Robust media processing in a flexible and cost-effective network of multi-tasking coprocessors

Evaluating schedulers for multimedia processing on buffer-constrained SoC platforms

“…The most promising way to overcome this drawback is to use intelligent platform management and scheduling techniques. 6,7 A typical SoC platform for multimedia applications consists of a heterogeneous collection of fully programmable processing elements (PEs)-for example, MIPS Technologies, ARM, and TriMedia processorsand coarse-grained application-specific coprocessors optimized for specific tasks. The task structure representing the application to be executed on such a plat-form includes a set of concurrently executing tasks that exchange information solely through unidirectional data streams.…”

“…The task structure representing the application to be executed on such a plat-form includes a set of concurrently executing tasks that exchange information solely through unidirectional data streams. 5,6 Figure 1 shows a simple system-level model of a settop box device implementing an audio-video decoder. Audio and video streams enter the device, which includes two PEs: PE 1 and PE 2 .…”

Eclipse: heterogeneous multiprocessor architecture for flexible media processing

“…The task scheduler is based on round-robin style task selection as this can be efficiently implemented in hardware. The scheduler uses a weighted round-robin scheme, where the weights or budgets are configured as a guaranteed minimum number of cycles that a task may continuously execute, irrespective of the resource requirements of other tasks [21]. These budgets typically range from 1000 up to 10,000 clock cycles (10-100 processing steps).…”

“…Therefore, the runnable test of Fig. 4 provides a 'best guess' by considering both the available data and room in the stream buffers as well as on denied data access (For details, see [21]). Section 4.3 shows that this information is locally available in the shell.…”

Meeting CPU constraints by delaying playout of multimedia tasks

“…macroblocks or frames in the case of MPEG decoding), the number of processor cycles required to process each data item is highly variable. The ratio of the worst-case and the average load on a processor running such an application can be as high as a factor of 10 [15]. The playout rate associated with the application (e.g.…”