Register Allocation and Binding for Low Power
read more
Citations
Power minimization in IC design: principles and applications
A predictive system shutdown method for energy saving of event-driven computation
Energy minimization using multiple supply voltages
Inherently lower-power high-performance superscalar architectures
High-level power modeling, estimation, and optimization
References
Probability, random variables and stochastic processes
Probability, random variables, and stochastic processes
Combinatorial optimization: algorithms and complexity
Algorithmic graph theory and perfect graphs
Synthesis and optimization of digital circuits
Related Papers (5)
Frequently Asked Questions (20)
Q2. What future works have the authors mentioned in the paper "Register allocation and binding for low power" ?
Their future work will focus on the register assignment for pipelined design and data ow graph with outer loops.
Q3. What are the current approaches to calculating the pdf of primary input variables?
The existing approaches include rule-based [6], greedy or iterative [10], branch and bound [13], linear programming [1], and graph theoretic, as in the Facet system [18], the HAL system [16] and the EASY system [17].
Q4. What is the driving factor behind the push for low power design?
One driving factor behind the push for low power design is the growing class of personal computing devices as well as wireless communications and imaging systems that demand high-speed computations and complex functionalities with low power consumption.
Q5. What is the probability of switching on R?
In the register allocation phase, if several compatible arcs are assigned to the same register R, the switching on R will occur whenever one stored data value is replaced by another data value.
Q6. What is the simplest way to calculate the switching activity between pairs of arcs?
Having calculated the switching activity between pairs of arcs that could potentially share the same register and given the number of registers that are to be used, the register assignment problem for minimum power consumption is formulated as a minimum cost clique covering of the compatibility graph.
Q7. What is the meaning of the word pdf?
In addition to the calculation of pairwise joint pdfs, the pdf of any internal arc is needed to calculate the total switching activity of the set of registers.
Q8. What is the life time of each arc in a scheduled data ow graph?
The life time of each arc (data value) in a scheduled data ow graph is the time during which the data value is active (valid) and is de ned by an interval [birth time; death time].
Q9. what is the way to solve the max-cost ow problem?
The easiest method to solve the max-cost ow problem is to negate the cost of each arc in the network, and run the min-cost ow algorithm on the new network [14].
Q10. What is the ow value for the arc?
To demonstrate that the switching activity calculation based on the joint pdf is necessary to obtain a low power register assignment the authors performed an experiment where every arc weight in the compatibility graph was set to some constant (C = 100) and then ran the max-cost ow for di erent ow values.
Q11. What is the cdf of the new random variable y?
The cdf (cumulated distribution function) of the new random variable y is de ned as G(y) = prob(Y y), which is equal to prob(w(x1; x2; : : : ; xn) y).
Q12. What is the simplest way to minimize the total power consumption on the registers?
To minimize the total power consumption on the registers, a network NG = (s; t; Vn; En; C;K) is constructed from the compatibility graph G0 = G(V; A).
Q13. What is the power consumption of a register?
In any case, minimizing the switching activity at the output of the registers will minimize the power consumption regardless of the speci c load seen at the output of the registers.
Q14. What is the method for calculating the joint pdf of two random variables described in section 2.1?
The method for calculating the joint pdf of two random variables described in section 2.1 is mainly suitable for the case when the variables in the system are of continuous type.
Q15. What is the simplest way to calculate the joint pdf?
When however the precision used to represent thediscrete numbers is high enough or the variance of the underlying distribution is not too large, the continuous type pdf gxy(x; y) can be used as a good approximation for the discrete type pdf fxy(x; y) after being multiplied by the scaling factor ( P (x;y)2B gxy(x;y))
Q16. What is the compatiblity graph for the arcs in a scheduled?
The authors will show that the unoriented compatiblity graph for the arcs (data values) in a scheduled data ow graph without cycles and branches is a comparability graph (or transitively orientable graph) which is a perfect graph [5].
Q17. What is the ow value on each path?
The ow value on each path is one, this implies that the total cost on each individual path is the sum over all individual arcs on that path according to their topological order in the graph G0 = G(V; A), where the cost on each arc is a linear function of the \\Saved Power".
Q18. what is the arc from v to v0 in the unoriented compatibility graph?
with j f j = k, in the network N 0G corresponds to a set of vertex disjoint cliques 1; 2; : : : ; k in the unoriented compatibility graph G00.
Q19. What is the simplest way to calculate the joint pdf of the input variables?
If the authors are given the joint pdf of the input random variables of a data ow graph, then the joint pdf of any pair of values (arcs in the data ow graph) can be calcualted [15].
Q20. What is the ow value on the graph G0?
Note that, the max-cost ow on N 0G always nds the clique covering that covers all of the vertices in the original graph G0 whenever the ow value j f j is larger than or equal to kmin.