In various examples, one or more components or regions of a processing unit—such as a processing core, and/or component thereof—may be tested for faults during deployment in the field. To perform testing while in deployment, the state of a component subject to test may be retrieved and/or stored during the test to maintain state integrity, the component may be clamped to communicatively isolate the component from other components of the processing unit, a test vector may be applied to the component, and the output of the component may be compared against an expected output to determine if any faults are present. The state of the component may be restored after testing, and the clamp removed, thereby returning the component to its operating state without a perceivable detriment to operation of the processing unit in deployment.

Embodiments of the invention are directed to a computer-implemented method of unit environment verification. The method includes monitoring, by a processor, a data stream between a first driver and a device under test (DUT) in a unit verification environment. The processor retrieves a transaction value from a database, wherein the transaction value was generated in a higher-level verification environment than the unit verification environment. The processor transmits the retrieved transaction value to the DUT. The processor compares a response from the DUT to the transmitted transaction value to an expected value. In response to the comparison indicating an error, the processor initiates a repair of the error at the unit verification environment.

The disclosure describes techniques for a self-test of a graphics processing unit (GPU) independent of instructions from another processing device. The GPU may perform the self-test in response to a determination that the GPU enters an idle mode. The self-test may be based on information indicating a safety level, where the safety level indicates how many faults in circuits or memory blocks of the GPU need to be detected.

Method, apparatus and product for automatically introducing register dependency into tests. A test template represents an abstract test scenario to be utilized for testing a target processor. The abstract test scenario requires that a value be assigned to a register. A test that implements the abstract test scenario is generated. The test is a set of instructions that are executable by the target processor. The generation of the test comprises: determining a memory address to retain the value in a memory that is accessible to the target processor; and adding to the test an instruction to load to the register the value from the memory address, whereby adding a register dependency to the test that is not required by the abstract test scenario. The test can be executed on the target processor or simulation thereof.

Lockstep comparators and related methods are described. An example apparatus includes self-test logic circuitry having first outputs, and comparator logic including selection logic having first inputs and second outputs, ones of the first inputs coupled to the first outputs, first detection logic having second inputs and third outputs, the second inputs coupled to the second outputs, second detection logic having third inputs and fourth outputs, the third inputs coupled to the third outputs, latch logic having fifth inputs and fifth outputs, the third output and the fourth output coupled to the fifth inputs, and error detection logic having sixth inputs coupled to the fifth inputs.

The technology disclosed relates to inter-node execution of configuration files on reconfigurable processors using network interface controller (NIC) buffers. In particular, the technology disclosed relates to a runtime logic that is configured to execute configuration files that define applications and application data for applications using a first reconfigurable processor connected to a first host, and a second reconfigurable processor connected to a second host. The first reconfigurable processor is configured to push input data for the applications in a first plurality of buffers. The first host is configured to cause a first network interface controller (NIC) to stream the input data to a second plurality of buffers from the first plurality of buffers. The second host is configured to cause a second NIC to stream the input data to the second reconfigurable processor from the second plurality of buffers.

A system for training parameters of a neural network includes a processing node with a processor reconfigurable at a first level of configuration granularity and a controller reconfigurable at a finer level of configuration granularity. The processor is configured to execute a first dataflow segment of the neural network with training data to generate a predicted output value using a set of neural network parameters, calculate a first intermediate result for a parameter based on the predicted output value, a target output value, and a parameter gradient, and provide the first intermediate result to the controller. The controller is configured to receive a second intermediate result over a network, and execute a second dataflow segment, dependent upon the first intermediate result and the second intermediate result, to generate a third intermediate result indicative of an update of the parameter.

A system for executing a graph partitioned across a plurality of reconfigurable computing units includes a processing node that has a first computing unit reconfigurable at a first level of configuration granularity and a second computing unit reconfigurable at a second, finer, level of configuration granularity. The first computing unit is configured by a host system to execute a first dataflow segment of the graph using one or more dataflow pipelines to generate a first intermediate result and to provide the first intermediate result to the second computing unit without passing through the host system. The second computing unit is configured by the host system to execute a second dataflow segment of the graph, dependent upon the first intermediate result, to generate a second intermediate result and to send the second intermediate result to a third computing unit, without passing through the host system, to continue execution of the graph.

Disclosed is a quantum diagnostic circuit, which includes an input unit having an input of at least first to fourth qubits, a diagnostic circuit unit receiving the first to fourth qubits from the input unit and providing a quantum superposition and a quantum entanglement, and an output unit receiving an output of the diagnostic circuit unit and determining whether the output is in a Bell-state, and the diagnostic circuit unit includes a Hadamard gate processing the first qubit to provide the quantum superposition of the first to fourth qubits, a first CNOT gate providing the quantum entanglement between an output of the Hadamard gate and the second qubit, a second CNOT gate providing the quantum entanglement between an output of the first CNOT gate and the third qubit, and a third CNOT gate providing the quantum entanglement between an output of the second CNOT gate and the fourth qubit.

An application processor includes a central processing unit, a root complex that communicates with at least one external device under control of the central processing unit and generates a state change interrupt when an operation state changes, and an interrupt aggregation and debug unit that performs debugging on at least one component associated with the state change interrupt depending on the state change interrupt.