Processors having virtually clustered cores and cache slices
A processor of an aspect includes a plurality of logical processors each having one or more corresponding lower level caches. A shared higher level cache is shared by the plurality of logical processors. The shared higher level cache includes a distributed cache slice for each of the logical processors. The processor includes logic to direct an access that misses in one or more lower level caches of a corresponding logical processor to a subset of the distributed cache slices in a virtual cluster that corresponds to the logical processor. Other processors, methods, and systems are also disclosed.
1. A processor comprising:
a plurality of cores, the plurality of cores comprising symmetric multi-threaded cores;
a cache subsystem, the cache subsystem comprising a plurality of first-level caches and at least one higher-level distributed cache comprising a plurality of distributed cache portions that are physically distributed across a die and shared by the plurality of cores, each first-level cache integral to one of the plurality of cores and each distributed cache portion accessible to each of the plurality of cores;
cache management circuitry operative to provide coherent, non-uniform access to the plurality of distributed cache portions by the plurality of cores; and
power management circuitry operative to enable a first frequency of operation for a first cluster of the plurality of cores which are physically proximate to one another and a second frequency of operation for a second cluster of the plurality of cores which are physically proximate to one another, wherein an average distance between cores in the first cluster is less than an average distance between all of the cores, the power management circuitry operative to selectively gate power to the first cluster of the plurality of cores and distributed cache portions of the at least one higher-level distributed cache that correspond to the first cluster and/or the second cluster of the plurality of cores and distributed cache portions of the at least one higher-level distributed cache that correspond to the second cluster.
2. The processor of claim 1 wherein each distributed cache portion is physically closer to at least one cluster of the plurality of cores than it is to another cluster of the plurality of cores.
3. The processor of claim 1 wherein the power management circuitry is operative to selectively gate power in accordance with a detected workload on the first and/or second clusters of the plurality of cores.
4. The processor of claim 1 wherein, based on a workload on each of the first and second clusters of the plurality of cores, the processor is operative to selectively migrate a thread from the first cluster to the second cluster.
5. The processor of claim 1 wherein the non-uniform access to the plurality of distributed cache portions by the plurality of cores is based, at least in part, on each distributed cache portion being coupled to a proximate cluster of the plurality of cores over a first interconnect having first characteristics and being coupled to a non-proximate cluster of the plurality of cores over a second interconnect having second characteristics.
6. The processor of claim 1 , wherein, based on workload on each of the first and second clusters of the plurality of cores, the system is operative to selectively migrate a thread from the first cluster to the second cluster, and further comprising a Peripheral Component Interface Express (PCIe) interface coupled with the plurality of cores.
7. The processor of claim 1 , wherein each distributed cache portion is a cache slice and is physically co-located with one of the cores, wherein, based on workload on each of the first and second clusters of the plurality of cores, the system is operative to selectively migrate a thread from the first cluster to the second cluster, wherein the first cluster of the plurality of cores form a first virtual processor and the second cluster of the plurality of cores form a second virtual processor, and further comprising a Peripheral Component Interface Express (PCIe) interface coupled with the plurality of cores.
8. The processor of claim 1 , wherein the first cluster of the plurality of cores form a first virtual processor and the second cluster of the plurality of cores form a second virtual processor, and further comprising a Peripheral Component Interface Express (PCIe) interface coupled with the plurality of cores.
9. The processor of claim 1 , wherein each distributed cache portion is a cache slice and is physically co-located with one of the cores, wherein, based on workload on each of the first and second clusters of the plurality of cores, the system is operative to selectively migrate a thread from the first cluster to the second cluster, and wherein the first cluster of the plurality of cores form a first virtual processor and the second cluster of the plurality of cores form a second virtual processor.
10. The processor of claim 1 wherein the first cluster of the plurality of cores form a first virtual processor and the second cluster of the plurality of cores form a second virtual processor.
11. The processor of claim 10 further comprising a storage storing a virtual machine monitor (VMM) that when executed by the processor is to form the first virtual processor through combination of the first cluster of the plurality of cores and to form the second virtual processor through combination of the second cluster of the plurality of cores.
12. The processor of claim 1 wherein a first core of the first cluster is physically proximate to a first distributed cache portion, and wherein the first cluster is a predetermined cluster.
13. The processor of claim 12 further comprising:
a local core-cache interconnect coupling the first core to the first distributed cache portion; and
an on-die interconnect coupling one or more other cores to the first distributed cache portion.
14. The processor of claim 13 wherein the local core-cache interconnect and the on-die interconnect provide different access latency and/or data throughput to the first distributed cache portion.
15. The processor of claim 13 wherein the local core-cache interconnect is operative to provide relatively lower latency and/or higher data throughput to the first core when the first core accesses the first distributed cache portion than the on-die interconnect is operative to provide to the one or more other cores when the one or more other cores access the first distributed cache portion.
16. A method comprising:
executing instructions and processing data with a plurality of cores, the plurality of cores comprising symmetric multi-threaded cores;
storing the instructions and the data in a cache subsystem, the cache subsystem comprising a plurality of first-level caches and at least one higher-level distributed cache comprising a plurality of distributed cache portions that are physically distributed across a die, each first-level cache integral to one of the plurality of cores and each distributed cache portion accessible to each of the plurality of cores;
sharing the plurality of distributed cache portions among the plurality of cores;
providing coherent, non-uniform access to the plurality of distributed cache portions by the plurality of cores;
enabling a first frequency to be set for a first cluster of the plurality of cores which are physically proximate to one another and a second frequency to be set for a second cluster of the plurality of cores which are physically proximate to one another, wherein an average distance between cores in the first cluster is less than an average distance between all of the cores; and
selectively gating power to the first cluster of the plurality of cores and distributed cache portions of the at least one higher-level distributed cache that correspond to the first cluster and/or the second cluster of the plurality of cores and distributed cache portions of the at least one higher-level distributed cache that correspond to the second cluster.
17. The method of claim 16 wherein said selectively gating the power comprises selectively gating the power in accordance with a detected workload on the first and/or second clusters of the plurality of cores.
18. The method of claim 16 further comprising selectively migrated a thread from the first cluster to the second cluster based on a workload on each of the first and second clusters of the plurality of cores.
19. The method of claim 16 further comprising forming the first cluster of the plurality of cores as a first virtual processor and forming the second cluster of the plurality of cores as a second virtual processor.
20. The method of claim 19 further comprising executing instructions to implement a virtual machine monitor (VMM), the VMM to form the first virtual processor by combining the first cluster of the plurality of cores and to form the second virtual processor by combining the second cluster of the plurality of cores.
21. A machine comprising:
a processor comprising:
a plurality of cores, the plurality of cores comprising symmetric multi-threaded cores;
a cache subsystem, the cache subsystem comprising a plurality of first-level caches and at least one higher-level distributed cache comprising a plurality of distributed cache portions that are physically distributed across a die and shared by the plurality of cores, each first-level cache integral to one of the plurality of cores and each distributed cache portion accessible to each of the plurality of cores; and
cache management circuitry configured to provide coherent, non-uniform access to the plurality of distributed cache portions by the plurality of cores; and
an article of manufacture comprising a machine-readable storage medium storing a plurality of instructions, the plurality of instructions, when performed by the machine, cause the machine to perform operations comprising:
enable a first frequency of operation for a first cluster of the plurality of cores which are physically proximate to one another and a second frequency of operation for a second cluster of the plurality of cores which are physically proximate to one another, wherein an average distance between cores in the first cluster is less than an average distance between all of the cores; and
selectively gate power to the first cluster of the plurality of cores and distributed cache portions of the at least one higher-level distributed cache that correspond to the first cluster and/or the second cluster of the plurality of cores and distributed cache portions of the at least one higher-level distributed cache that correspond to the second cluster.
22. The machine of claim 21 wherein the instructions further comprise instructions that when performed by the machine cause the machine to form the first cluster of the plurality of cores as a first virtual processor and form the second cluster of the plurality of cores as a second virtual processor.