In this paper, we propose a novel microarchitectural technique for run-time power-gating caches of GPUs to save leakage energy. The L1 cache (private to a core) can be put in a low-leakage sleep mode when there are no ready threads to be scheduled, and the L2 cache can be put in sleep mode when there is no memory request. The sleep mode is state-retentive, which precludes the necessity to flush the caches after they are woken up. The primary reason for the effectiveness our technique lies in the fact that the latency of detecting cache inactivity, putting a cache to sleep and waking it up before it is accessed, is completely hidden microarchitecturally. The technique incurs insignificant overheads in terms of power and area. Experiments were performed using the GPGPU-Sim simulator on benchmarks that was set up using the CUDA framework. The power and latency modeling of the cache arrays for measuring the wake-up latency and the break-even periods is performed using a 32-nm SOI IBM technology model. Based on experiments on 16 different GPU workloads, the average energy savings achieved by the proposed technique is 54%.

In this work, we investigate state-retentive power gating of register files for leakage reduction in multicore processors supporting multithreading. In an in-order core, when a thread gets blocked due to a memory stall, the corresponding register file can be placed in a low leakage state through power gating for leakage reduction. When the memory stall gets resolved, the register file is activated for being accessed again. Since the contents of the register file are not lost and restored on wakeup, this is referred to as state-retentive power gating of register files. While state-retentive power gating in single cores has been studied in the literature, it is being investigated for multicore architectures for the first time in this work. We propose specific techniques to implement state-retentive power gating for three different multicore processor configurations based on the multithreading model: 1) coarse-grained multithreading, 2) fine-grained multithreading, and 3) simultaneous multithreading. The proposed techniques can be implemented as design extensions within the control units of the in-order cores. Each technique uses two different modes of leakage states: low-leakage savings and low wake-up and high-leakage savings and high wake-up latency. The overhead due to wake-up latency is completely avoided in two techniques while it is hidden for most part in the third approach, either by overlapping the wake-up process with the thread context switching latency or by executing instructions from other threads ready for execution. The proposed techniques were evaluated through simulations with multiprogrammed workloads comprised of SPEC 2000 integer benchmarks. Experimental results show that in an 8-core processor executing 64 threads, the average leakage savings were 42 percent in coarse-grained multithreading, while they were between seven percent and eight percent for finegrained and simultaneous multithreading.

Compiler-directed power gating is an approach in which sleep instructions are inserted appropriately at compile time into the application code to selectively deactivate the functional units in microprocessors during their idle periods to reduce power dissipation due to leakage. Although the effect of code transformations on dynamic and system power has been investigated and reported in the literature, such a study is lacking in the context of power gating. In this paper, we investigate and report how the leakage savings in both integer and floating point units can be improved using machine-dependent and independent optimizations in a compiler-directed power gating framework. In our study, it is ensured that power gating is applied only when the leakage savings are considerably more than the various overheads incurred in its implementation. The target embedded processor is modeled on the ARMv4 architecture, which is modified to support the power gating of its arithmetic functional units. For experimentation, GCC is used as the compiler infrastructure and Simplescalar-ARM is used as the detailed architectural simulator for reporting power and performance metrics for embedded applications belonging to the MiBench and MediaBench benchmark suites. Experimental results suggest that the additional savings in leakage energy due to one or more of the optimizations may vary largely depending on the benchmark. Moreover, the overhead of sleep instructions can be reduced by up to 50 times by performing procedure inlining.

Power gating is a circuit level technique for reducing standby leakage in a circuit block by cutting off paths in it between the supply and the ground. A processor architecture that supports power gating of its resources may provide instructions that activate and deactivate those resources as part of the instruction set architecture level. Adequate compiler support is then required so that the power gating instructions can be inserted into the code to deactivate the resources that remain idle for long periods of time during program execution. However, the resource usage in a program depends on the code generated by the compiler. Thus, the code transformations performed by the compiler has an influence on the power gating opportunities of the processor resources. In this work, we explore target independent compiler optimizations that modify the functional unit usage in the loops of a procedure to enhance the opportunities to deactivate functional units in an embedded processor architecture. The optimizations performed on the code are sparse conditional constant propagation, lazy code motion, weak strength reduction, and operator strength reduction. Insertion of power gating instructions is performed by inspecting the idleness of the units in the regions enclosed within loops. We model the processor architecture with power gating support around an ARM core and use the SUIF framework for compiler support. Finally, we use the Simplescalar-ARM distribution to perform power and performance evaluation with a set of benchmarks from MiBench and MediaBench suites. Experimental results indicate that the integer multiplier in the processor core can be power gated for upto 99% of its idle cycles, for integer benchmarks, and upto 93%, for floating point benchmarks, when all the optimizations are performed. Moreover, the energy due to leakage in the functional units for the code with all the optimizations performed can be upto 51% lower, for integer benchmarks, and upto 21% lower, for floating point benchmarks, than that for the unoptimized code.

Power gating is a technique commonly used for leakage reduction in integrated circuits. In microprocessors, power gating is implemented by using sleep transistors to selectively deactivate circuit modules that remain idle for sustained periods of time during program execution. In this work, we develop a new framework for power gating the functional units in embedded system microprocessors without degradation in performance. The proposed framework includes an efficient algorithm for idle time estimation, appropriate insertion of sleep instructions within the code, and a method for reactivating the sleeping units only when needed without the use of wakeup instructions. We introduce the notion of loop hierarchy trees (LHTs) to represent the partial ordering of the nested loops within the program. From the control flow graph (CFG) representation of the source program, a forest of LHTs is constructed and is used to identify the maximal subgraphs representing the long idle periods for the functional units. For each subgraph thus identified, a sleep instruction is introduced in the program with a list of corresponding functional units to be deactivated. When an instruction is decoded, the functional units needed for that instruction are automatically activated by the control unit such that the units are ready before the instruction reaches the execute stage. This eliminates the need for wakeup instructions to be inserted into the object code reducing the overheads. In our implementation, the ARM processor architecture was modified and resynthesized to include power gating by developing a CMOS cell library of functional units with the above capabilities. Experimental results are reported for a set of 12 benchmarks chosen from the MiBench suite, which indicate that, on average, our technique reduces the leakage energy in functional units by 31.1% for integer benchmarks and 26.8% for floating-point benchmarks.