System and method for a snooping and snarfing cache in a multiprocessor computer system

• US 5,588,131 A
• Filed: 03/09/1994
• Issued: 12/24/1996
• Est. Priority Date: 03/09/1994
• Status: Expired due to Term

• Current Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Original Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Inventors: Borrill, Paul L.
• Primary Examiner(s): Asta, Frank J.
• Application Number:
  US08/208,170
• Time in Patent Office: 1,021 Days
• Field of Search: 395/446, 395/468, 395/470, 395/473
• US Class Current: 711/146
• CPC Class Codes: G06F 12/0831 using a bus scheme, e.g. with bus monitoring or watching means

Abstract

An improved multiprocessor computer system with an improved snarfing cache is disclosed. The multiprocessor system includes a main memory, I/O interface, and a plurality of processor nodes. Each processor node includes a CPU, and a cache. A shared interconnect couples the main memory, I/O interface, and the plurality of processor nodes. The snarfing cache of each processor node snarfs valid data that appears on the shared interconnect, regardless of whether the cache of the processor node has an invalid copy or no copy of the data. The net effect is that each processor node locally caches additional valid data, resulting in an expected improved cache hit rate, reduced processor latency, and fewer transactions on the shared interconnect.

Claims

1. An apparatus configured to be inserted into a multiprocessor system having a plurality of data elements and a plurality of processor nodes coupled together by a shared interconnect, the shared interconnect configured to carry a plurality of data elements, the apparatus comprising:

a processor node;

a cache configured to cache a group of the plurality of data elements for the processor node; and

a snarfing mechanism, coupled to the cache, the snarfing mechanism, when inserted into the multiprocessor system, configured to check the cache to determine if the cache contains a copy of a data element appearing on the shared interconnect, and further programmed to snarf the data element off the shared interconnect if the cache has no copy of the data element.

2. The apparatus of claim 1, wherein the snarfing mechanism is further configured to snarf the data element off the shared interconnect in the event the cache has an invalid copy of the data element.

3. The apparatus of claim 1, wherein the snarfing mechanism is further configured to not snarf the data element off the shared interconnect in the event the cache has a valid copy of the data element.

4. The apparatus of claim 1, wherein the snarfing mechanism further comprises a program element configured to store program information to program the snarfing mechanism to selectively snarf selected data elements among the plurality of data elements appearing on the shared interconnect in accordance with a selection criteria.

5. The apparatus of claim 4, wherein the selection criteria directs the snarfing mechanism to snarf all the data elements updated and placed on the shared interconnect by a second processor node in the multiprocessor system.

6. The apparatus of claim 4, wherein the selection criteria directs the snarfing mechanism to snarf the data element addressed by a specified memory location.

7. The apparatus of claim 4, wherein the selection criteria directs the snarfing mechanism to snarf a first subset of the plurality data elements generated by processing a first thread in the multiprocessor system and to not scarf a second subset of the plurality data elements generated by processing a second thread in the multiprocessor system.

8. The apparatus of claim 4, wherein the selection criteria directs the snarfing mechanism to snarf the data elements derived from a specified source.

9. The apparatus of claim 4, wherein the selection criteria directs the mechanism to snarf a first subset of the plurality data elements having a first label and to not scarf a second subset of the plurality data elements having a second label.

10. The apparatus of claim 4, wherein the selection criteria directs the snarfing mechanism to snarf the data elements located at a specified memory location.

11. A multiprocessor system having a plurality of data elements and a plurality of processor nodes coupled together by a shared interconnect, the shared interconnect configured to carry a plurality of data elements, each of the processor nodes comprising:

a cache configured to cache a group of the plurality of data elements for the processor node; and

a snarfing mechanism, coupled to the cache, the snarfing mechanism, when inserted into the multiprocessor system, configured to check the cache to determine if the cache contains a copy of a data element appearing on the shared interconnect, and further programmed to snarf the data element off the shared interconnect if the cache has no copy of the data element.

12. The apparatus of claim 11, wherein the snarfing mechanism is further configured to snarf the data element off the shared interconnect in the event the cache has an invalid copy of the data element.

13. The apparatus of claim 11, wherein the ,snarfing mechanism is further configured to not snarf the data element off the shared interconnect in the event the cache has a valid copy of the data element.

14. The apparatus of claim 11, wherein the snarfing mechanism further comprises a program element configured to store program information to program the snarfing mechanism to selectively snarf selected data elements among the plurality of data elements appearing on the shared interconnect in accordance with a selection criteria.

15. The apparatus of claim 14, wherein the selection criteria directs the snarfing mechanism to snarf all the data elements updated and placed on the shared interconnect by a second processor node in the multiprocessor system.

16. The apparatus of claim 14, wherein the selection criteria directs the snarfing mechanism to snarf the data element addressed by a specified memory location.

17. The apparatus of claim 14, wherein the selection criteria directs the snarfing mechanism to snarf a first subset of the plurality data elements generated by processing a first thread in the multiprocessor system and to not scarf a second subset of the plurality data elements generated by processing a second thread in the multiprocessor system.

18. The apparatus of claim 14, wherein the selection criteria directs the snarfing mechanism to snarf the data elements derived from a specified source.

19. The apparatus of claim 14, wherein the selection criteria directs the mechanism to snarf a first subset of the plurality data elements having a first label and to not scarf a second subset of the plurality data elements having a second label.

20. A data caching method for a processor node and cache in a multiprocessor system having a plurality of processor nodes coupled together by a shared interconnect, the shared interconnect configured to carry a plurality of data elements, the method comprising the steps of:

providing one of the plurality of data elements over the shared interconnect;

determining if the cache contains a copy of the data element appearing on the shared interconnect;

snarfing the data element off the shared interconnect if the cache has no copy of the data element; and

storing the data element snarfed from the shared interconnect in the cache.

21. The method of claim 20, wherein the snarfing step further occurs off the shared interconnect in the event the cache has an invalid copy of the data element.

22. The method of claim 20, wherein the snarfing step does not snarf the data element off the shared interconnect in the event the cache has a valid copy of the data element.

23. The method of claim 20, wherein the snarfing step further snarls selected data elements among the plurality of data elements appearing on the shared interconnect in accordance with a selection criteria.

24. The method of claim 23, wherein the selection criteria causes snarfing of all the data elements updated and placed on the shared interconnect by a second processor node in the multiprocessor system.

25. The method of claim 23, wherein the selection criteria causes the snarfing step to snarl the data element addressed by a specified memory location.

26. The method of claim 23, wherein the selection criteria causes snarfing to occur from a first subset of the plurality data elements generated by processing a first thread in the multiprocessor system and no snarfing to occur from a second subset of the plurality data elements generated by processing a second thread in the multiprocessor system.

27. The method of claim 23, wherein the selection criteria causes snarfing of the data elements derived from a specified source.

28. The method of claim 23, wherein the selection criteria causes the snarfing of a first subset of the plurality data elements having a first label and no snarfing of a second subset of the plurality data elements having a second label.