Abhishek, The link provided by Jayakrishnan is a good starting point. In addition, TI has a whole set of FAQs and usage notes for DVSDK available at: http://tiexpressdsp.com/index.php?title=Main_Page
IMHO, the following should be your approach: 1. Decide whether the video processing algorithm will be implemented on ARM or DSP. If it involves a lot of signal processing, the answer is clear. The best option would be to use DSP. You can find a few notes on developing codecs at: http://tiexpressdsp.com/index.php?title=Category:Codecs 1. Develop the codec on DSP using CCS, verify functionality and fine tune and optimize for performance. You can refer to the following for CCS: http://tiexpressdsp.com/index.php?title=Category:CCS 1. Ensure that the codec is xDAIS and xDM compliant. The xDAIS is an algorithm standard followed by all codecs on TI platforms. It has a defined interface for resource request, algorithm instantiation, actual processing and cleanup. xDM is an extension of xDAIS that caters to Multimedia Codecs. These standards ensure that the codec adhering to the standard can be easily plugged into any application using standard frameworks. Topics on xDAIS are at: http://tiexpressdsp.com/index.php?title=Category:XDAIS http://wiki.davincidsp.com/index.php?title=XDM_FAQ 1. When the algorithm is robust, we need a framework to control it. Someone needs to allocate memory, grant DMA channels, take care of ARM-DSP message transfers, power off/on DSP when required, etc. TI's DVSDK (Digital Video Software Development Kit) provides a framework for exactly this purpose. The Codec Engine framework handles all the internal details and leaves the application writer to worry about only the application control logic. An overview of DVSDK and details on Codec Engine here: http://tiexpressdsp.com/index.php?title=DVEVM_and_DVSDK_overview http://tiexpressdsp.com/index.php?title=Category:DVSDK http://tiexpressdsp.com/index.php?title=Category:Codec_Engine http://wiki.davincidsp.com/index.php?title=RTSC_Codec_And_Server_Package_Wizards 1. When the codec gets integrated with the framework, the result is a DSP image which can be loaded on the DSP. This is called as a codec server/codec combo: http://wiki.davincidsp.com/index.php?title=RTSC_Codec_And_Server_Package_Wizards 1. You would still need an ARM application to call the codec on DSP and control it. The Digital Video Test Bench provides this functionality and enables easy plug-n-play of codecs: http://wiki.davincidsp.com/index.php?title=I_just_want_my_video_codec_to_work_with_the_DVSDK This page provides a wholistic view of what's involved in getting a video codec to be called from ARM application: http://wiki.davincidsp.com/index.php?title=I_just_want_my_video_codec_to_work_with_the_DVSDK Best Regards, Anand Balagopalakrishnan Texas Instruments India ________________________________ From: [email protected] [mailto:[email protected]] On Behalf Of Jaya krishnan Sent: Friday, April 24, 2009 6:34 AM To: Abhishek Botadra; [email protected] Subject: Re: Real-time video processing on DM6467 EVM Abhishek, You don't need to connect two hosts to the DVEVM. You develop the algorithm in CCS, debug,optimize, make it xDM compliant and produce a library file. The rest of the development work is done on Linux host. Pls go through an earlier post http://www.mail-archive.com/[email protected]/msg01049.html You may also go through several TI documents such as Algorithm creator user's guide, Server Integartor user's guide and Algorithm development user's guide. Regards Jayakrishnan ------- Original Message ------- Sender : Abhishek Botadra<[email protected]> Date : Apr 24, 2009 07:58 (GMT+09:00) Title : Real-time video processing on DM6467 EVM Hi, I need some help in figuring out how to implement a video-processing algorithm on ARM+DSP system. >From various sources, I am able to understand that first I need to develop the >code on DSP because real-time execution can be achieved only on DSP (using >code composer studio). I need to work on Windows host for this work. Then, I >need to package up the code, i.e. write C/C++ wrappers so that it can be >loaded and executed from ARM running linux. Again, this development work needs >to be done on Linux host. So, in the end I will need to connect both the hosts >to the EVM, windows host through JTAG emulator and serial port and Linux host >through Ethernet. Also, there are no reference frameworks available for DaVinci Processors. So, I was wondering if you guys can let me know if this would be the best approach in implementing the video processing algorithm on DM 6467 EVM. Or, is there any other better way I should opt for. Thanks for the help. Regards, -- Abhishek size=2 width=410 style='width:307.5pt' align=left> [http://www.samsungcamera.co.kr/images/images/logo.gif] Jayakrishnan M M Research Engineer Development Team Digital Imaging Division SAMSUNG TECHWIN CO.,LTD TEL +82-31-740-8677 FAX +82-31-740-8750 Mobile +82-10-6409-3619 E-mail:[email protected]<mailto:[email protected]> ________________________________
_______________________________________________ Davinci-linux-open-source mailing list [email protected] http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
