Lowering the latency of interfaces for rationally-related frequencies

Abstract

We have introduced the Globally-Ratiochronous, Locally-Synchronous (GRLS) design paradigm, a design style based on rationally-related frequencies, with the objective to overcome the limitations of traditional multi-frequency systems by providing a flexibility close that of Globally-Asynchronous, Locally-Synchronous (GALS) systems but introducing performance penalties and overheads close to those of mesochronous systems. In this paper we focus on performances and improve the latency figures of our original GRLS interfaces by introducing two new interfaces, called GRLS-F and GRLS-noF, the first suitable for blocks with long computation time and the second for blocks with short computation time. The latency figures of the original GRLS interfaces are improved up to 50% without increasing complexity. The average latency figures of the resulting interfaces are lower than 1 Receiver clock cycle, the latency of a synchronous interface.

A GALS Network-on-Chip based on rationally-related frequencies

“…When f T > f R , the worst-case latency, assuming zero-waitstate buffers can be calculated as: [14]. Latency figures can be reduced using techniques presented in [26]. However, these enhancements are not considered in this work.…”

“…if the setup/hold constraints of strobe or data samplers are violated, then the output of the samplers stabilizes to a random value. Setup and hold times of the samplers are fixed respectively to 10ps and 30ps, to model the values in the reference 90nm technology used in [14], [26]. For the purpose of this analysis, in order to keep the results general, other non-idealities are not considered and the propagation delay through the channel is considered to be null.…”

“…If no contention is encountered, the worstcase latency of a packet travelling between two nodes is ideally given by [14], [26]:…”

“…While standalone GRLS links have been characterized in [14], [26], this is the first study of a complete GRLS system. Our previous work on GRLS focused either on performance [14], [26] or flexibility [15] analysis. Here, the interaction between the two aspects comes into play and their combined effect is studied.…”

Μεθοδολογίες Και Εργαλεία Διαχείρισης Πόρων Και Παραμετροποίησης Εφαρμογών Κατά Το Χρόνο Εκτέλεσης Σε Πολυπύρηνες Ενσωματωμένες Πλατφόρμες

“…all submultiples of f H Σχήμα 3.22: Αρχιτεκτονική της διαχείρισης ισχύος στην πολυπυρήνη πλατφόρμα [30] Τέλος, η εφαρμογή Matrix Multiplication εκτελεί πολλαπλασιασμό πίνακα επί πίνακα.…”