Abstract: A sparse convolutional neural network accelerator system that dynamically and efficiently identifies fine-grained parallelism in sparse convolution operations. The system determines matching pairs of non-zero input activations and weights from the compacted input activation and weight arrays utilizing a scalable, dynamic parallelism discovery unit (PDU) that performs a parallel search on the input activation array and the weight array to identify reducible input activation and weight pairs.

Abstract: A distributed deep neural net (DNN) utilizing a distributed, tile-based architecture includes multiple chips, each with a central processing element, a global memory buffer, and a plurality of additional processing elements. Each additional processing element includes a weight buffer, an activation buffer, and vector multiply-accumulate units to combine, in parallel, the weight values and the activation values using stationary data flows.

Abstract: A distributed deep neural net (DNN) utilizing a distributed, tile-based architecture includes multiple chips, each with a central processing element, a global memory buffer, and a plurality of additional processing elements. Each additional processing element includes a weight buffer, an activation buffer, and vector multiply-accumulate units to combine, in parallel, the weight values and the activation values using stationary data flows.

Abstract: A distributed deep neural net (DNN) utilizing a distributed, tile-based architecture includes multiple chips, each with a central processing element, a global memory buffer, and a plurality of additional processing elements. Each additional processing element includes a weight buffer, an activation buffer, and vector multiply-accumulate units to combine, in parallel, the weight values and the activation values using stationary data flows.

Abstract: A sparse convolutional neural network accelerator system that dynamically and efficiently identifies fine-grained parallelism in sparse convolution operations. The system determines matching pairs of non-zero input activations and weights from the compacted input activation and weight arrays utilizing a scalable, dynamic parallelism discovery unit (PDU) that performs a parallel search on the input activation array and the weight array to identify reducible input activation and weight pairs.

Abstract: Methods and apparatuses to control cache line coherence are described. A hardware processor may include a first processor core with a cache to store a cache line, a second set of processor cores that each include a cache to store a copy of the cache line, and cache coherence logic to aggregate in a tag directory an acknowledgment message from each of the second set of processor cores in response to a request from the first processor core to modify the copy of the cache line in each of the second set of processor cores and send a consolidated acknowledgment message to the first processor core.

Abstract: Methods and apparatuses to control cache line coherence are described. A hardware processor may include a first processor core with a cache to store a cache line, a second set of processor cores that each include a cache to store a copy of the cache line, and cache coherence logic to aggregate in a tag directory an acknowledgment message from each of the second set of processor cores in response to a request from the first processor core to modify the copy of the cache line in each of the second set of processor cores and send a consolidated acknowledgment message to the first processor core.

Abstract: A processor is described having an out-of-order core to execute a first thread and a non-out-of-order core to execute a second thread. The processor also includes statistics collection circuitry to support calculation of the following: the first thread's performance on the out-of-order core; an estimate of the first thread's performance on the non-out-of-order core; the second thread's performance on the non-out-of-order core; an estimate of the second thread's performance on the out-of-order core.

Abstract: A processor is described having an out-of-order core to execute a first thread and a non-out-of-order core to execute a second thread. The processor also includes statistics collection circuitry to support calculation of the following: the first thread's performance on the out-of-order core; an estimate of the first thread's performance on the non-out-of-order core; the second thread's performance on the non-out-of-order core; an estimate of the second thread's performance on the out-of-order core.

Abstract: A processing engine includes separate hardware components for control processing and data processing. The instruction execution order in such a processing engine may be efficiently determined in a control processing engine based on inputs received by the control processing engine. For each instruction of a data processing engine: a status of the instruction may be set to “ready” based on a trigger for the instruction and the input received in the control processing engine; and execution of the instruction in the data processing engine may be enabled if the status of the instruction is set to “ready” and at least one processing element of the data processing engine is available. The trigger for each instruction may be a function of one or more predicate register of the control processing engine, FIFO status signals or information regarding tags.

Abstract: A technique to enable resource allocation optimization within a computer system. In one embodiment, a gradient partition algorithm (GPA) module is used to continually measure performance and adjust allocation to shared resources among a plurality of data classes in order to achieve optimal performance.

Abstract: In one embodiment, the present invention includes an apparatus having a register file to store vector data, an address generator coupled to the register file to generate addresses for a vector memory operation, and a controller to generate an output slice from one or more slices each including multiple addresses, where the output slice includes addresses each corresponding to a separately addressable portion of a memory. Other embodiments are described and claimed.

Abstract: In one embodiment, the present invention includes an apparatus having a register file to store vector data, an address generator coupled to the register file to generate addresses for a vector memory operation, and a controller to generate an output slice from one or more slices each including multiple addresses, where the output slice includes addresses each corresponding to a separately addressable portion of a memory. Other embodiments are described and claimed.

Abstract: In one embodiment, the present invention includes an apparatus having a register file to store vector data, an address generator coupled to the register file to generate addresses for a vector memory operation, and a controller to generate an output slice from one or more slices each including multiple addresses, where the output slice includes addresses each corresponding to a separately addressable portion of a memory. Other embodiments are described and claimed.

Abstract: A technique to enable resource allocation optimization within a computer system. In one embodiment, a gradient partition algorithm (GPA) module is used to continually measure performance and adjust allocation to shared resources among a plurality of data classes in order to achieve optimal performance.

Abstract: In one embodiment, the present invention includes an apparatus having a register file to store vector data, an address generator coupled to the register file to generate addresses for a vector memory operation, and a controller to generate an output slice from one or more slices each including multiple addresses, where the output slice includes addresses each corresponding to a separately addressable portion of a memory. Other embodiments are described and claimed.

Abstract: In one embodiment, the present invention includes an apparatus having a register file to store vector data, an address generator coupled to the register file to generate addresses for a vector memory operation, and a controller to generate an output slice from one or more slices each including multiple addresses, where the output slice includes addresses each corresponding to a separately addressable portion of a memory. Other embodiments are described and claimed.

Abstract: A method and apparatus for partitioning a shared cache of a chip multi-processor are described. In one embodiment, the method includes a request of a cache block from system memory if a cache miss within a shared cache is detected according to a received request from a processor. Once the cache block is requested, a victim block within the shared cache is selected according to a processor identifier and a request type of the received request. In one embodiment, selection of the victim block according to a processor identifier and request type is based on a partition of a set-associative, shared cache to limit the selection of the victim block from a subset of available cache ways according to the cache partition. Other embodiments are described and claimed.

Abstract: In one embodiment, the present invention includes an apparatus having a register file to store vector data, an address generator coupled to the register file to generate addresses for a vector memory operation, and a controller to generate an output slice from one or more slices each including multiple addresses, where the output slice includes addresses each corresponding to a separately addressable portion of a memory. Other embodiments are described and claimed.

Abstract: Embodiments of apparatuses and methods for reducing the uncorrectable error rate in a lockstepped dual-modular redundancy system are disclosed. In one embodiment, an apparatus includes two processor cores, a micro-checker, a global checker, and fault logic. The micro-checker is to detect whether a value from a structure in one core matches a value from the corresponding structure in the other core. The global checker is to detect lockstep failures between the two cores. The fault logic is to cause the two cores to be resynchronized if there is a lockstep error but the micro-checker has detected a mismatch.