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 B2FD442FA8 for ; Fri, 14 Oct 2022 23:42:28 +0000 (UTC) Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 7715268BD62; Sat, 15 Oct 2022 02:42:25 +0300 (EEST) Received: from EUR04-HE1-obe.outbound.protection.outlook.com (mail-oln040092073031.outbound.protection.outlook.com [40.92.73.31]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id BEBB468BBB8 for ; Sat, 15 Oct 2022 02:42:19 +0300 (EEST) ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=M9vuYD1qmXeFN9HJwnxbFDFaN80ogaDZNolqQ1PVAoi4vlr0Awrl/mwgoj9gzzoEud4qCkJbK/l2brobOmuakX2TPZwibaXPZWgvTOMj9R4YezYM6eX8mLng2T22nr8AKblXWntcAt4+OujYdVyQXpBRzPpIzrMZOqrEmdGzSkAkwR8deRt5rTGuPD6LO6rtXmpyvrHLH6VBSU3spolJBYj/SPk06V8y7JRpupGY6CxAzCU/ZMDhRypzyfsud2SkUayo9UnQ3knBrt8YVVHUDXS31sNutYNuhL23ypPkI8yIXyGj+BBo+iocpBpczQa3OTgU1QUeMjkHmYLF14jrGQ== 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=8IgK4dnn1KPhWjWv/kO5Vbi64X9DbnZ0/4j4Ji5rw3k=; b=dqM9/rRlma/dYYq5uauYd77CBUhhZte7R1ori/LOeUZHvflafEKGf4D1Ivf1oMMOwjmCqNglFiiP07woxbdtjWLGHH+2GC6gMWK/Oy8z9Xpg/f9ZdV6VZVJJoKsHfpg4F2ngSTDjKDulTTDtOp0v+H9lrFqi7r1ylycqqBag89ml99weOa98F9QSmZWd4moAUGV02IZMz9T5Hgu3D8rlzhgliqbM3lXSkwjqLuLaLAZjTlm9pd6HXG/MRVF1Ga5e0KB0ZRTgU33lrs+jj5kKbPJSRe6FAlGoRh6r7CWpcTrq/hk4zoPDLt1DGPzvudeaFA+izreoLOpdZbiiyjJgEw== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=none; dmarc=none; dkim=none; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=outlook.com; s=selector1; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=8IgK4dnn1KPhWjWv/kO5Vbi64X9DbnZ0/4j4Ji5rw3k=; b=EOWtelFkkk59N2CL+4M8RH28b1chrw2QwI8S243SjZlpGVKOF+9C+wTf3uIUtXcJ1PdlUzFm29Nx0LxRwfLWs7RlVyXKEs3IAFcypfpqDNA2cQ2SJg/UxzGEvSkd1DHDSjHbg0XXRxSLgq9WVf3ccnD8JM1nqoln0JdXuuu0sPfmgj7QpNo++Ac1I18sME+JCyB+knh4c0Y6vxMc31D8YE68Cln7f1LEJvp4dBny6Lr+qu8/95BfJnGtmUIGZBMPaVzB07ZNxK0v47zeSIGVw59kjcI8/JfGwcQ3498pnt7A7dwnqRXFjkVXbb1XYf3KBhgtEaeJuDS8Ku4vemFcvw== Received: from AS8P250MB0744.EURP250.PROD.OUTLOOK.COM (2603:10a6:20b:541::14) by AS1P250MB0405.EURP250.PROD.OUTLOOK.COM (2603:10a6:20b:4ad::7) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.5709.15; Fri, 14 Oct 2022 23:42:17 +0000 Received: from AS8P250MB0744.EURP250.PROD.OUTLOOK.COM ([fe80::f9d7:680f:70c4:44fe]) by AS8P250MB0744.EURP250.PROD.OUTLOOK.COM ([fe80::f9d7:680f:70c4:44fe%7]) with mapi id 15.20.5709.015; Fri, 14 Oct 2022 23:42:17 +0000 Message-ID: Date: Sat, 15 Oct 2022 01:42:27 +0200 Content-Language: en-US To: FFmpeg development discussions and patches References: <166572977016.12287.1847463242431058317@lain.khirnov.net> From: Andreas Rheinhardt In-Reply-To: <166572977016.12287.1847463242431058317@lain.khirnov.net> X-TMN: [8YXHIscuivT5jpwCdzpyVHSy60+INlno5P9Mke41ry0=] X-ClientProxiedBy: FR3P281CA0127.DEUP281.PROD.OUTLOOK.COM (2603:10a6:d10:94::13) To AS8P250MB0744.EURP250.PROD.OUTLOOK.COM (2603:10a6:20b:541::14) X-Microsoft-Original-Message-ID: <0ea8a7b9-471c-f6ad-7c27-9651e42cd6ed@outlook.com> MIME-Version: 1.0 X-MS-Exchange-MessageSentRepresentingType: 1 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: AS8P250MB0744:EE_|AS1P250MB0405:EE_ X-MS-Office365-Filtering-Correlation-Id: 5ba70836-fa4f-4e58-8f7b-08daae3dbac0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: =?utf-8?B?STl3bDBXMk5lUlp6cFMyVUpUbElZZG9tamxYenFuVys3QXMrdXc3WS93QTgz?= =?utf-8?B?b2J3WEttZHFTN2ZFbE9xRG81WlhpVGQza3RCckxuRFo5THBKOEsvUkt6NlIx?= =?utf-8?B?NkhHbzFqTEpBMmFkejZTb0t3NVlWK3hPSElFT2x0bzhjNmYzM0IrbE9pdm5V?= =?utf-8?B?elM4clZyMFFjaFRLZ1JFaExydDFpUDBQMXNvcTRIQnUwTlBMSncyeWs5Q3pt?= =?utf-8?B?L1dJSkpGTFpwVmtUNGN5d1dya1dWVDROekxCdU8vUTJLdUcyN014d0x5cUpM?= =?utf-8?B?MTZSSFE4U25pd3pSeFhHME5TQ3VJODJPb2NCRG5Xck9iMjZ5elVNb1FUYmRh?= =?utf-8?B?SHRFNm1FcTdxOURCMXp3N1ZyU3JoQTVpNGZlOGZJWXVkTFdMR2NrdUdOUWtw?= =?utf-8?B?NW9Tajh0cCtPYmg2SFJDMGk3bXVBOXNkUUd6YnNmYTJDUlJLc25LM2o0RERF?= =?utf-8?B?Vmltb1RXZTRna3pGanhNTnY4aEFVSkx2K3JqaUwwc2ExdUJQcTZyY1NpUGtv?= =?utf-8?B?dTFabXZsYlVMcmxWZ0Q4dERyU1lROGtydEp6alFHR05GTTNoci8rUVh1K3Bt?= =?utf-8?B?MHJkWC82M2VNemV5dVF3TFJjYkhaRTZvdDRMZDV0NVBDV05QcmF1TXM5cXRW?= =?utf-8?B?b0ViTW9FSDZqb1N1VEFJYU0wbDUwdHUzYnBXREx3Q0lYWlVZZnF4QWFqTFNY?= =?utf-8?B?ektMck1SREJoTmxDRXJEZ29RZnJLa2Y1b2I0VC9HQ2NzUkxHWlRlczJReVNI?= =?utf-8?B?NEZ1WWRXQ3pNSW5xdTA5Qzc4R09SazA2bFk5SEFiaVRFNi9jV2JQQndEdTBk?= =?utf-8?B?WURvaFFxR1hwdTl5UUJVdkhtYmZhRDlKMEFnMktHckFjRDBPOTVpamlWWnFF?= =?utf-8?B?WGtwWmZHSm9SMXU2SVBhMU1BVEFGaVNPZm84MGJjNDZBRUlXTDN1WHljc2Vp?= =?utf-8?B?RmNXQUN5Tk1ZK2N1TndvVmhrRk96YUNxM3RNbTVKS0p2VHQyL3BuWmhHdk1V?= =?utf-8?B?UGIyb051eHNhQjBtbG9OMUgwd2dhWHdGNmpUTGVmR25RNGtpMkxnKzIvL2ty?= =?utf-8?B?TExmeXFqSkVKcHpxVHRBb1g4YVZOcHlOTjVCTEF0bHV1YXByT2p0SlBJczAv?= =?utf-8?B?Nzh4L3B1KzZhS1c5bXcwek5NZFVlK1ZvczdIbTJBQjJLQU5NczFCSGg2Wnc3?= =?utf-8?B?Y0MydklTZVQ5QTB5Tm1tR3htQkhsVVhBdmtocGFreHUzMStKUmVEWWVldVE1?= =?utf-8?B?ajZBNmk4OE5TZjFSOXMvSjVmT2NVeVY1NUloaEJGald6Nm9jd2lOajZPYTJv?= =?utf-8?B?eE85Wm9BekpacUhLYUwwdml3WHRoVEU3dWZhMGxVWHprbVZDeVl2dzdyMWM3?= =?utf-8?B?bTdtUDVDdlJIN2ljZXNXVjdha21Fdnh1Z0pMZ2YxZjl4c0tnTUZwZWtNMW01?= =?utf-8?B?cUZxNXY3ODJqdkJIQzR5bjd2c1JxbHA2dkhyTVpOUTV2MWNsWGxkTUZrM1Zq?= =?utf-8?B?aGcyN0tRUEg5RGdlamozMnAzRGFDRE5lYk4zTndUZ3pKci9sSTlVNkVVY3hX?= =?utf-8?B?VXB6enE0enFkd3d3Z3ZYTTF2RVJQM0lSSGtIK0h3ZjFuc2djOUN3bFUrOTFz?= =?utf-8?B?QXR1Q2M1SnJhTTlsV1QxU0N5L09JemtvKy9XajBoNHBMSnBUZHYwWUl5bjFU?= =?utf-8?B?blZqTitHSjhNdkRwN3RtNG51QjN3QmRuL244cVRvWWRQQUtEY3ZTb1poN2Nu?= =?utf-8?Q?j/zXVqHum9Bd36YEstC4xcDzmsO0pn5z9rD69Ho?= X-OriginatorOrg: outlook.com X-MS-Exchange-CrossTenant-Network-Message-Id: 5ba70836-fa4f-4e58-8f7b-08daae3dbac0 X-MS-Exchange-CrossTenant-AuthSource: AS8P250MB0744.EURP250.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 14 Oct 2022 23:42:17.7797 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 84df9e7f-e9f6-40af-b435-aaaaaaaaaaaa X-MS-Exchange-CrossTenant-RMS-PersistedConsumerOrg: 00000000-0000-0000-0000-000000000000 X-MS-Exchange-Transport-CrossTenantHeadersStamped: AS1P250MB0405 Subject: Re: [FFmpeg-devel] [PATCH v2] avcodec/startcode: Avoid unaligned accesses 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 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: Anton Khirnov: > Quoting Andreas Rheinhardt (2022-10-12 00:20:23) >> Up until now, ff_startcode_find_candidate_c() simply casts >> an uint8_t* to uint64_t*/uint32_t* to read 64/32 bits at a time >> in case HAVE_FAST_UNALIGNED is true. Yet this ignores the >> alignment requirement of these types as well as effective type >> rules of the C standard. This commit therefore replaces these >> direct accesses with AV_RN64/32; this also improves >> readability. >> >> UBSan reported these unaligned accesses which happened in 233 >> FATE-tests involving H.264 and VC-1 (this has also been reported >> in tickets #8138 and #8485); these tests are fixed by this commit. >> >> The output of GCC with -O3 is unchanged for aarch64, alpha, arm, >> loongarch, ppc and x64. There was only a slight difference for mips. >> >> Signed-off-by: Andreas Rheinhardt >> --- >> This is v2 of https://patchwork.ffmpeg.org/project/ffmpeg/patch/20200122145210.6898-1-andreas.rheinhardt@gmail.com/ >> >> Here is the mips code before this change: >> >> startcode_old_O3.o: file format elf64-tradlittlemips >> >> >> Disassembly of section .text: >> >> 0000000000000000 : >> 0: 18a00029 blez a1,a8 >> 4: 3c08ff7f lui a4,0xff7f >> 8: 3c07ff01 lui a3,0xff01 >> c: 65087f7f daddiu a4,a4,32639 >> 10: 64e70101 daddiu a3,a3,257 >> 14: 00084438 dsll a4,a4,0x10 >> 18: 00073c38 dsll a3,a3,0x10 >> 1c: 65087f7f daddiu a4,a4,32639 >> 20: 64e70101 daddiu a3,a3,257 >> 24: 00084478 dsll a4,a4,0x11 >> 28: 00073df8 dsll a3,a3,0x17 >> 2c: 00803025 move a2,a0 >> 30: 00001025 move v0,zero >> 34: 3508feff ori a4,a4,0xfeff >> 38: 10000005 b 50 >> 3c: 34e78080 ori a3,a3,0x8080 >> 40: 24420008 addiu v0,v0,8 >> 44: 0045182a slt v1,v0,a1 >> 48: 10600015 beqz v1,a0 >> 4c: 00000000 nop >> 50: dcc30000 ld v1,0(a2) >> 54: 0068482d daddu a5,v1,a4 >> 58: 44a30000 dmtc1 v1,$f0 >> 5c: 44a90800 dmtc1 a5,$f1 >> 60: 4be10002 pandn $f0,$f0,$f1 >> 64: 44230000 dmfc1 v1,$f0 >> 68: 00671824 and v1,v1,a3 >> 6c: 1060fff4 beqz v1,40 >> 70: 64c60008 daddiu a2,a2,8 >> 74: 0045182a slt v1,v0,a1 >> 78: 10600009 beqz v1,a0 >> 7c: 0082182d daddu v1,a0,v0 >> 80: 10000005 b 98 >> 84: 90640000 lbu a0,0(v1) >> 88: 24420001 addiu v0,v0,1 >> 8c: 10a20008 beq a1,v0,b0 >> 90: 00000000 nop >> 94: 90640000 lbu a0,0(v1) >> 98: 1480fffb bnez a0,88 >> 9c: 64630001 daddiu v1,v1,1 >> a0: 03e00008 jr ra >> a4: 00000000 nop >> a8: 03e00008 jr ra >> ac: 00001025 move v0,zero >> b0: 03e00008 jr ra >> b4: 00a01025 move v0,a1 >> ... >> >> And here after this change: >> >> startcode_new_O3.o: file format elf64-tradlittlemips >> >> >> Disassembly of section .text: >> >> 0000000000000000 : >> 0: 18a0002b blez a1,b0 >> 4: 3c08ff7f lui a4,0xff7f >> 8: 3c07ff01 lui a3,0xff01 >> c: 65087f7f daddiu a4,a4,32639 >> 10: 64e70101 daddiu a3,a3,257 >> 14: 00084438 dsll a4,a4,0x10 >> 18: 00073c38 dsll a3,a3,0x10 >> 1c: 65087f7f daddiu a4,a4,32639 >> 20: 64e70101 daddiu a3,a3,257 >> 24: 00084478 dsll a4,a4,0x11 >> 28: 00073df8 dsll a3,a3,0x17 >> 2c: 00803025 move a2,a0 >> 30: 00001025 move v0,zero >> 34: 3508feff ori a4,a4,0xfeff >> 38: 10000005 b 50 >> 3c: 34e78080 ori a3,a3,0x8080 >> 40: 24420008 addiu v0,v0,8 >> 44: 0045182a slt v1,v0,a1 >> 48: 10600017 beqz v1,a8 >> 4c: 00000000 nop >> 50: 68c30007 ldl v1,7(a2) >> 54: 6cc30000 ldr v1,0(a2) >> 58: 0068482d daddu a5,v1,a4 >> 5c: 44a30000 dmtc1 v1,$f0 >> 60: 44a90800 dmtc1 a5,$f1 >> 64: 4be10002 pandn $f0,$f0,$f1 >> 68: 44230000 dmfc1 v1,$f0 >> 6c: 00671824 and v1,v1,a3 >> 70: 1060fff3 beqz v1,40 >> 74: 64c60008 daddiu a2,a2,8 >> 78: 0045182a slt v1,v0,a1 >> 7c: 1060000a beqz v1,a8 >> 80: 0082182d daddu v1,a0,v0 >> 84: 10000006 b a0 >> 88: 90640000 lbu a0,0(v1) >> 8c: 00000000 nop >> 90: 24420001 addiu v0,v0,1 >> 94: 10a20008 beq a1,v0,b8 >> 98: 00000000 nop >> 9c: 90640000 lbu a0,0(v1) >> a0: 1480fffb bnez a0,90 >> a4: 64630001 daddiu v1,v1,1 >> a8: 03e00008 jr ra >> ac: 00000000 nop >> b0: 03e00008 jr ra >> b4: 00001025 move v0,zero >> b8: 03e00008 jr ra >> bc: 00a01025 move v0,a1 >> >> As one can see, the difference is that an ld has been replaced >> by a pair of ldl and ldr. I don't know the performance implications >> of this. > > google tells me ld is an aligned load, so this change is then correct > whatever its performance implications are. > I was about to write the same after reading https://www.cs.cmu.edu/afs/cs/academic/class/15740-f97/public/doc/mips-isa.pdf, but then I found out that certain loongson processors allow unaligned accesses (see https://lists.gnu.org/archive/html/qemu-devel/2020-10/msg08563.html) and configure only enables fast_unaligned for loongson2e|loongson2f|loongson3*); if fast_unaligned would lead to crashes on these processors, it would probably have been noticed long ago. So the question of performance implications on such processors is still open. If no one bothers to test this in a reasonable time, I will treat this as sign that no one is interested in the affected arches and apply this. - Andreas _______________________________________________ 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".