During some performance-oriented benchmarking on a Cortex A9 platform, a slight performance degradation was noted on datasets that spanned into the L2 cache (<10%). This performance hit was minor, but it prompted investigation into the cause. One difference in the actual PL310 configuration that was concerning was the enabling of two PM-related changes: Dynamic Clock Gating and Standby Mode Enabling. As the kernel being tested was patched and configured to use the PREEMPT_RT patchset, it was desired to disable these settings for our use-case since anything PM can (and usually does) impact determinism. Making these changes resulted in a modest performance improvement and those wonderful warm-n-fuzzies regarding determinism and enabling system control without needing to change the kernel. In the following set, there's the actual change to control these features given DT presence of a couple of new bindings and the documenation to accompany those changes. Thanks! -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
