Abstract: Methods and systems are described. A system includes a redundant shader pipe array that performs rendering calculations on data provided thereto and a shader pipe array that includes a plurality of shader pipes, each of which performs rendering calculations on data provided thereto. The system also includes a circuit that identifies a defective shader pipe of the plurality of shader pipes in the shader pipe array. In response to identifying the defective shader pipe, the circuit generates a signal. The system also includes a redundant shader switch. The redundant shader switch receives the generated signal, and, in response to receiving the generated signal, transfers the data for the defective shader pipe to the redundant shader pipe array.

Abstract: Techniques for scheduling operations for a task graph on a processing device are provided. The techniques include receiving a task graph that specifies one or more passes, one or more resources, and one or more directed edges between passes and resources; identifying independent passes and dependent passes of the task graph; based on performance criteria of the processing device, scheduling commands to execute the passes; and transmitting scheduled commands to the processing device for execution as scheduled.

Abstract: A first workgroup is preempted in response to threads in the first workgroup executing a first wait instruction including a first value of a signal and a first hint indicating a type of modification for the signal. The first workgroup is scheduled for execution on a processor core based on a first context after preemption in response to the signal having the first value. A second workgroup is scheduled for execution on the processor core based on a second context in response to preempting the first workgroup and in response to the signal having a second value. A third context it is prefetched into registers of the processor core based on the first hint and the second value. The first context is stored in a first portion of the registers and the second context is prefetched into a second portion of the registers prior to preempting the first workgroup.

Abstract: Described herein are techniques for performing ray tracing operations. A command processor executes custom instructions for orchestrating a ray tracing pipeline. The custom instructions cause the command processor to perform a series of loop iterations, each at a particular recursion depth. In a first loop iteration, a ray generation shader is executed that triggers execution of a trace ray operation. In any other iteration, zero or more shaders are executed based on the contents of a shader queue. Any shader may trigger execution of a trace ray operation. The trace ray operation determines whether a ray specified by the shader intersects a triangle. The ray trace operation places shader entries into a shader queue, at the current recursion depth plus 1. The command processor updates the current recursion depth based on whether a trace ray operation is executed. The loop ends when the recursion depth is less than a threshold.

Abstract: Techniques for executing an atomic command in a distributed computing network are provided. A core cluster, including a plurality of processing cores that do not natively issue atomic commands to the distributed computing network, is coupled to a translation unit. To issue an atomic command, a core requests a location in the translation unit to write an opcode and operands for the atomic command. The translation unit identifies a location (a “window”) that is not in use by another atomic command and indicates the location to the processing core. The processing core writes the opcode and operands into the window and indicates to the translation unit that the atomic command is ready. The translation generates an atomic command and issues the command to the distributed computing network for execution. After execution, the distributed computing network provides a response to the translation unit, which provides that response to the core.

Abstract: Systems, apparatuses, and methods for implementing a dual-purpose millimeter-wave frequency band transmitter are disclosed. A system includes a dual-purpose transmitter sending a video stream over a wireless link to a receiver. In some embodiments, the video stream is generated as part of an augmented reality (AR) or virtual reality (VR) application. The transmitter operates in a first mode to scan and map an environment of the transmitter and receiver. The transmitter generates radio frequency (RF) signals in a first frequency range while operating in the first mode. Additionally, the transmitter operates in a second mode to send video data to the receiver, and the transmitter generates RF signals in the first frequency range while operating in the second mode.

Abstract: An electronic device includes a processor having a cache memory, a plurality of physical registers, and a promotion logic functional block. The promotion logic functional block promotes prefetched data from a portion of a cache block in the cache memory into a given physical register, the promoting including storing the prefetched data in the given physical register. Upon encountering a load micro-operation that loads data from the portion of the cache block into a destination physical register, the promotion logic functional block sets the processor so that the prefetched data stored in the given physical register is provided to micro-operations that depend on the load micro-operation.

Abstract: An electronic device that includes a controller functional block and a computational functional block having one or more computational elements performs operations associated with a training operation for a generative adversarial network, the generative adversarial network including a generative network and a discriminative network. The controller functional block determines one or more characteristics of the generative adversarial network. Based on the one or more characteristics, the controller functional block configures the one or more computational elements to perform processing operations for each of the generative network and the discriminative network during the training operation for the generative adversarial network.

Abstract: A communication network includes: a plurality of nodes in a topology, with each node having an upstream and a downstream neighboring node in the topology; a separate unidirectional communication link coupled between each node and that node's downstream neighboring node; and a separate unidirectional control link coupled between each node and that node's upstream neighboring node. A controller in each node keeps a count of packets sent by that node via the corresponding unidirectional communication link. The controller uses the count of packets sent to determine whether a given packet is allowed to be sent from that node to the downstream neighboring node and, if so, whether a full rate or a throttled rate is to be used for sending the given packet. Based at least in part on the determining, the controller selectively sends the given packet to the downstream neighboring node.

Abstract: Systems, apparatuses, and methods for executing a shader core instruction to invoke depth culling are disclosed. A shader core executes an instruction to invoke a culling function on a depth culling unit for one or more entities prior to completing a corresponding draw call. The shader core provides a mode and coordinates to the depth culling unit as a result of executing the instruction. The depth culling unit implements the culling function to access a live depth buffer to determine whether one or more primitives corresponding to the entities are occluded. The culling unit returns indication(s) to the shader core regarding the result(s) of processing the one or more primitives. For example, if the results indicate a primitive is occluded, the shader core cancels the draw call for the primitive.

Abstract: An electronic device has an activation function functional block that implements an activation function. During operation, the activation function functional block receives an input including a plurality of bits representing a numerical value. The activation function functional block then determines a range from among a plurality of ranges into which the input falls, each range including a separate portion of possible numerical values of the input. The activation function functional block next generates a result of a linear function associated with the range. Generating the result includes using a separate linear function that is associated with each range in the plurality of ranges to approximate results of the activation function within that range.

Abstract: A memory controller interfaces with a dynamic random access memory (DRAM). The memory controller selectively places memory commands in a memory interface queue and transmits the memory commands from the memory interface queue to a memory channel coupled to at least one dynamic random access memory (DRAM). An activate counter is maintained related to a number of activate commands sent over the memory channel to a memory region of the DRAM. In response to the activate counter being at or above a designated threshold, an arbiter is signaled that a refresh command should be sent to the memory region. In response to a designated condition, a value of the activate counter is adjusted by a total number based on a first fixed number and second varying number selected with one of random selection and pseudo-random selection.

Abstract: An integrated circuit device includes a memory controller coupleable to a memory. The memory controller to schedule memory accesses to regions of the memory based on memory timing parameters specific to the regions. A method includes receiving a memory access request at a memory device. The method further includes accessing, from a timing data store of the memory device, data representing a memory timing parameter specific to a region of the memory cell circuitry targeted by the memory access request. The method also includes scheduling, at the memory controller, the memory access request based on the data.

Abstract: Systems, apparatuses, and methods for identifying response data arriving out-of-order from two different memory types are disclosed. A computing system includes one or more clients for processing applications. A memory channel transfers memory traffic between a memory controller and a memory bus connected to each of a first memory and a second memory different from the first memory. The memory controller determines a given point in time when read data is to be scheduled to arrive on the memory bus from memory. The memory controller associates a unique identifier with the given point in time. The memory controller identifies a given command associated with the arriving read data based on the given point in time.

Abstract: Devices, methods, and systems for determining N-dimensional MaxPool or AvgPool for a M-dimensional input array. For each of N dimensions, in order from highest to lowest dimension i: the M dimensional input array is decomposed into 1 dimensional (1D) input arrays in the ith dimension, 1D MaxPool or AvgPool is performed on each of the 1D input arrays in the ith dimension to generate 1D output arrays in the ith dimension, and the M dimensional input array is recomposed from the 1D output arrays in the ith dimension to update the M-dimensional input array. In MaxPool, the updated M-dimensional input array is output as an M-dimensional output array. In AvgPool, each element of the updated M-dimensional input array is divided by a kernel size to form the M-dimensional output array.

Abstract: An electronic device includes a cache with a cache controller and a cache memory. The electronic device also includes a cache policy manager. The cache policy manager causes the cache controller to use two or more cache policies for cache operations in each of multiple test regions in the cache memory, with different configuration values for the two or more cache policies being used in each test region. The cache policy manager selects a selected configuration value for at least one cache policy of the two or more cache policies based on performance metrics for cache operations while using the different configuration values for the two or more cache policies in the test regions. The cache policy manager causes the cache controller to use the selected configuration value when using the at least one cache policy for cache operations in a main region of the cache memory.

Abstract: A single stage transmitter that operates at high speed is configured to operate as a driver in write mode and a termination in read mode. The driver configuration includes two circuits. The first circuit includes a PMOS cross-coupled device and a PMOS cascode circuit. The second circuit includes a NMOS cross-coupled device and a NMOS cascode circuit. The PMOS cross-coupled device and the NMOS cross-coupled device is connected in series by alternating current (AC) coupling capacitors. The termination configuration includes a third circuit including MOSFET transmission gates and an inverter controlled by a termination mode enable signal. In write mode, the third circuit of the single stage transmitter is turned off and the first and second circuits are operational. In read mode, the first and second circuits of the single stage transmitter are inactive and the third circuit is operational.

Abstract: Described herein are techniques for performing compilation operations for heterogeneous code objects. According to the techniques, a compiler identifies architectures targeted by a compilation unit, compiles the compilation unit into a heterogeneous code object that includes a different code object portion for each identified architecture, performs name mangling on functions of the compilation unit, links the heterogeneous code object with a second code object to form an executable, and generates relocation records for the executable.

Abstract: Systems, apparatuses, and methods for implementing a fastpath microcode sequencer are disclosed. A processor includes at least an instruction decode unit and first and second microcode units. For each received instruction, the instruction decode unit forwards the instruction to the first microcode unit if the instruction satisfies at least a first condition. In one implementation, the first condition is the instruction being classified as a frequently executed instruction. If a received instruction satisfies at least a second condition, the instruction decode unit forwards the received instruction to a second microcode unit. In one implementation, the first microcode unit is a smaller, faster structure than the second microcode unit. In one implementation, the second condition is the instruction being classified as an infrequently executed instruction.

Abstract: A processing system includes a processing unit and a memory device. The memory device includes a processing-in-memory (PIM) module that performs processing operations on behalf of the processing unit. An instruction set architecture (ISA) of the PIM module has fewer instructions than an ISA of the processing unit. Instructions received from the processing unit are translated such that processing resources of the PIM module are virtualized. As a result, the PIM module concurrently performs processing operations for multiple threads or applications of the processing unit.