Dynamically setting routing information to transfer input output data directly into processor caches in a multi processor system
View Patent ↗According to some embodiments, IO traffic is transferred directly into a target processor cache in accordance with routing information.
1. A write agent, comprising:
a write agent input to receive, from a requesting agent processor, routing information associated with Input Output (IO) traffic to be received at a target processor cache of a target processor, wherein the write agent, the requesting agent processor, and the target processor are different devices and further wherein the write agent comprises at least one of a direct memory access controller or an IO controller hub;
a write agent transfer portion to arrange for information packets of the IO traffic to be transferred directly into the target processor cache in accordance with the routing information; and
a write agent output to transfer status information associated with the IO traffic directly into a requesting agent cache of the requesting agent processor.
2. The write agent of claim 1 , wherein the routing information includes at least one of: (i) a memory address, (ii) a target processor identifier, (iii) a direct transfer on/off indication, (iv) a cache allocation/update indication, (v) a routing policy, (vi) a routing condition, (vii) a routing preference, (viii) coherence information, or (ix) an allocation policy.
3. The write agent of claim 1 , wherein the IO traffic is associated with at least one of: (i) a network, (ii) a network interface controller, (iii) a disk drive controller, (iv) a peripheral component interconnect interface, (v) a universal serial bus interface, or (vi) a 1394 interface.
4. The write agent of claim 1 , wherein the routing information indicates that one type of IO traffic should be transferred directly into the target processor cache while another type of IO traffic should be transferred directly into another target processor cache.
5. The write agent of claim 1 , wherein the IO traffic is received from at least one of: (i) a network fabric, and (ii) a disk drive, a (iii) a graphics device, or (iv) a peripheral device.
6. The write agent of claim 1 , wherein the IO traffic is transferred into the target processor cache in accordance with a chipset's platform routing function.
7. A method, comprising:
determining at a requesting agent processor that Input Output (IO) traffic is to be received at a target processor cache of a target processor, wherein the target processor is different than the requesting agent processor;
transmitting, to a write agent, routing information associated with the IO traffic, wherein the write agent is different than the requesting agent processor and the target processor and comprises at least one of: (i) a direct memory access controller, or (ii) an IO controller hub; and
receiving, from the write agent, status information associated with the IO traffic, the status information being directly received into a requesting agent cache of the requesting agent processor.
8. The method of claim 7 , wherein said transmitting is performed by an IO driver executing at the requesting agent processor.
9. The method of claim 7 , wherein the routing information includes at least one of: (i) a memory address, (ii) a target processor identifier, (iii) a direct transfer on/off indication, (iv) a cache allocation/update indication, (v) a routing policy, (vi) a routing condition, (vii) a routing preference, (viii) coherence information, or (ix) an allocation policy.
10. The method of claim 7 , wherein the IO traffic is associated with at least one of: (i) a network, (ii) a network interface controller, (iii) a disk drive controller, (iv) a peripheral component interconnect interface, (v) a universal serial bus interface, or (vi) a 1394 interface.
11. The method of claim 7 , wherein the arranging is performed in a multi-processor system that includes a plurality of potential target processor caches.
12. The method of claim 7 , wherein the routing information indicates that one type of IO traffic should be transferred directly into the target processor cache while another type of IO traffic should be transferred directly into another target processor cache.
13. A requesting agent, comprising:
a requesting agent processor to determine that Input Output (IO) traffic is to be received at a target processor cache of a target processor, wherein the target processor is different than the requesting agent;
a requesting agent output to transmit, to a write agent, routing information associated with the IO traffic, wherein the write agent is different than the requesting agent and the target processor; and
a requesting agent input to receive, from the write agent, status information associated with the IO traffic, the status information being directly received into a requesting agent cache of the requesting agent.
14. The requesting agent of claim 13 , wherein the routing information is transmitted from an IO driver executing at the requesting agent processor.
15. The requesting agent of claim 13 , wherein the IO traffic is associated with at least one of: (i) a network, (ii) a network interface controller, (iii) a disk drive controller, (iv) a peripheral component interconnect interface, (v) a universal serial bus interface, or (vi) a 1394 interface.
16. The requesting agent of claim 13 , wherein the routing information includes at least one of: (i) a memory address, (ii) a target processor identifier, (iii) a direct transfer on/off indication, (iv) a cache allocation/update indication, (v) a routing policy, (vi) a routing condition, (vii) a routing preference, (viii) coherence information, or (ix) an allocation policy.
17. The requesting agent of claim 13 , wherein the routing information indicates that one type of IO traffic should be transferred directly into the target processor cache while another type of IO traffic should be transferred directly into another target processor cache.