Laying the Groundwork for Heterogeneous Processing
AMD developer Vishal Badole has introduced two patches for the Linux kernel designed to implement Capacity Aware Scheduling (CAS). This development is crucial for optimizing how operating systems manage processors that feature varying core architectures.
Until this point, the amd-pstate driver for Linux functioned under the assumption that all processor cores were identical, distributing workloads without accounting for differences in power or performance. While current chips utilize a mix of standard Zen 5 and compact Zen 5C cores, the performance disparity between them is relatively minor compared to architectures found elsewhere in the industry.
The Shift to Low-Power Cores in Zen 6
The strategic move toward CAS appears to be driven by future hardware roadmaps. Reports suggest that the upcoming Zen 6 architecture will introduce dedicated Low-Power (LP) cores. Unlike previous AMD designs that utilized efficiency cores which were essentially miniaturized versions of their performance counterparts, these future LP cores are expected to operate with significantly different power profiles.
These anticipated LP cores are likely to mirror the Low-Power Efficiency (LPE) cores currently seen in architectures like Intel's Panther Lake. By implementing Capacity Aware Scheduling, the system will gain the ability to intelligently distribute tasks based on the specific power consumption and performance capabilities of each individual core type.
Benefits for Existing and Future Systems
While the feature is essential for the upcoming Zen 6 generation, it also offers immediate improvements for existing hardware. Regarding current heterogeneous setups, Badole noted:
«The scheduler sees every CPU as having identical capacity and treats performance and efficiency cores as interchangeable. Tasks are therefore placed and balanced without regard to core strength, which hurts both throughput and energy efficiency on hybrid systems.»
By moving away from treating every core as interchangeable, AMD aims to maximize both energy efficiency and overall throughput. For future processors with very low-power cores, this level of scheduling precision will be a requirement for optimal operation rather than just a performance enhancement.
