Patel, K.; Smith, B.C.; Rowe, L.A., "Performance of a software MPEG video decoder," p. 75-82, Proceedings ACM Multimedia 93. Anaheim, CA, Aug. 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....range of cache sizes. For the other cases, performance is significantly improved with a relative execution time minimum for software prefetching of 0.36 for the 256KB cache size. Again, performance is never degraded by applying software prefetching. C. Mpeg Finally, data for Berkcley s mpeg play [17] is pre sented in figure 6. The movie caster.mpg is chosen as the training data, and the movie hula.mpg is chosen as the reference data. In the range from approximately 16KB to 256KB, relative execution time is approximately 0.70 of what it would be with no prefetching. A thorough investigation ....

K. Patel, B.C. Smith, and L.A. Rowe, "Performance of a software MPEG video decoder," in Proceedings A CM Multimedia g3, August 1993, pp. 75-82.

....Earmarked Research Grant under RGC 97 98 54, and by the FRG under FRG 96 97 II 103. 1 Introduction The advancement and development in multimedia systems and network applications have led to the development of distributed video systems for streaming video over an open network, like the Internet [2, 3, 6, 10, 11, 12, 15]. In the meantime, MPEG is becoming the de facto standard of video compression, making MPEG video players the mainstream of video display in multimedia applications. The functionality of these MPEG video players is simple: it continuously carries out the execution cycle of fetching for the MPEG ....

....a MPEG video player system is in human perspective. It is common to trade off the quality of services with the smoothness of video playing when a MPEG video player is not operating in an ideal environment, such as with a slow network or slow processor. This is done by skipping some of the frames. [6, 8, 12] The best choice is to skip as many frames as necessary and at the same time providing the highest possible video quality. This paper reports our research on the development of a scheme on optimizing the skipping of frames with respect to QoS Human. In the rest of this paper, we first give an ....

K. Patel, B.C. Smith, and L. A. Rowe, "Performance of a Software MPEG Video Decoder", Proceedings of A CM Multimedia 95, pp. 75-82.

....and Devetsikiotis, Izquierdo and Reeves, and Krunz, Sass, and Hughes [10 12] For the quality of service support, various QoS control mechanisms have been proposed. While Bolot and Turletti proposed a rate control mechanism for transmitting packet video over the Internet [34] Reibman et al. [24,35] and Reininger et al. [25] made use of an adaptive congestion control scheme to study the problem of transmitting VBR video over ATM networks. Furthermore, Hasegawa and Kato had implemented and evaluated a video system with congestion control based on a two level rate control [33] Also in our ....

K. Patel, B.C. Smith, and L. A. Rowe "Performance of a Software MPEG Video Decoder", Proceedings of ACM Multimedia 93, pp 75-82

....have several options (none of which are adequate) They can develop code from scratch, but the complex nature of common multimedia encoding schemes (e.g. MPEG) makes this approach impractical. For example, several person years of work went into writing the Berkeley MPEG player (mpeg play) [58]. A more commonly used option is to modify an existing code base to add the desired functionality. For example, many researchers have hacked up mpeg play to test their ideas. However, this approach requires understanding thousands of lines of code and usually results in complex, unmanageable ....

K. Patel, B. C. Smith, and L. Rowe. Performance of a software MPEG video decoder. In Proceedings of the First ACM International Conference on Multimedia, pages 75--82, Anaheim, CA, August 1993.

....in half. In order to hold the idea, we need to assume that there is no memory operation on MPEG decoding because frame rate scaling does not scale the memory access time. This assumption is valid as memory operation time in MPEG decoding is relatively small compared to the CPU operation time [20]. Secondly, we exclude the display function of MPEG players because the state of the art technology executes those routines separately on other hardware devices. Only IDCT and reconstruction are considered in our experiment, which are the main part of the decoder, and parsing and other minor ....

K. Patel, B. Smith and L. Rowe, "Performance of a Software MPEG Video Decoder," First ACM Int'l Conf. on Multimedia, pp. 75 82, 1993.

....predicted frames that use motion estimation to extrapolate from previous I frames, while B frames result from a linear combination of predictions from both preceding and ensuing I or P frames. The MPEG implementations used in this work are mpeg play and mpeg encode from the Berkeley MPEG Group [PSR93] The lower portion of figure 2.1 illustrates the three frame types discussed above. MPEG2 is an enhancement of MPEG to allow for an extended range of applications. The primary application intended for MPEG2 was digital television, but it is considered an improvement over MPEG for a number of ....

....machine. The time command divides the time spent into user, system, and other time. Finally, gprof was used as a third check on total execution time. Gprof gives additional information that indicates the temporal breakdown by procedure. We divide the procedures into functional blocks after [PSR93] The procedural breakdown was double checked using ATOM to count execution cycles for time spent in individual routines and comparing that to the time reported by gprof. In general, the times were close, but did not compare exactly. The particular data is not presented here.The experimental ....

[Article contains additional citation context not shown here]

K. Patel, B.C. Smith, and L.A. Rowe. Performance of a software MPEG video decoder. In Proceedings ACM Multimedia 93, pages 75--82, August 1993.

....in half. In order to hold the idea, we need to assume that there is no memory operation on MPEG decoding because frame rate scaling does not scale the memory access time. This assumption is valid as memory operation time in MPEG decoding is relatively small compared to the CPU operation time [25]. Secondly, we exclude the display function of MPEG players because the state of the art technology executes those routines separately on other hardware devices. Only IDCT and reconstruction are considered in our experiment, which are the main part of the decoder, and parsing and other minor ....

K. Patel, B. Smith and L. Rowe, "Performance of a Software MPEG Video Decoder," First ACM Int'l Conf. on Multimedia, pp. 75 - 82, 1993. 16

....two pass onedimensional DCT (column row) or may implement a one pass two dimensional DCT [8, 2] Tweaks to optimize performance at the cost of accuracy. In order to maximize the decoding rate, a decoder may not perform all the calculations. One such example is using the Berkeley MPEG 1 decoder [5] with half pixel motion estimation searches. If the motion vectors in the B frames are both half pixel in size, the Berkeley decoder will not completely average all four pixels that make up the motion compensated pixel. Their results show that there is very little visual difference. However, all ....

B. S. K. Patel and L. A. Rowe. Performance of a software mpeg video decoder. In Proc. of the conference on Multimedia '93, August 1993.

....computer architects. Before architects commit a Mach 3.0 X11 Display BSD 4.3 Server Kernel Server Emulation Library mpeg play 40 5 30 25 Figure 1. 1 The MPEG Decoder Workload mpeg play decodes an MPEG compressed video stream and displays the resulting sequence of frames in an X window [Patel92]. We will frequently use mpeg play throughout this dissertation as an example of an application that presents a several difficulties to accurate memory system simulation. The components of mpeg play, depicted in this figure, include a dynamicallylinked emulation library, an X11 display server, ....

....variance if the simulation for a given memory configuration is repeated. The precise sequence of traps that drive a Tapeworm simulation, however, are Workload Description mpeg play mpeg play (version 2.0) from the Berkeley Plateau Research Group. Displays 85 frames from a compressed video file [Patel92]. jpeg play The xloadimage (version 3.0) program written by Jim Frost. Displays two JPEG images. gs Ghostscript (version 2.4.1) from the Free Software Foundation. Renders and displays a single postscript page with text and graphics in an X window. verilog Verilog XL (version 1.6b) simulating ....

Patel, K., Smith, B. C. and Rowe, L. A. Performance of a Software MPEG Video Decoder. Technical Report, University of California, Berkeley. 1992.

....To stress the protection mechanism, we created a set of soft tasks characterized by a very high variance in their execution times. A typical periodic task characterized by a large execution time variance is an MPEG [8] player (all experiments were performed using the Berkeley MPEG library [16], which was ported to HARTIK) Then, we performed a simple experiment with 2 tasks, 1 and 2 , having periods T 1 = 125ms (8 frames per second FpS) and T 2 = 30ms (33 FpS) Figure 3 plots the number of decoded frames as a function of time when the two periodic tasks are scheduled by the ....

K. Patel, B. C. Smith, and L. A. Rowe. Performance of a software mpeg video decoder. In Proceedings of ACM Multimedia 93, Anaheim, CA, 1993.

....create terminate, requesting reservation parameters, binding threads to reserves, and extracting usage information about reserves. Several real applications were modified to use processor reserves including: a QuickTime video player developed at CMU called QTPlay, an MPEG decoder called mpeg play [93], and a version of the X Server [34] A version of the user level socket library [70] was modified to use reserves as well. This socket implementation supports predictable performance for applications that send and receive network packets. A QOS manager was implemented to allow for more ....

....period, the thread displays a frame of the video and then saves the start time and completion time for the frame in a buffer in memory. Just before the player exits, it dumps the contents of this timestamp buffer to a file for subsequent analysis. 5.3. 2 MPEG decoder The Berkeley MPEG decoder [93] was modified to use processor reserves in RT Mach. This version of mpeg play reserves processor capacity during its initialization and periodically evaluates its performance and makes adjustments to its processor reservation and timing constraints as necessary. The original Berkeley MPEG decoder ....

K. Patel, B. C. Smith, and L. A. Rowe. Performance of a Software MPEG Video Decoder. In Proceedings of the First ACM International Conference on Multimedia, pages 75--82, August 1993.

....extension for MPEG 1 Although layered scalability extensions are present in several of the common video compression standards, freely available implementations are not available. For our purposes, the easiest solution was to develop a rapid prototype with an open source MPEG 1 software codec [14]. We call our modified MPEG 1 format SPEG[10] SPEG adds spatial scalability to MPEG 1, through layered quantization of DCT data. Our current implementation has four levels of SNR scalability in each MPEG 1 picture. Our implementation does transcoding from MPEG 1 to add the spatial scalability, ....

K. Patel, B. C. Smith, and L. A. Rowe. Performance of a software mpeg video decoder. In Proceedings ACM Multimedia 93, pages pp. 75--82, Anaheim, CA, August 1993.

....amount of information to be transmitted on digital channels by up to two order of magnitude. Unfortunately motion estimation is also the weak point of multimedia protocols in terms of computing power required. While multimedia decoding routines can run on last generation workstations in real time[1][2] on the contrary software real time multimedia coders are very far from being successfully implemented on desktop computers unless equipped with special purpose hardware[6] 5] because of the computing intensive nature of the motion estimation modules[4] This work was supported by MURST. ....

B.C. Smith K. Patel and L.A. Rowe. "Performance of a Software MPEG video decoder". "UCB/ERLM93/2 ", 1993.

....bit rate cannot be reached by changing QP alone, the input sequences are subsampled to result in lower frame rates. This occurs when we try to compress Raider using GOP sizes of 4 and 8 and Mother using GOP size of 4 at a rate below 128 kbps. The MPEG results are generated using an MPEG 1 software [14] with 10 slices per frame and exhaustive search. Target bit rates are set so as to match those produced by H.263. The MPEG 1 software dynamically adjusts the compression quality of each frame using its own rate control mechanism, which is essentially based on a leaky bucket al..gorithm. For our ....

....the same range varies from 60 to 25 frames per second. The reported times exclude disk access and display time [8] We see that real time software only encoding and decoding are possible on current computers. Fig. 8 shows the speed comparison of our proposed scheme and that of an MPEG 1 software [14] on an 170 MHz UltraSparc. For the sequence at bit rates ranging from 500 kbps to 3 Mbps, our encoding proceeds at 18.2 to 11.2 fps and decoding proceeds at 17.5 to 12.5 fps, respectively. On the same machine MPEG encoding proceeds at 0.4 1.6 fps using exhaustive and logarithmic motion vector ....

K. Patel, B. Smith, and L. Rowe, "Performance of a software MPEG video decoder," in Proc. ACM Multimedia '93 pp. 75--82.

No context found.

K. Patel, B. Smith, and L. Rowe. Performance of a software MPEG video decoder. In Proceedings of the First ACM International Conference on Multimedia, pages 75--82, August 1993.

No context found.

K. Patel, B. Smith, and L. Rowe. Performance of a software MPEG video decoder. In Proceedings of the First ACM International Conference on Multimedia, pages 75-82, August 1993.

....the input sequences are subsampled to result in lower frame rates. This input subsampling occurs when we try to compress Raider using GOP sizes of 4 and 8 and Mother using GOP size of 4 at a rate below 128 kbps. The MPEG results are generated using the Berkeley MPEG 1 encoder [49] and decoder [96]. We used frame and the reconstructed frame: PSNR = 10 log 255 MPEG 4 FGS uses a non scalable MPEG 4 base layer and many fine granularity enhancement layers. The software from the University of British Columbia is used. Further information about the Berkeley MPEG encoder decoder and ....

....in the bit rate range of 200 kbps to 2 Mbps respectively. The decoding speeds in the same bit rate range varies from 60 to 25 frames per second. The reported times exclude disk access and display time [124] Fig. 3. 20 shows the speed comparison of SVIP and that of the Berkeley MPEG 1 decoder [96] on an 170 MHz UltraSparc 1. For the Raider sequence at bit rates ranging from 500 kbps to 3 Mbps, the SVIP encoding rate is 18.2 to 11.2 fps and the decoding rate is 17.5 to 12.5 fps, respectively. On the same machine the MPEG encoding rate is 0.4 to 1.6 fps using exhaustive and logarithmic ....

K. Patel, B. Smith, and L. Rowe. Performance of a software mpeg video decoder. In Proc. ACM Intl. Multimedia Conf. '93, pp. 75--82, Anaheim, CA., August 1993.

....for each task. 6.4 Video playback Multimedia applications are becoming more common on portable computers [6] Therefore, we include a movie player as one of our workloads. We use the MPEG player included with the Berkeley MPEG Tools developed by the Berkeley Multimedia Research Center (BMRC) [16]. Since they provide full source code for their tool, we were easily able to instrument it to measure and output the CPU time taken for each frame. Thus, each task of the workload represents the processing of one frame. We obtained animations to use from the same BMRC FTP site as the MPEG ....

K. Patel, B. Smith, and L. Rowe. Performance of a software MPEG video decoder. In Proceedings of the First ACM International Conference on Multimedia, pages 75--82, August 1993.

No context found.

Patel, K.; Smith, B.C.; Rowe, L.A., "Performance of a software MPEG video decoder," p. 75-82, Proceedings ACM Multimedia 93. Anaheim, CA, Aug. 1993.

No context found.

K. Patel, B. C. Smith, and L. A. Rowe, "Performance of a software MPEG video decoder," in Proc. ACM Multimedia'93, Aug. 1993, pp. 75--82.

No context found.

K. Patel et al., \Performance of a software mpeg video decoder," in Proceedings of ACM Multimedia 1993.

No context found.

K. Patel, B. Smith, and L. Rowe, "Performance of a software MPEG video decoder," in Proc. ACM Multimedia '93, (Anaheim, CA), pp. 75--82, 1993.

No context found.

K. Patel, B.C. Smith, L.A. Rowe,"Performance of a Software MPEG Video Decoder", In Proceedings ACM Multimedia

No context found.

K.Patel, B.Smith and L.Rowe, Performance of a Software MPEG Video Decoder, Proc. ACM Multimedia '93.

No context found.

K. Patel, B.C. Smith, L.A. Rowe; Performance of a Software MPEG Video Decoder, Proceedings 1st ACM Conference on Multimedia, pp.83-90, Aug 1993.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC