Voltage Assignment with Guaranteed Probability Satisfying Timing Constraint for Real-time Multiproceesor DSP

Abstract

Dynamic Voltage Scaling (DVS) is one of the techniques used to obtain energy-saving in real-time DSP systems. In many DSP systems, some tasks contain conditional instructions that have different execution times for different inputs. Due to the uncertainties in execution time of these tasks, this paper models each varied execution time as a probabilistic random variable and solves the Voltage Assignment with Probability (VAP) Problem. VAP problem involves finding a voltage level to be used for each node of an date flow graph (DFG) in uniprocessor and multiprocessor DSP systems. This paper proposes two optimal algorithms, one for uniprocessor and one for multiprocessor DSP systems, to minimize the expected total energy consumption while satisfying the timing constraint with a guaranteed confidence probability. The experimental results show that our approach achieves significant energy saving than previous work. For example, our algorithm for multiprocessor achieves an average improvement of 56.1% on total energy-saving with 0.80 probability satisfying timing constraint.

Towards Power-Efficient Smartphones by Energy-Aware Dynamic Task Scheduling

“…At runtime, the real execution time and energy consumption may exhibit high variations [11], [12], due to process variability, physical faults, and voltage/frequency changes. In our model, we expect each core in the same processor can adjust its voltage and frequency independently.…”

Energy saving for memory with loop scheduling and prefetching

“…Using algorithm Mode_Assign [4], which is a near-optimal algorithm with dynamic programming, we obtain the solution table for the mode assignment problem. The solution table of modes assignment for schedule in Figure 1 (a) is shown in Table 1.…”

“…They can change to low power operating modes (with low operating velocity) independently when they are not used for a while. Hence, same type of operations can be processed under different mode with different energy consumptions [4]. Therefore, an important problem arises: how to assign a mode to each operation of an application such that the requirements can be met and the total energy consumption can be minimized with a guaranteed confidence probability satisfying timing constraints.…”

Dynamic and Leakage Power Minimization with Loop Voltage Scheduling and Assignment

“…[x, y] ← The earliest available location of u from row 1 to row t in S based on the precedence relation in G r ; 10: if not find then 11: [x, y] ← The earliest available location of u in row t + 1 in S; 12: end if 13: Put u into [x, y]; 14: end for 15: for all u ∈ U do 16: a ← The switching activity of node u: Computed by Hamming distance between u and its immediately before node on the same processor; 17: b ← 0, b is the voltage switching overhead of node u; 18: b ← 1, when there is a voltage switching; 19: end for 20: Use algorithm VAP M [23] to obtain MIN(E) voltages assignment A with (Prob. (T ≤ L)) ≥ P for the schedule; 21: S i ← S; r i ← r; A i,j ← A; 22: end for 23: E min ← MIN(E total ); 24: Store the S i , r i , and A i,j corresponding to E min ; 25: Output S i , r i , and A i,j ;…”

“…Then for each new schedule, use algorithm VAP M [23], which using a novel dynamic programming approach, to obtain the minimum total energy E with Prob. (T ≤ L) ≥ P. LSAMP will repeat the above procedure R times, where R is a user specified amount.…”