This paper describes a design framework for TCPfriendly and media-friendly rate control algorithms for multimedia streaming applications. The idea of this framework is to start from TFRC's (TCP-Friendly Rate Control) transmission rate and then alter this transmission rate so that it tracks the media characteristics of the stream (e.g., bitrate) or other application characteristics like the client buffer fill level. In this way, the media-friendly property of the algorithm is achieved. We give three rules that guide how the TFRC throughput should track the evolution of the stream's media characteristics and remain TCPfriendly in the long term. We also present, as proof of concept, four simple media-friendly and TCP-friendly congestion control algorithms built using the aforementioned framework. These congestion control algorithms are better suited for multimedia streaming applications than traditional TCP congestion control or smooth congestion control algorithms like TFRC. We have performed evaluations of two of the four proposed media-friendly and TCP-friendly congestion control algorithms under various network conditions and validated that they represent viable transport solutions, better than TFRC, for variable bitrate video streams. More specifically, our two media-friendly and TCPfriendly congestion control algorithms maintained a TCP-friendly throughput in the long term in all experiments and avoided an empty buffer at the client side in situations when TFRC could not achieve this.

No abstract

In compiler theory, the Banerjee test is a dependence test. The Banerjee test assumes that all loop indices are independent, however in reality, this is often not true. The Bannerjee test is a conservative test. That is, it will not break a dependence that does not exist.This means that the only thing the test can guarantee is the absence of dependence.This paper proposes an innovative algorithm which allows precise determination of information about dependences and can act in situation where certain cycling limits are known.

This paper is intended to be a follow up of the work done by the authors in previous articles. On one hand it is concluded with a theorem that proves to be a definite answer to one very important research direction and on the other hand it is an opening of new research directions in the field of loop structures automatic parallelization.

No abstract