512 > +#define PIXELS (WIDTH * HEIGHT) > + > +#define randomize_buffers(buf, size) \ > + do { \ > + int j; \ > + uint8_t *tmp_buf = (uint8_t *)buf;\ > + for (j = 0; j< size; j++) \ > + tmp_buf[j] = rnd() & 0xFF; \ > + } while (0) > + > +static void check_sobel(const char * report_name) > +{ > + LOCAL_ALIGNED_32(uint8_t, src, [PIXELS]); > + LOCAL_ALIGNED_32(uint8_t, dst_ref, [PIXELS]); > + LOCAL_ALIGNED_32(uint8_t, dst_new, [PIXELS]); > + const int height = WIDTH; > + const int width = HEIGHT; > + const int stride = SRC_STRIDE; > + const int dstride = SRC_STRIDE; > + int mode = 0; > + const uint8_t *c[49]; > + const int radius = 1; > + const int bpc = 1; > + const int step = mode == MATRIX_COLUMN ? 16 : 1; > + const int slice_start = 0; > + const int slice_end = height; > + int y; > + const int sizew = mode == MATRIX_COLUMN ? height : width; > + float scale = 2; > + float delta = 10; > + > + ConvolutionContext s; > + > + s.scale = scale; > + s.delta = delta; > + s.depth = 8; > + s.nb_planes = 3; > + ff_convolution_init(&s, "sobel"); > + > + declare_func(void, uint8_t *dst, int width, float scale, float delta, > const int *const matrix, > + const uint8_t *c[], int peak, int radius, int dstride, int > stride, int size); > + > + memset(dst_ref, 0, PIXELS); > + memset(dst_new, 0, PIXELS); > + randomize_buffers(src, PIXELS); > + > + if (check_func(s.filter[0], "%s", report_name)) { > + for (y = slice_start; y < slice_end; y += step) { > + const int xoff = mode == MATRIX_COLUMN ? (y - slice_start) * bpc > : radius * bpc; > + const int yoff = mode == MATRIX_COLUMN ? radius * dstride : 0; > + > + s.setup[0](radius, c, src, stride, radius, width, y, height, > bpc); > + call_ref(dst_ref + yoff + xoff, sizew - 2 * radius, > + scale, delta, NULL, c, 0, radius, > + dstride, stride, slice_end - step); > + call_new(dst_new + yoff + xoff, sizew - 2 * radius, > + scale, delta, NULL, c, 0, radius, > + dstride, stride, slice_end - step); > + if (memcmp(dst_ref + yoff + xoff, dst_new + yoff + xoff, > slice_end - step)) > + fail(); > + bench_new(dst_new + yoff + xoff, sizew - 2 * radius, > + scale, delta, NULL, c, 0, radius, > + dstride, stride, slice_end - step); > + if (mode != MATRIX_COLUMN) > + dst_ref += dstride; > + } > + } > + > +} > + > +void checkasm_check_vf_convolution(void) > +{ > + check_sobel("sobel"); > + report("convolution:sobel"); > +} Failed to apply your patch in fate testing, see https://patchwork.ffmpeg.org/project/ffmpeg/patch/BN6PR11MB1746B88D2FEC63B85650B4379B6D9@BN6PR11MB1746.namprd11.prod.outlook.com/ Thanks Haihao _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".