Abstract: In one embodiment, a processor includes an accelerator, a decoder to decode a first instruction into a decoded first instruction, and a second instruction into a decoded second instruction, and an execution unit to execute the first decoded instruction to, for a thread executing on the accelerator that is to be placed in an inactive state, cause a save of context information for the thread, and a save of a vector identifying the accelerator corresponding to the context information, and execute the second decoded instruction to read the vector to determine the accelerator to restore saved context information into for the thread, read the saved context information, and restore the saved context information into the accelerator.

Abstract: Embodiments of a near optimal configurable adder tree for arbitrary shaped 2D block sum of absolute differences (SAD) calculation engine are generally described herein. Other embodiments may be described and claimed. In some embodiments, a configurable two-dimensional adder tree architecture for computing a sum of absolute differences (SAD) for various block sizes up to 16 by 16 comprises a first stage of one-dimensional adder trees and a second stage of one-dimensional adder trees, wherein each one-dimensional adder tree comprises an input routing network, a plurality of adder units, and an output routing network.

Abstract: In one embodiment, a processor includes an accelerator, a decoder to decode a first instruction into a decoded first instruction, and a second instruction into a decoded second instruction, and an execution unit to execute the first decoded instruction to, for a thread executing on the accelerator that is to be placed in an inactive state, cause a save of context information for the thread, and a save of a vector identifying the accelerator corresponding to the context information, and execute the second decoded instruction to read the vector to determine the accelerator to restore saved context information into for the thread, read the saved context information, and restore the saved context information into the accelerator.

Abstract: An apparatus is described having multiple cores, each core having: a) an accelerator; and, b) a general purpose CPU coupled to the accelerator. The general purpose CPU has functional unit logic circuitry to execute an instruction that returns an amount of storage space to store context information of the accelerator.

Abstract: Some implementations disclosed herein provide techniques and arrangements for an specialized logic engine that includes translation lookaside buffer to support multiple threads executing on multiple cores. The translation lookaside buffer enables the specialized logic engine to directly access a virtual address of a thread executing on one of the plurality of processing cores. For example, an acceleration compute engine may receive one or more instructions from a thread executed by a processing core. The acceleration compute engine may retrieve, based on an address space identifier associated with the one or more instructions, a physical address associated with the one or more instructions from the translation lookaside buffer to execute the one or more instructions using the physical address.

Abstract: Some implementations disclosed herein provide techniques and arrangements for an specialized logic engine that includes translation lookaside buffer to support multiple threads executing on multiple cores. The translation lookaside buffer enables the specialized logic engine to directly access a virtual address of a thread executing on one of the plurality of processing cores. For example, an acceleration compute engine may receive one or more instructions from a thread executed by a processing core. The acceleration compute engine may retrieve, based on an address space identifier associated with the one or more instructions, a physical address associated with the one or more instructions from the translation lookaside buffer to execute the one or more instructions using the physical address.

Abstract: Apparatus, systems and methods are described including dividing cache lines into at least most significant portions and next most significant portions, storing cache line contents in a register array so that the most significant portion of each cache line is stored in a first row of the register array and the next most significant portion of each cache line is stored in a second row of the register array. Contents of a first register portion of the first row may be provided to a barrel shifter where the contents may be aligned and then stored in a buffer.

Abstract: Embodiments of a near optimal configurable adder tree for arbitrary shaped 2D block sum of absolute differences (SAD) calculation engine are generally described herein. Other embodiments may be described and claimed. In some embodiments, a configurable two-dimensional adder tree architecture for computing a sum of absolute differences (SAD) for various block sizes up to 16 by 16 comprises a first stage of one-dimensional adder trees and a second stage of one-dimensional adder trees, wherein each one-dimensional adder tree comprises an input routing network, a plurality of adder units, and an output routing network.