|
From: Médéric B. <Med...@in...> - 2011-02-16 09:46:22
|
Dear Siyuan, I think that the JSVM report that your encoding resolution is not a multiple of 16. So it encodes with 320x192 resolution and crop the output video with the 320x180. Concerning the Open SVC Decoder crash, you can send me the video in order to correct the bug. Kind regards, Médéric Le 15/02/2011 03:29, Siyuan Xiang a écrit : > Dear Médéric, > > I use the source video with resolution 320x180, 640x360, 1280x720. > The ratio should be 2. I do not know why JSVM report the resolution > with 320x192, 640x368 and 1280x720. > > I removed MGSVector. When I encode the video with these 3 dependency > layers, layer 1280x720 with dependency id equal to 32 can be decoded. > > Then I config 2 layers(with different QP) for each resolution > resulting in total 6 layers. The layer with DQ id equal to 17 can be > decoded. But when the DQ id is 32 there is an error in I_BL.c > > LeftBorder + (x_mbPartIdx << 2) + (y_mbPartIdx << 2) * LumaStride *is > larger than* LumaStride * LumaStride, > then > tmp_luma[LeftBorder + (x_mbPartIdx << 2) + (y_mbPartIdx << 2) * > LumaStride] leads to memory access violation. > > Is this a bug or wrong configuration? > > > profile & level info: > ===================== > DQ= 0: Baseline @ Level 1.2 > DQ= 1: Scalable Baseline @ Level 1.2 > DQ= 16: Scalable Baseline @ Level 3 > DQ= 17: Scalable Baseline @ Level 3 > DQ= 32: Scalable Baseline @ Level 4 > DQ= 33: Scalable Baseline @ Level 4 > > > Regards, > Siyuan > --- > > > > On Mon, Feb 14, 2011 at 01:42, Médéric Blestel > <Med...@in... > <mailto:Med...@in...>> wrote: > > Dear Siyuan, > > The Open SVC Decoder is still not compatible with MGS Control. > You have to remove the MGSVectorMode option from your > configuration files. > > I have noticed that you are using ratio different from 1.5 or 2 > between two consecutives enhancement layer, which is not > compatible with the SVC Baseline Profile. > To be compatible with the Open SVC Decoder, you can use these size > as example: 320x176, 640x 352, 1280x704. > > Kind regards, > > Médéric > > > > Le 14/02/2011 10:22, Siyuan Xiang a écrit : >> Hi, Guys, >> >> I used JSVM 9.19.10(beta) Encoder to encode a video >> with mixed CGS/MGS configuration. There are 3 resolutions >> 320x180, 640x360, 1280x720 with total number of 8 layers. when I >> use opensvcdecoder to decode, an error is prompted "A layer is >> using a different scan for residual block" in >> SliceHeaderExtension.c:306. >> >> I use MGS setting in configure file as following: >> >> MGSVectorMode 1 # Write coefficients into >> multiple slices >> MGSVector0 8 # Number of coefficients >> that are contained in slice 1 >> MGSVector1 8 # Number of coefficients >> that are contained in slice 2 >> >> Is there anything wrong with the configuration? >> >> >> >> profile & level info: >> ===================== >> DQ= 0: Main @ Level 1.2 >> DQ= 1: Scalable High @ Level 1.3 320x180 >> DQ= 2: Scalable High @ Level 1.3 >> DQ= 16: Scalable High @ Level 3 >> DQ= 17: Scalable High @ Level 3.1 >> DQ= 18: Scalable High @ Level 3.1 >> DQ= 32: Scalable High @ Level 4 >> DQ= 33: Scalable High @ Level 4 >> >> # I do not know why JSVM shows wrong resolution 320x192, 640x368, >> but the JSVM decoded resolution is right. >> >> SUMMARY: >> bitrate Min-bitr Y-PSNR U-PSNR >> V-PSNR >> --------- ---------- -------- -------- >> -------- >> 320x192 @ 3.0000 105.3180 3.2400 75.1867 77.6071 75.0764 >> 320x192 @ 6.0000 223.8720 8.4840 67.7323 76.1392 74.2387 >> 320x192 @ 12.0000 338.7904 18.6560 57.3909 77.7631 76.0774 >> 320x192 @ 24.0000 838.7456 34.9696 49.8598 70.5339 67.9347 >> 640x368 @ 3.0000 596.0040 452.5980 65.9925 69.5955 66.0604 >> 640x368 @ 6.0000 1227.1800 928.7520 60.0840 67.7073 65.3157 >> 640x368 @ 12.0000 1904.9728 1436.2048 54.8221 65.1333 62.7191 >> 640x368 @ 24.0000 4268.8896 3183.4624 52.2276 63.3834 60.6160 >> 1280x720 @ 3.0000 861.6300 607.2900 65.9835 69.4145 66.7894 >> 1280x720 @ 6.0000 1745.2800 1222.5960 59.6164 67.9509 64.2980 >> 1280x720 @ 12.0000 2785.9584 1945.9072 54.6809 66.0932 63.4704 >> 1280x720 @ 24.0000 5871.8528 4134.2336 54.5693 63.6624 61.2874 >> >> Regards, >> Siyuan >> --- >> >> >> >> On Tue, Feb 8, 2011 at 10:38, Siyuan Xiang <xia...@gm... >> <mailto:xia...@gm...>> wrote: >> >> Thank you Mickaël and Médéric. >> >> I was using debug mode and SVC_decoder_v9 in >> \src_1.10\Libs\SVC\build\PC.c. >> Now it is much faster with Release_MMX configuration. >> >> >> Regards, >> Siyuan >> --- >> >> >> >> >> On Tue, Feb 8, 2011 at 02:34, Médéric Blestel >> <Med...@in... >> <mailto:Med...@in...>> wrote: >> >> Dear Siyuan, >> >> I'm using TCPMP or mplayer to do the benchmarcks with >> give a timing and a frame rate more accurate. >> The decoder is compilled with the Release_MMX configuration. >> I am using windows 7 with visual studio 2008 professional. >> >> kind regards, >> Médéric >> >> Le 08/02/2011 10:56, Siyuan Xiang a écrit : >>> Dear Médéric, >>> >>> Thanks for your reply. I am using windows 7 with visual >>> studio 2008 express. >>> >>> >>> Regards, >>> Siyuan >>> --- >>> >>> >>> >>> On Tue, Feb 8, 2011 at 01:30, Médéric Blestel >>> <Med...@in... >>> <mailto:Med...@in...>> wrote: >>> >>> Dear Siyuan, >>> >>> Thank you for using OpenSVCDecoder. >>> >>> The Benchmarks on this page are quite old (version >>> 1.07, sorry for that!!), but seems to still be updated. >>> I just try on my computer and the sequence was >>> decoded at 80 fps on a 2.93Ghz processor. >>> >>> What is your operating system? >>> >>> kind regards, >>> Médéric >>> >>> Le 08/02/2011 10:12, Siyuan Xiang a écrit : >>>> Hi, all, >>>> >>>> I played with the open SVC codec recently. I am >>>> quite impressive with the excellent work. When I >>>> use the main program PC.c to decode SVCBST-15 with >>>> 1280x720, but I did not think the playback frame >>>> rate is as fast as reported in >>>> http://sourceforge.net/apps/mediawiki/opensvcdecoder/index.php?title=Decoder_Features >>>> >>>> >>>> My computer is intel core 2 2.24GHz, the source >>>> code is the src 1.1.0 version. Is there anything I >>>> am missing? Should MMX , SSE or other optimization >>>> be enabled? >>>> >>>> >>>> Regards, >>>> Siyuan >>>> --- >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: >>>> Pinpoint memory and threading errors before they happen. >>>> Find and fix more than 250 security defects in the development cycle. >>>> Locate bottlenecks in serial and parallel code that limit performance. >>>> http://p.sf.net/sfu/intel-dev2devfeb >>>> >>>> >>>> _______________________________________________ >>>> Opensvcdecoder-support mailing list >>>> Ope...@li... <mailto:Ope...@li...> >>>> https://lists.sourceforge.net/lists/listinfo/opensvcdecoder-support >>> >>> >>> >>> ------------------------------------------------------------------------------ >>> The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: >>> Pinpoint memory and threading errors before they happen. >>> Find and fix more than 250 security defects in the development cycle. >>> Locate bottlenecks in serial and parallel code that limit performance. >>> http://p.sf.net/sfu/intel-dev2devfeb >>> >>> >>> _______________________________________________ >>> Opensvcdecoder-support mailing list >>> Ope...@li... <mailto:Ope...@li...> >>> https://lists.sourceforge.net/lists/listinfo/opensvcdecoder-support >> >> >> >> >> ------------------------------------------------------------------------------ >> The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: >> Pinpoint memory and threading errors before they happen. >> Find and fix more than 250 security defects in the development cycle. >> Locate bottlenecks in serial and parallel code that limit performance. >> http://p.sf.net/sfu/intel-dev2devfeb >> >> >> _______________________________________________ >> Opensvcdecoder-support mailing list >> Ope...@li... <mailto:Ope...@li...> >> https://lists.sourceforge.net/lists/listinfo/opensvcdecoder-support > > > > ------------------------------------------------------------------------------ > The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: > Pinpoint memory and threading errors before they happen. > Find and fix more than 250 security defects in the development cycle. > Locate bottlenecks in serial and parallel code that limit performance. > http://p.sf.net/sfu/intel-dev2devfeb > > > _______________________________________________ > Opensvcdecoder-support mailing list > Ope...@li... > https://lists.sourceforge.net/lists/listinfo/opensvcdecoder-support |