Claude Castelluccia and Phillip Hoschka. A compiler-based approach to protocol optimization. In High Performance Communication Subsystem, August 1995.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... automaton is dened by the x transitions that use y of the protocol execution time (where x is typically equal to 10 and y to 90) Therefore the common path can be identied from the probabilities of the transitions [30] In HIPPCO, those probabilities are computed by a classic Markov analysis [9] of the automaton. In the automata generated by Esterel, each state tree is composed of initialization actions followed by independent subtrees processing the dioeerent types of input events. A test at the root of each subtree determines whether the incoming event should be processed by this ....

....is then equal to the product of the following terms: the visit probability of state S 1 , the occurrence probability of the ring input, and the probabilities of the visited branches. Thus: P (T 1;2 ) P (S 1 ) Theta P (packet) Theta P ( size T ) false) Theta P (checksum = true) In [9], we showed that the visit counts V j of the state S j can be computed by solving the following system of n linear equations (where n is the number of states) V j = ffi 1j n X k=1 V k Theta P (T k;j ) where ffi ij = 1 for i = j, and 0 otherwise. The visit probability of a state S j is ....

Claude Castelluccia and Phillip Hoschka. A compiler-based approach to protocol optimization. In High Performance Communication Subsystem, August 1995.

....sometimes rely on user level communication stacks or on specific networks. A completely different approach to minimize the memory access costs, that does not alter the communication stack architecture, consists in improving the instruction cache hit ratio, either with a compiler oriented technique [Mosberger95, Castelluccia95] or run time linker oriented technique [Montz94] Optimization of the communication system software environment Two optimized communication system frameworks have been designed: x Kernel [Hutchinson91] and an improved BSD derived architecture [Jacobson92, Jacobson93] x Kernel, like STREAMS, ....

C. Castelluccia, P. Hoschka, "A Compiler-Based Approach to Protocol Optimization". Third IEEE Workshop on the Architecture and Implementation of High Performance Communication Subsystems, August 1995.

....version of this system. ffl Common Uncommon path separation: Another way to enhance the performance is to reduce the amount of processing per operation. Improving the performance of the protocol operations is classically achieved by optimizing the operations that are on the common path [HH93, CH95] The common path, also called the critical path or the fast path, is the part of the code that is the most frequently executed. By optimizing this path, the execution RR n Sigma2748 14 C. Castelluccia , W. Dabbous time of the overall program under normal conditions will be optimized. This ....

Claude Castelluccia and Phillip Hoschka. A compiler-based approach to protocol optimization. In High Performance Communication Subsystem, August 1995.

No context found.

Castelluccia, C., & Hoschka, P. (1995). A Compiler-Based Approach to Protocol Optimization. Proceedings "Third IEEE Workshop on the Architecture and Implementation of High Performance Communication Subsystems".

....a Markov analysis of the code using prediction information provided by the application designer at the protocol compilation time. This solution is easy to perform because it is performed at compile time, but it still gives accurate results because the prediction is customized to the application. [7, 8] give more details on the algorithms implemented in HIPPCO to identify the common path. 4 Tree Optimizer: Common Path Optimization In this section, we present the tree transformations that optimize the protocol common path in HIPPCO. These transformations improve the common path performance ....

Claude Castelluccia and Phillip Hoschka. A compiler-based approach to protocol optimiza tion. In S.-P. Chang, editor, Proceeding of the High Performance Communication Subsystem Workshop, Mistic, Connecticut, USA, August 1995.

....identied from some user or proler provided predictions about the probability of events and the probability that a given test will succeed or fail. This is done hierarchically: the common states are identied using a Markov analysis, the common branches are detected in each of those states [CH95] Second the code is systematically reorganized to take advantage of information about the common path. In fact, HIPPCO performs a variety of optimizations designed to reduce the number of instructions executed, lower the instruction cache miss rate, and improve pipelining performance. For ....

Claude Castelluccia and Phillip Hoschka. A compiler-based approach to protocol optimization. In High Performance Com munication Subsystem, August 1995.

....version of this system. ffl Common Uncommon path separation: Another way to enhance the performance is to reduce the amount of processing per operation. Improving the performance of the protocol operations is classically achieved by optimizing the operations that are on the common path [HH93, CH95] The common path, also called the critical path or the fast path, is the part of the code that is the most frequently executed. By optimizing this path, the execution time of the overall program under normal conditions will be optimized. This optimization technique has been widely utilized to ....

....of program sections that will be executed with high frequency, the optimizations we consider (such as test outcomes rescheduling or function inlining) require more exact information on the execution frequency. We therefore use a full Markov analysis of the control AEow graph of an Esterel program [CH95] The general idea of Markov analysis is to regard the control AEow graph of a program as a nite Markov chain. The control AEow graph is composed of nodes connected by arcs. It is constructed as follows: the nodes in the graph correspond to sequential sequences of assignement statements or ....

Claude Castelluccia and Phillip Hoschka. A compiler-based approach to protocol optimization. In High Performance Communication Subsystem, August 1995.

....fonction est #conomis# (IC diminue) Ensuite, l #limination de l indirection permet l application de plus d optimisations par le compilateur de bas niveau. La contre partie est une augmentation de la taille du code, qui pourrait ralentir l ex#cution du programme par suite de son eoeet sur le cache [6, 7, 16, 12]. L insertion d une fonction est donc 11 b#n#que si cette fonction est appel#e une seule fois, ou tr#s fr#quemment ou bien si la taille du corps de la fonction est inf#rieur ou #gal au nombre d instructions n#cessaires pour appeler cette fonction. Les compilateurs traditionnels ins#rent en ....

Claude Castelluccia and Phillip Hoschka. A compiler-based approach to protocol optimization. In S.-P. Chang, editor, Proceeding of the High Performance Communication Subsystem Workshop, Mistic, Connecticut, USA, August 1995.

.... of an automaton is dened by the x transitions that use y of the protocol execution time (where x is typically equal to 10 and y to 90) Therefore, the common path can be identied from the probabilities of the transitions [31] In HIPPCO, these probabilities are computed by a Markov analysis [10] of the automaton. The transitions of a state are dened by the dioeerent paths that an input can follow through the tree of this state. They are therefore dened by an input and the se quence of tests (and their results) that lead to the leaf node. In gure 2, the set of transitions of the state S1 ....

....of this transition is then equal to the product of the following terms: the visit probability of state S1 , the occur rence probability of the ring input, and the probabilities of the visited branches. Thus: P (T 1;2 ) P (S1) Theta P (pack) Theta P ( size T ) Theta P (check) In [10], we have shown that the visit counts V j of the state S j can be computed by solving the following system of n linear equations (where n is the number of states) V j = ffi 1j n X k=1 Vk Theta P (T k;j ) where ffi ij = 1 for i = j, and 0 otherwise. The visit probabil ity of a state S j ....

Claude Castelluccia and Phillip Hoschka. A compiler-based approach to protocol optimization. In S.-P. Chang, editor, Proceeding of the High Performance Communication Subsys tem Workshop, Mistic, Connecticut, USA, August 1995.

No context found.

Castelluccia, C., & Hoschka, P. (1995). A Compiler-Based Approach to Protocol Optimization. Proceedings "Third IEEE Workshop on the Architecture and Implementation of High Performance Communication Subsystems".

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