The Microarchitecture of a Low Power Register File

The access time, energy and area of the register file are often critical to overall performance in wide-issue microprocessors, because these terms grow superlinearly with the number of read and write ports that are required to support wide-issue. This paper presents two techniques to reduce the number of ports of a register file intended for a wide-issue microprocessor without hardly any impact on IPC. Our results show that it is possible to replace a register file with 16 read and 8 write ports, intended for an eight-issue processor, with a register file with just 8 read and 8 write ports so that the impact on IPC is a few percent. This is accomplished with the addition of several small auxiliary memory structures - a "delayed write-back queue" and a "operand prefetch buffer." We examine several configurations employing these structures separately and in combination. In the case of just the delayed write-back queue, we show an energy per access savings of about 40% and an area savings of 40%. This incurs a performance loss of just 4%. The area savings in turn has the potential for further savings by shortening global interconnect in the layout. We also show that the performance loss can be almost eliminated if both techniques are used in combination, although some area and power savings is lost.