System-wide low power management
Systems, apparatuses, and methods for performing efficient power management for a multi-node computing system are disclosed. A computing system includes multiple nodes. When power down negotiation is distributed, negotiation for system-wide power down occurs within a lower level of a node hierarchy prior to negotiation for power down occurring at a higher level of the node hierarchy. When power down negotiation is centralized, a given node combines a state of its clients with indications received on its downstream link and sends an indication on an upstream link based on the combining. Only a root node sends power down requests.
1. A computing system comprising:
a plurality of nodes including at least a first node and a second node, each comprising:
one or more clients configured to process applications;
a plurality of links coupled to one or more other nodes of the plurality of nodes, wherein two or more of the plurality of nodes are not fully connected; and
a power controller; and
wherein the second node is configured to send an indication to the first node that each client in a first subset of nodes that includes at least two nodes of the plurality of nodes is idle;
wherein each node of the plurality of nodes is configured to power down the plurality of links and the one or more clients, in response to the first node of the plurality of nodes determining each client in the plurality of nodes is idle.
2. The computing system as recited in claim 1 , wherein a third node of the plurality of nodes is configured to send an indication to the second node of the plurality of nodes specifying that each client in a second subset of the plurality of nodes is idle, wherein the third node is not directly connected to the first node.
3. The computing system as recited in claim 1 , wherein the second node is configured to send said indication, in response to:
determining each client of the one or more clients in the second node is idle; and
receipt of an indication from a third node that each client in a second subset of the plurality of nodes is idle.
4. The computing system as recited in claim 2 , wherein in response to receiving an indication that each client in a given subset of the plurality of nodes is idle on each link of the first node, the first node is configured to send a request to power down to each directly connected neighboring node.
5. The computing system as recited in claim 1 , wherein the first subset of the plurality of nodes includes the second node and at least one additional node other than the first node.
6. The computing system as recited in claim 1 , wherein a power controller of the second node is configured to power down each link and each client of the second node responsive to transferring on each link of the second node a power down request and a power down response.
7. The computing system as recited in claim 1 , wherein the plurality of nodes are not fully connected.
8. The computing system as recited in claim 1 , wherein the second node is configured to send an indication that each client in the second node is idle, based at least in part on receipt of an indication that clients in a third node of the plurality of nodes are idle.
9. The computing system as recited in claim 2 , wherein,
the plurality of nodes are connected in a tree hierarchy, wherein messages travel on links in an upstream direction from a leaf node to a root node and messages travel on links in a downstream direction from the root node to the leaf node; and
the first node is a root node of the tree hierarchy.
10. The computing system as recited in claim 9 , wherein a power controller of the second node is configured to send on a downstream link a request to power down the plurality of nodes, in response to:
determining each client in the second node is idle; and
receiving on an upstream link an indication to power down the plurality of nodes.
11. The computing system as recited in claim 1 , wherein the second node is configured to send an indication to the first node specifying that at least one client in the plurality of nodes is non-idle, in response to:
receiving a logical disconnect link indication from the first node.
12. The computing system as recited in claim 2 , wherein the first node is a root node, the second node is an intermediate node, and the third node is a leaf node.
13. A method, comprising:
processing applications by a plurality of nodes including a first node and a second node, each node comprising:
one or more clients configured to process the applications;
a plurality of links coupled to one or more other nodes of the plurality of nodes, wherein two or more of the plurality of nodes are not fully connected; and
a power controller; and
sending, by the second node, an indication to the first node that each client in a first subset of nodes that includes at least two nodes of the plurality of nodes is idle;
powering down, by each node of the plurality of nodes, the plurality of links and the one or more clients, in response to the first node of the plurality of nodes determining each client in the plurality of nodes is idle.
14. The method as recited in claim 13 , further comprising sending, by a third node of the plurality of nodes, an indication to second node of the plurality of nodes specifying that each client in a second subset of the plurality of nodes is idle, wherein the third node is not directly connected to the first node.
15. The method as recited in claim 13 , further comprising the second node sending said indication in response to
determining each client of the one or more clients in the second node is idle; and
receiving of an indication from a third node that each client in a second subset of the plurality of nodes is idle.
16. The method as recited in claim 14 , wherein in response to receiving an indication that each client in a given subset of the plurality of nodes is idle on each link of the first node, the method further comprises sending, by the first node, a request to power down to each directly connected neighboring node.
17. A processing node comprising:
one or more clients configured to process applications;
an upstream link;
a downstream link; and
a power controller;
wherein the power controller is configured to send an indication via the upstream link specifying that each client in a subset of processing nodes including at least two processing nodes is idle, in response to:
a determination that the one or more clients are idle; and
receipt of an indication via the downstream link that each client in a node coupled to the downstream link is idle; and
wherein subsequent to sending the indication via the upstream link, the power controller is further configured to power down the upstream link, the downstream link, and the one or more clients, in response to receipt of an indication via the upstream link from another node.
18. The processing node as recited in claim 17 , wherein the subset of processing nodes includes the processing node and at least the node coupled to the downstream link.
19. The processing node as recited in claim 18 , wherein the upstream link is configured to be coupled to a root node and the downstream link is configured to be coupled to either a leaf node or an intermediate node.
20. The processing node as recited in claim 17 , wherein the power controller is configured to power down each link and each client responsive to transferring on each link a power down request and a power down response.