On Tue, 2012-11-13 at 19:58 +0800, Pavel Machek wrote: > On Tue 2012-11-13 10:04:48, Joseph Lo wrote: > > The L2 RAM is in different power domain from the CPU cluster. So the > > L2 content can be retained over CPU suspend/resume. To do that, we > > need to disable L2 after the MMU is disabled, and enable L2 before > > the MMU is enabled. But the L2 controller is in the same power domain > > with the CPU cluster. We need to restore it's settings and re-enable > > it after the power be resumed. > > Is it good idea to retain L2 over suspend? I guess this is > power-vs-speed balance...? > Hi Pavel, Thanks for your review. Because one of the Tegra's CPU idle state will shut off the CPU power, it only contain CPU cluster, L1$, GIC and L2 controller. The L2 ram is still powered. The L2 flush is a slow operation at low frequency. The CPU idle "power-down" state is likely to happen often at low frequency for low power use case (e.g. low power audio playback). So yes, it's power-vs-speed balance. For the system-wide sleep mode (i.e. suspend), the power of L2 RAM would be shut off at the deepest sleep mode. There would be different procedure to handle L2. Thanks, Joseph -- To unsubscribe from this list: send the line "unsubscribe linux-tegra" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
