Hi, OK, my initial impression was that migration to offline state involved many un-necessary operations, such as full de-allocation of all data-structures related to this processor, but after reading your description it seems that it indeed does only what's necessary to achieve the state transition.
I want to confirm that this function only does the minimum required in order to transition to the offline state. If it already does this then I'm happy, because I don't need to do any changes. In my project I'm envisioning a number of power states where each of them are associated with a performance "cost" in terms of cycles and power. My hope is that i can model the "shallowest" state (that only uses simple clock gating) using a idle thread with a HLT instruction or something similar. So, using simple clock gating would not involve de-allocation of the processor from the active list. If I would be forced to use the offline function for simple clock-gating, then the performance-cost would be huge, which is usually not the case in a real implementation. However, the "deeper" power states that utilize techniques involving actual shutdown of processor components would in my mind need to use the offline function to redirect interrupts etc...this is how I intend to use the different mechanisms within the kernel, feel free to comment or ask questions if something is unclear. Regards, Mladen This message posted from opensolaris.org _______________________________________________ opensolaris-code mailing list opensolaris-code@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
