Heterogeneous core microarchitecture
Embodiments relate to a heterogeneous core microarchitecture. An aspect includes binding, by an operating system that is executing on a processor comprising a core comprising a heterogeneous microarchitecture comprising two or more flows, a job that is being executed by the operating system to a flow of the two or more flows. Another aspect includes issuing an instruction corresponding to the job with a tag indicating the binding of the job to which the instruction corresponds. Yet another aspect includes executing the instruction by the flow in the core that is indicated by the tag.
1. A computer implemented method for implementing a heterogeneous microarchitecture, the method comprising:
binding, by an operating system that is executing on a homogeneous processor comprising a plurality of cores, each core comprising a heterogeneous microarchitecture comprising two or more flows, a job that is being executed by the operating system to a flow of the two or more flows, wherein the operating system is running on a virtual machine of the homogeneous processor;
issuing an instruction corresponding to the job with a tag indicating the binding of the job to which the instruction corresponds;
determining the tag associated with the instruction;
responsive to the determining, dispatching the instruction to a first hardware or a second hardware according to the tag, wherein the first hardware comprises a dedicated L1 instruction cache coupled to two 3-way flow dispatching units, the two 3-way dispatching units being coupled to a plurality of execution units and an L1 data cache, and the second hardware comprises a 64 kilobyte local instruction cache which loads instructions from an L2 cache, at least one instruction pipeline, and an L1 data cache;
wherein at least one L2 instruction cache is shared between the first hardware and the second hardware;
wherein the L1 data cache of the second hardware is coupled to a first L2 data cache, and the L1 data cache of the first hardware is coupled to a second L2 data cache; and
executing the instruction by the flow in the core that is indicated by the tag.
2. The method of claim 1 , wherein the two or more flows comprise the same hardware operated in different modes corresponding to each flow.
3. The method of claim 1 , wherein the two or more flows comprise distinct hardware.
4. The method of claim 3 , wherein each core in the homogeneous processor comprises the same microarchitecture.
5. The method of claim 3 , wherein each flow comprises a respective cache, and wherein coherency is maintained in the caches in the core during operation.