31 results about "Hardware thread" patented technology

Described herein is a graphics processor having processing resources with configurable thread and register configurations. Program code can configure a number of registers and accumulators that will be used by hardware threads during execution of the program code by the graphics processor. Processing resources within the graphics processor can be configured to assign different numbers of registers and accumulators to hardware threads based on the configuration requested by program code to be executed by the processing resource.

Described herein is a graphics processor comprising a graphics processing cluster coupled with the memory interface, the graphics processing cluster including a plurality of processing resources, a processing resource of the plurality of processing resources including a register file including a first plurality of registers associated with a first hardware thread of a plurality of hardware threads of the processing resource and a second plurality of registers associated with a second hardware thread of the plurality of hardware threads of the processing resource and first circuitry configured to facilitate access to memory on behalf of the plurality of hardware threads and store metadata for memory access requests from the plurality of hardware threads.

The invention provides a processor table look-up method, a processor, equipment and a readable medium, and belongs to the field of processors. The method comprises the steps that the hardware thread obtains a target process corresponding to a table look-up key value under the condition that parallel table look-up is executed every time; wherein the target process is a process in an idle state in the at least two processes; calling the target process to look up a table according to the table look-up key value, and updating the state of the target process to a state of waiting for returning a table look-up result; and the hardware thread updates the state of the corresponding target process to a result returned state when monitoring return information of the table look-up result, and stores the table look-up result. The method is used for improving processor performance.

Described herein is a graphics processor having processing resources with configurable thread and register configurations. Program code can configure a number of registers and accumulators that will be used by hardware threads during execution of the program code by the graphics processor. Processing resources within the graphics processor can be configured to assign different numbers of registers and accumulators to hardware threads based on the configuration requested by program code to be executed by the processing resource.

A system can include a plurality of hardware threads, one or more schedulers, and one or more execution pipelines. At least one hardware thread of the plurality of hardware threads can include one or more finite state machines. At least one finite state machine of the one or more finite state machines can be of a first type. The at least one finite state machine can process hazards on instructions. The at least one finite state machine can determine cycles in which the instructions are safe to issue to at least one of the one or more execution pipelines.

A method of microprocessor validation includes creating a plurality of actionable items in a system memory for execution by multiple hardware threads, creating a hardware thread table structure in the system memory, wherein the hardware thread table structure contains the multiple hardware threads, creating a Process Table Structure (PTS) table in the system memory, where the PTS table includes entry points into the plurality of actionable items, creating a re-entrant entry point function (FSCHED) in the system memory to allow the multiple hardware threads to enter the plurality of actionable items, inspecting each of actionable items to identify an inactive actionable item, where if an inactive actionable item is identified, the FSCHED atomically locks the inactive actionable item and a hardware thread enters the inactive actionable item and executes the inactive actionable item until finished, where the FSCHED locks the inactive actionable item and continues searching for inactive actionable items.

A method of microprocessor validation includes creating a plurality of actionable items in a system memory for execution by multiple hardware threads, creating a hardware thread table structure in the system memory, wherein the hardware thread table structure contains the multiple hardware threads, creating a Process Table Structure (PTS) table in the system memory, where the PTS table includes entry points into the plurality of actionable items, creating a re-entrant entry point function (FSCHED) in the system memory to allow the multiple hardware threads to enter the plurality of actionable items, inspecting each of actionable items to identify an inactive actionable item, where if an inactive actionable item is identified, the FSCHED atomically locks the inactive actionable item and a hardware thread enters the inactive actionable item and executes the inactive actionable item until finished, where the FSCHED locks the inactive actionable item and continues searching for inactive actionable items.

The present disclosure provides a processor table lookup method, a processor, a device, and a readable medium. The processor table lookup method is applied to any hardware thread in a processor, and the hardware thread comprises at least two processes. The processor table lookup method comprises: in each execution of a parallel table lookup, obtaining a target process corresponding to a table lookup key value, wherein the target process is an idle process among the at least two processes; calling the target process to perform a table lookup on the basis of the table lookup key value, and updating the state of the target process to waiting for a table lookup result to be returned; and upon detecting return information of a table lookup result, updating the state of the target process to an idle state and saving the table lookup result.

The invention discloses a message interruption transmission system and a message interruption transmission method, and belongs to the technical field of integrated circuit processing systems. The system comprises an interrupt controller, a system bus, a first conversion module and a first hardware thread, the interrupt controller is used for sending a first message interrupt to the system bus, a destination node of the first message interrupt is the first hardware thread, and a destination node of the first message interrupt is the second hardware thread. The first hardware thread does not have an adaptive incoming message interrupt controller (IMSIC); the system bus is used for interrupting and routing the first message to the first conversion module connected with the first hardware thread; and the first conversion module is used for converting the first message interrupt into a line interrupt request and providing the line interrupt request for the first hardware thread.

The invention provides a network-on-chip many-core architecture and a data processing method thereof, relates to the technical field of data communication, and is used for solving the technical problem that the capability of supporting service characteristics based on the network-on-chip many-core architecture is limited. According to the method, a plurality of physical firmware micro-cores are virtualized into a logic firmware micro-core, and each physical firmware micro-core is equivalent to a hardware thread of the logic firmware micro-core. According to the method, instruction codes of a plurality of business characteristics to be operated on a logic firmware micronucleus are divided into a plurality of task groups, and each physical firmware micronucleus only operates the instruction codes in one of the plurality of task groups, so that the codes of different task groups are distributed on different physical firmware micronucleus in the logic firmware micronucleus; therefore, the resource requirements of each firmware micro-core on the IRAM and the DRAM are reduced while various service characteristics are supported.

An apparatus and method for hard-partitioned threading in a clustered processor core. For example, one embodiment of a processor comprises: front end circuitry to fetch instructions of a number of software threads from a memory; out-of-order execution circuitry comprising a set of partitionable execution resources to execute the instructions; and circuitry to dynamically allocate the set of partitionable execution resources to a plurality of hardware threads, wherein a different isolated subset of the partitionable execution resources are allocated to each hardware thread based, at least in part, on characteristics of each respective software thread and the number of software threads.

A deterministic replay of a multi-threaded trace on a multi-threaded processor is described. An example of a computer-readable storage medium includes instructions to cause at least one processor to receive graphics processing unit (GPU) program code for tracing, the program code including a plurality of instructions; analyze the plurality of instructions to identify instructions of the program code that are events requiring synchronization; instrument each of the identified events to generate instrumented program code; execute the instrumented program code on a plurality of hardware threads of the GPU to generate trace data; and emulate the trace data utilizing an emulator on a plurality of hardware traces of a central processing unit (CPU), including replaying the identified events according to an order of occurrence of the identified events.

A method includes receiving a network packet into a hardware pipeline of a network device; parsing and retrieving information of the network packet; determining, by the hardware pipeline, a packet-processing action to be performed by matching the information to a data structure of a set of flow data structures; sending, by the hardware pipeline, an action request to a programmable core, the action request being populated with data to trigger the programmable core to execute a hardware thread to perform a job, which is associated with the packet-processing action and that generates contextual data; retrieving the contextual data updated by the programmable core; and integrating the contextual data into performing the packet-processing action.

The invention discloses a chip detection method and device, and relates to the field of chips. The computing device determines the number of the hardware threads in the first state, and determines the chip utilization rate according to the proportion of the hardware threads in the first state in all the hardware threads included in the programmable chip. In the embodiment of the invention, the computing equipment directly determines the chip utilization rate according to the state of the hardware thread included in the programmable chip, and under the condition that the service flow processed by the programmable chip does not reach the processing upper limit of the programmable chip, the computing equipment can also determine the chip utilization rate, and the determined chip utilization rate is accurate. And the computing device does not need to determine the chip utilization rate by detecting packet loss, and normal operation of services is not affected.

The invention provides a hardware thread switching method and equipment. The method comprises the steps that a program execution direction control unit sends a PackSave instruction; wherein the PackSave instruction is used for triggering a storage site, and the PackSave instruction carries a thread identifier; the processor blocks the assembly line based on the PackSave instruction and stops distributing a new instruction to enter the assembly line; storing field related data according to the thread identifier; recovering the assembly line; wherein the field related data comprises but is not limited to register file information, a logic register and physical register mapping table, condition register information and state register information. According to the method provided by the invention, the switching of the hardware threads can be realized through one PackSave instruction, so that the switching time of the hardware threads is shortened, and the switching speed of the hardware threads is improved.

The invention relates to the technical field of computers, and particularly discloses a multi-core processor-oriented dynamic thread scheduling and cache optimization method, which comprises the following steps of: monitoring and collecting performance monitoring counting data of active hardware threads in real time, calculating cache conflict factors among the threads, and identifying thread pairs with serious conflicts. And generating a thread migration decision to migrate one thread in the conflict thread pair to a target processor core, and pre-distributing a protected cache partition in a final-stage cache of the target core for the thread to be migrated before a thread migration action is executed, and pre-loading historical access hot data. According to the method and the device, the effects of effectively reducing the cache conflict between the threads, improving the cache utilization rate and the performance of the multi-core processor, and further optimizing the system performance by reasonably scheduling the threads and preferentially processing the threads with high branch prediction misjudgment rate are achieved.

The application discloses a message interrupt transmission system and a message interrupt transmission method, and belongs to the technical field of integrated circuit processing systems. The system comprises an interrupt controller, a system bus, a first conversion module and a first hardware thread. The interrupt controller is used for sending a first message interrupt to the system bus. The destination node of the first message interrupt is the first hardware thread, and the first hardware thread does not have an adaptive incoming message interrupt controller (IMSIC). The system bus is used for routing the first message interrupt to the first conversion module connected with the first hardware thread. The first conversion module is used for converting the first message interrupt into a line interrupt request, and providing the line interrupt request to the first hardware thread.

The invention discloses a distributed register file cache for reducing L1 bandwidth requirements. Described herein is a graphics processor comprising a graphics processing cluster coupled with a memory interface, the graphics processing cluster comprising a plurality of processing resources, the processing resources of the plurality of processing resources comprising: a register file, the register file comprises a plurality of first registers associated with a first hardware thread in the plurality of hardware threads for processing the resources and a plurality of second registers associated with a second hardware thread in the plurality of hardware threads for processing the resources; and first circuitry configured to facilitate access to the memory on behalf of the plurality of hardware threads and store metadata of memory access requests from the plurality of hardware threads.

At least one computer-readable storage medium comprising instructions for execution by at least one graphics processing unit (GPU) that, when executed, cause the at least one GPU to:obtain program code for tracing, the program code including a plurality of instructions;identify from the plurality of instructions of the program code events to be synchronized;instrument the program code corresponding to one or more of the events identified, by inserting instructions that support monitoring code;execute the instrumented program code on at least a plurality of hardware threads of the GPU and generate trace data;replay the identified events according to an order of occurrence of the events identified; andreport a GPU state indicating a utilization of the GPU based; andwherein to report the GPU state includes to indicate when the GPU executes non-graphics related tasks.

This application provides a hardware thread switching method and device. The method includes: sending a PackSave instruction to a control unit; wherein the PackSave instruction is used to trigger saving the context, and the PackSave instruction carries a thread identifier; the processor, based on the PackSave instruction, blocks the pipeline and stops dispatching new instructions into the pipeline; saving relevant context data according to the thread identifier; and restoring the pipeline; wherein the relevant context data includes, but is not limited to: register file information, logical register-physical register mapping table, condition register information, and status register information. The method provided by this application can achieve hardware thread switching with a single PackSave instruction, reducing hardware thread switching time and improving hardware thread switching speed.

A multi-thread processor can include logic that can be implemented in a form of physical logic gates, a plurality of hardware threads, one or more execution units that can execute one or more instructions, a least a level 1 cache that can be for data, and selection logic. Each hardware thread of the plurality of hardware threads can fetch instructions from a software thread of execution assigned to a corresponding hardware thread of the plurality of hardware threads. The selection logic can include one or more fair ready-schedulers. Each fair ready-scheduler of the one or more fair ready-schedulers can select instructions that are ready for execution from among the plurality of hardware threads. A distribution of which a hardware thread is chosen can be consistent with a pattern expected from choosing a ready instruction according to a uniform probability distribution.

A DMA architecture capable of performing multi-level, multi-step, and parallel determination of multiple memory addresses is described. In one aspect, the DMA system includes one or more hardware DMA threads. Each DMA thread includes a request generator configured to generate m memory addresses for a multidimensional tensor in parallel during each parallel memory address computation cycle, and to generate a corresponding request for performing a memory operation on the memory system for each address. The request generator includes m memory address units, each memory address unit including a step size tracker configured to generate a corresponding step size index value for each dimension of the tensor, and to generate a corresponding step offset value for that dimension based on the corresponding step size index value. Each memory address unit includes a memory address computation element configured to generate memory addresses for tensor elements and send requests to perform memory operations.

A device includes a hardware data processing node configured to perform corresponding tasks, and a hardware thread scheduler including a hardware task scheduler (300). The hardware task scheduler (300) is coupled to the hardware data processing node and has a producer socket (304), a consumer socket (302), and a spare socket (306, 307). The spare socket is configured to provide data control signals, also provided by the first socket of the producer and consumer sockets, in response to a first value of the memory-mapped register (224). The spare socket is configured to provide data control signals, also provided by the second socket of the producer and consumer sockets, in response to a second value of the memory-mapped register (224).

Techniques for broadcast translation lookaside buffer invalidation are described. In some examples, an instance of a single instruction includes fields for a single instruction having fields to indicate one or more operations to perform, an identifier of a first source operand, and, in some examples, an identifier of a second source operand is to be handled by execution circuitry. The execution circuitry of a first hardware thread is to execute the decoded instruction according to the fields to indicate one or more operations to broadcast a command to a plurality of hardware threads to cause an invalidation one or more translation lookaside buffer (TLB) entries for each of the plurality of hardware threads in accordance with a value of the first source operand and / or a value of the second source operand.

The present invention provides a many-core architecture chip on a chip and its data processing method, which relates to the technical field of data communication and is used to solve the technical problem of limited ability to support service characteristics based on the many-core architecture on a chip. The present invention virtualizes multiple physical firmware micro-cores into a logical firmware micro-core, and each physical firmware micro-core is equivalent to a hardware thread of the logical firmware micro-core. The instruction codes of multiple service characteristics to be run on the logical firmware micro-core are divided into multiple task groups, and each physical firmware micro-core only runs the instruction codes in one task group among the multiple task groups, so that the codes of different task groups are distributed on different physical firmware micro-cores in the logical firmware micro-core, thereby realizing the support for multiple service characteristics while reducing the resource requirements of each firmware micro-core for IRAM and DRAM.

DMA architectures capable of performing multi-level multi-striding and determining multiple memory addresses in parallel are described. In one aspect, a DMA system includes one or more hardware DMA threads. Each DMA thread includes a request generator configured to generate, during each parallel memory address computation cycle, m memory addresses for a multi-dimensional tensor in parallel and, for each memory address, a respective request for a memory system to perform a memory operation. The request generator includes m memory address units that each include a step tracker configured to generate, for each dimension of the tensor, a respective step index value for the dimension and, based on the respective step index value, a respective stride offset value for the dimension. Each memory address unit includes a memory address computation element configured to generate a memory address for a tensor element and transmit the request to perform the memory operation.

This invention comprises three layers of parallelism: 1. A 64-bit parallel layer, using a single 64-bit instruction within a single 64-bit CPU core to simultaneously process arithmetic operations on eight base-64 data columns, each column consisting of 6 data bits and 2 carry bits. 2. A multi-core parallel layer, executing 64-bit instructions simultaneously on multiple CPU cores, with each core executing independently. 3. A hyper-threaded parallel layer, using two hardware threads within each core to simultaneously execute different tasks, sharing an execution unit. These three layers are independent, requiring no synchronization between them, resulting in a total performance improvement to multiplicative level rather than additive level. Real-world testing shows a 128x speedup on an 8-core, 16-thread CPU, breaking the Amdahl's Law limitation and achieving linear scalability without a saturation point compared to traditional architectures.

A method for a hardware thread allocation controller of a computing device includes receiving, at the hardware thread allocation controller, a processor thread including one or more processing commands to be executed, and one or more preconditions for execution of the processor thread. The processor thread is stored in a thread queue of the hardware thread allocation controller. Based at least in part on detecting that the one or more preconditions for execution are met, the processor thread is assigned to a hardware accelerator of the computing device for execution, wherein the hardware thread allocation controller is configured to receive the processor thread, store the processor thread in the thread queue, and assign the processor thread to the hardware accelerator through hardware programming of the hardware thread allocation controller.

The invention is suitable for the technical field of multi-thread queue processing, and provides an RDMA queue asynchronous processing method and device based on hardware multithreading, and the method comprises the steps: creating a first hardware thread context and a second hardware thread context in a hardware multi-thread engine of an RDMA network card; according to the first asynchronous event, triggering a first hardware thread context to access a submission queue and analyzing a first work queue element; triggering a second hardware thread context to access a completion queue according to the second asynchronous event and generating a completion queue element; and executing an RDMA protocol operation corresponding to the first working queue element through the hardware-level queue management unit, and writing a completion queue element into a host memory through a second hardware thread context after the operation is completed. According to the invention, the problem of ultra-high-speed network performance bottleneck caused by serialization processing delay, resource utilization imbalance and atomic operation efficiency loss in a high-concurrency scene is solved, and the utilization rate of resources is effectively improved.

A deterministic replay of a multi-threaded trace on a multi-threaded processor is described. An example of a computer-readable storage medium includes instructions to cause at least one processor to receive graphics processing unit (GPU) program code for tracing, the program code including a plurality of instructions; analyze the plurality of instructions to identify instructions of the program code that are events requiring synchronization; instrument each of the identified events to generate instrumented program code; execute the instrumented program code on a plurality of hardware threads of the GPU to generate trace data; and emulate the trace data utilizing an emulator on a plurality of hardware traces of a central processing unit (CPU), including replaying the identified events according to an order of occurrence of the identified events.