Adjustment of kernels for compilation on device

Abstract

Adjustments for a kernel compiled on a device are described. An example of a device includes a computer memory to store data for processing; in one embodiment, a computing device includes a computing core, and a processing resource including a GPU, the GPU including a compilation circuitry module, where the compilation circuitry module includes a core evaluation circuitry module to evaluate a computing core received for compilation and determine one or more characteristics of the computing core, and a device compiler circuitry module to support compilation of the computing core, where the compilation circuitry module includes a core evaluation circuitry module to evaluate the computing core received for compilation and to determine one or more characteristics of the computing core, and a device compiler circuitry module to support compilation of the computing core. Wherein the device compiler circuitry module is to adjust the compiling of the compute kernel based at least in part on one or more characteristics of the compute kernel.

Description

Priority Declaration

[0001] This application claims priority to U.S. Provisional Application Serial No. 63 / 565,508, filed by Radhakrishna Giduthuri et al. on March 14, 2024, entitled “GPU DEVICE FINE-TUNING OF KERNELS WITH GRAPH EXECUTION,” the entire contents of which are incorporated herein by reference. Background Art

[0002] In graphics processing operations, multiple different kernels (e.g., multiple different matrix operations) may be received for processing by a general-purpose processing unit (GPU) or other processing resources. Multiple kernels may be compiled for execution, where the kernels may have multiple different shapes.

[0003] When there is a kernel that achieves one or more operational shapes, there is typically an attempt to optimize the compilation for the one or more shapes. However, the optimization of the compilation depends at least in part on the kernel shape, which may vary significantly when the kernel is received. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] The embodiments described herein are illustrated by way of example and not limitation in the figures of the accompanying drawings in which like reference numerals indicate similar elements and in which:

[0005] Figure 1 is a block diagram illustrating a computer system configured to implement one or more aspects of the embodiments described herein;

[0006] Figures 2A-2E A parallel processor assembly comprising a graphics multiprocessor is described;

[0007] Figure 3 A graphics processing unit comprising a collection of dedicated graphics processing resources arranged into a multi-core group is described;

[0008] Figures 4A-4E An exemplary architecture is described in which multiple GPUs are communicatively coupled to multiple multi-core processors;

[0009] Figure 5 The graphics processing pipeline is explained;

[0010] Figure 6 Explains the machine learning software stack;

[0011] Figure 7 A general purpose graphics processing unit is described;

[0012] Figure 8A multi-GPU computing system is described;

[0013] Figures 9A-9B illustrates the layers of an exemplary deep neural network;

[0014] Figures 10A-10B An exemplary language model is described;

[0015] Figure 11 Demonstrates the training and deployment of deep neural networks;

[0016] Figure 12A is a block diagram illustrating distributed learning;

[0017] Figure 12B is a block diagram illustrating a programmable network interface and a data processing unit;

[0018] Figure 13 An exemplary inference system-on-chip (SOC) suitable for performing inference using a trained model is described;

[0019] Figure 14 It is a block diagram of the processing system;

[0020] Figures 15A-15C A computing system and a graphics processor are described;

[0021] Figure 16 is a block diagram of a graphics processor, which may be a discrete or integrated graphics processing unit;

[0022] Figures 17A-17B A block diagram illustrating an additional graphics processor and computing accelerator architecture;

[0023] Figures 18A-18C Thread execution logic comprising an array of processing elements employed in a graphics processor core is described;

[0024] Figure 19 A tile of a multi-chip processor according to an embodiment is described;

[0025] Figure 20 is a block diagram illustrating the graphics processor instruction format;

[0026] Figure 21 is a block diagram of the attached graphics processor architecture;

[0027] Figures 22A-22B Describes the graphics processor command format and command sequence;

[0028] Figure 23 An exemplary graphics software architecture for a data processing system is described;

[0029] Figure 24is a block diagram illustrating an IP core development system;

[0030] Figure 25A illustrates a cross-sectional side view of an integrated circuit package assembly including multiple units of hardware logic chiplets connected to a substrate (e.g., a base die);

[0031] Figure 25B A package assembly including interchangeable chiplets is described;

[0032] Figure 26 is a block diagram illustrating a system-on-chip integrated circuit;

[0033] Figure 27 is an illustration of an apparatus or system for providing kernel tuning using graphics execution according to some embodiments;

[0034] Figure 28 is an illustration of receiving a varying compute kernel for compilation at a processor core according to some embodiments;

[0035] Figure 29 is a diagram of a circuit module that supports on-device adjustment of a compiled kernel according to some embodiments;

[0036] Figure 30 is an illustration of a process compiled on a device including a kernel according to some embodiments; and

[0037] Figure 31 is an illustration of a process for on-device tuning of a compiled kernel, according to some embodiments. DETAILED DESCRIPTION

[0038] In some embodiments, an apparatus, system, or process provides tuning of kernels for on-device compilation. In some embodiments, to enable a graphics processing unit (GPU) or other accelerator to make accurate and efficient decisions regarding on-device compilation of kernels, information about one or more characteristics of the kernel is determined and applied to the compilation decisions.

[0039] Current parallel graphics data processing includes systems and methods developed to perform specific operations on graphics data, such as linear interpolation, tessellation, rasterization, texture mapping, depth testing, etc. Traditionally, graphics processors have used fixed-function compute units to process graphics data. However, recently, portions of graphics processors have been made programmable, enabling such processors to support a wider variety of operations for processing vertex and fragment data.

[0040] To further improve performance, graphics processors often implement processing techniques such as pipelining, which attempts to process as much graphics data as possible in parallel across different parts of the graphics pipeline. Parallel graphics processors with a single instruction multiple thread (SIMT) architecture are designed to maximize the amount of parallel processing in the graphics pipeline. In a SIMT architecture, groups of parallel threads attempt to execute program instructions together synchronously as often as possible to improve processing efficiency.

[0041] A graphics processing unit (GPU) is communicatively coupled to a host / processor core to accelerate, for example, graphics operations, machine learning operations, pattern analysis operations, and / or various general-purpose GPU (GPGPU) functions. The GPU can be communicatively coupled to the host processor / core via a bus or another interconnect (e.g., a high-speed interconnect such as PCIe or NVLink). Alternatively, the GPU can be integrated on the same package or chip as the core and can be communicatively coupled to the core via an internal processor bus / interconnect (i.e., inside the package or chip). Regardless of the manner in which the GPU is connected, the processor core can assign work to the GPU in the form of a sequence of commands / instructions contained in a work descriptor. The GPU then uses dedicated circuit modules / logic to efficiently process these commands / instructions.

[0042] Although the techniques described herein are primarily discussed in the context of GPUs, the techniques may also be implemented in other types of processors, including but not limited to general-purpose processors and accelerator devices such as artificial intelligence accelerators, vision processors, and neural processing units.

[0043] In the following description, numerous specific details are set forth to provide a more thorough understanding. However, it will be apparent to those skilled in the art that the embodiments described herein can be practiced without one or more of these specific details. In other instances, well-known features have not been described to avoid obscuring the details of the present embodiments. System Overview

[0044] Figure 1is a block diagram of a computing system 100 configured to implement one or more aspects of the embodiments described herein. The computing system 100 includes a processing subsystem 101 having one or more processors 102 and a system memory 104 that communicate via an interconnect path that may include a memory hub 105. The memory hub 105 may be a separate component within a chipset component or may be integrated within the one or more processors 102. The memory hub 105 is coupled to an I / O subsystem 111 via a communication link 106. The I / O subsystem 111 includes an I / O hub 107 that may enable the computing system 100 to receive input from one or more input devices 108. Additionally, the I / O hub 107 may enable a display controller that may be included in the one or more processors 102 to provide output to one or more display devices 110A. In one embodiment, the one or more display devices 110A coupled to the I / O hub 107 may include local, internal, or embedded display devices.

[0045] The processing subsystem 101 includes, for example, one or more parallel processors 112 coupled to the memory hub 105 via a communication link 113 (such as a bus or structure). The communication link 113 can be one of any number of standard-based communication link technologies or protocols (such as, but not limited to, PCI Express), or can be a vendor-specific communication interface or communication structure. The one or more parallel processors 112 can form a parallel or vector processing system in a computational cluster that can include a large number of processing cores and / or processing clusters, such as a many integrated core (MIC) processor. For example, the one or more parallel processors 112 form a graphics processing subsystem that can output pixels to one of the one or more display devices 110A coupled via the I / O hub 107. The one or more parallel processors 112 can also include a display controller and a display interface (not shown) to enable a direct connection to the one or more display devices 110B.

[0046] Within the I / O subsystem 111, a system storage unit 114 can be connected to the I / O hub 107 to provide a storage mechanism for the computing system 100. An I / O switch 116 can be used to provide an interface mechanism to enable connection between the I / O hub 107 and other components, such as a network adapter 118 and / or a wireless network adapter 119 that can be integrated into the platform, as well as various other devices that can be added with the aid of one or more plug-in devices 120. The plug-in device(s) 120 can also include, for example, one or more external graphics processor devices, graphics cards, and / or computing accelerators. The network adapter 118 can be an Ethernet adapter or another wired network adapter. The wireless network adapter 119 can include one or more of the following: Wi-Fi, Bluetooth, near field communication (NFC), or other network devices including one or more wireless radio devices.

[0047] The computing system 100 may include other components not explicitly shown, including USB or other port connections, optical storage drives, video capture devices, etc., which may also be connected to the I / O hub 107. Any suitable protocol may be used, such as a PCI (Peripheral Component Interconnect) based protocol (e.g., PCI Express) or any other bus or point-to-point communication interface and / or protocol(s) (such as NVLink high-speed interconnect, Compute Express Link TM (CXL TM ) ), such as CXL.mem, Infinity Fabric (IF), Ethernet (IEEE 802.3), Remote Direct Memory Access (RDMA), InfiniBand, Internet Wide Area RDMA Protocol (iWARP), Transmission Control Protocol (TCP), User Datagram Protocol (UDP), Quick UDP Internet Connection (QUIC), RDMA over Converged Ethernet (RoCE), Intel Quick Path Interconnect (QPI), Intel Ultra Path Interconnect (UPI), Intel System-on-Chip Fabric (IOSF), Omni-Path, HyperTransport, Advanced Microcontroller Bus Architecture (AMBA) Interconnect, OpenCAPI, Gen-Z, Cache Coherent Interconnect for Accelerators (CCIX), 3GPP Long Term Evolution (LTE), (4G), 3GPP 5G and variants thereof, or any other wired or wireless interconnect protocol known in the art. Figure 1 In some examples, data may be copied or stored to a virtualized storage node using a protocol such as Non-Volatile Memory Express (NVMe) over Fabric (NVMe-oF) or NVMe.

[0048] One or more parallel processors 112 can be combined with circuit modules optimized for graphics and video processing (including, for example, a video output circuit module) and constitute a graphics processing unit (GPU). Alternatively or in addition, as described in more detail herein, one or more parallel processors 112 can be combined with circuit modules optimized for general-purpose processing while retaining the underlying computing architecture. The components of the computing system 100 can be integrated on a single integrated circuit along with one or more other system elements. For example, one or more parallel processors 112, the memory hub 105, (one or more) processors 102, and the I / O hub 107 can be integrated into a system-on-chip (SoC) integrated circuit. Alternatively, the components of the computing system 100 can be integrated into a single package to form a system-in-package (SIP) configuration. In one embodiment, at least a portion of the components of the computing system 100 can be integrated into a multi-chip module (MCM), which can be interconnected with other multi-chip modules into a modular computing system.

[0049] In some configurations, in addition to the processor(s) 102 and the parallel processor(s) 112, the computing system 100 also includes one or more accelerator devices 130 coupled to the memory hub 105. The accelerator device(s) 130 are configured to perform domain-specific acceleration of workloads to handle computationally intensive or high-throughput tasks. The accelerator device(s) 130 can reduce the burden placed on the processor(s) 102 and / or the parallel processor(s) 112 of the computing system 100. The accelerator device(s) 130 can include, but are not limited to, intelligent network interface cards, data processing units, cryptographic accelerators, storage accelerators, artificial intelligence (AI) accelerators, neural processing units (NPUs), storage accelerators, and / or video transcoding accelerators. It will be appreciated that the computing system 100 shown herein is illustrative and that variations and modifications are possible. The connection topology, including the number and arrangement of bridges, the number of processor(s) 102, and the number of parallel processor(s) 112, can be modified as desired. For example, the system memory 104 can be connected directly to the processor(s) 102 rather than through a bridge, while other devices communicate with the system memory 104 via the memory hub 105 and the processor(s) 102. In other alternative topologies, the parallel processor(s) 112 are connected to the I / O hub 107 or directly to one of the processor(s) 102 rather than to the memory hub 105. In other embodiments, the I / O hub 107 and the memory hub 105 can be integrated into a single chip. It is also possible to attach two or more sets of processor(s) 102 via multiple sockets that can couple with two or more instances of the parallel processor(s) 112.

[0050] Some of the specific components shown herein are optional and may not be included in all implementations of computing system 100. For example, any number of plug-in cards or peripherals may be supported, or some components may be eliminated. In addition, some architectures may use different terminology for the same. Figure 1 For example, memory hub 105 may be referred to as a north bridge in some architectures, while I / O hub 107 may be referred to as a south bridge.

[0051] Figure 2AA parallel processor 200 is illustrated. The parallel processor 200 may be a GPU, GPGPU, or the like as described herein. Various components of the parallel processor 200 may be implemented using one or more integrated circuit devices such as a programmable processor, an application specific integrated circuit (ASIC), or a field programmable gate array (FPGA). The parallel processor 200 may be Figure 1 One or more of the parallel processor(s) 112 shown in .

[0052] Parallel processor 200 includes parallel processing unit 202. Parallel processing unit includes I / O unit 204, which enables communication with other devices including other instances of parallel processing unit 202. I / O unit 204 can be directly connected to other devices. For example, I / O unit 204 is connected to other devices by using a hub or switch interface such as memory hub 105. The connection between memory hub 105 and I / O unit 204 forms communication link 113. Within parallel processing unit 202, I / O unit 204 is connected to host interface 206 and memory crossbar switch 216, wherein host interface 206 receives commands directed to perform processing operations and memory crossbar switch 216 receives commands directed to perform memory operations. In one embodiment, I / O unit 204 is configured to enable secure I / O operations by means of trusted execution environment (TEE)-I / O support. TEE-IO enables trusted I / O virtualization, where a trust relationship can be directly established between a secure virtual environment, such as a trusted virtual machine, and a parallel processor 200 or a secure partition of a parallel processor.

[0053] When host interface 206 receives command buffers via I / O unit 204, it can direct work operations to execute those commands to front-end 208. In one embodiment, front-end 208 is coupled to scheduler 210, which is configured to dispatch commands or other work items to processing cluster array 212. Scheduler 210 ensures that processing cluster array 212 is properly configured and in a valid state before dispatching tasks to processing clusters within the processing cluster array 212. Scheduler 210 can be implemented using firmware logic executed on a microcontroller. A microcontroller-implemented scheduler 210 can be configured to perform complex scheduling and work dispatch operations at both coarse and fine granularity, thereby enabling rapid preemption and context switching of threads executing on processing cluster array 212. Preferably, host software can confirm workloads for scheduling on processing cluster array 212 via one of multiple graphics processing doorbells. In other examples, polling for new workloads or interrupts can be used to identify or indicate the availability of work to be executed. The workload may then be automatically distributed across the processing cluster array 212 by scheduler 210 logic within the scheduler microcontroller.

[0054] Processing cluster array 212 may include up to "N" processing clusters (e.g., cluster 214A, cluster 214B, through cluster 214N). Each cluster 214A-214N in processing cluster array 212 may execute a large number of concurrent threads. Scheduler 210 may allocate work to clusters 214A-214N in processing cluster array 212 using various scheduling and / or work distribution algorithms that may vary depending on the workload generated for each type of program or computation. Scheduling may be handled dynamically by scheduler 210 or may be assisted in part by compiler logic during the compilation of program logic configured for execution by processing cluster array 212. Optionally, different clusters 214A-214N in processing cluster array 212 may be assigned to process different types of programs or to perform different types of computations.

[0055] Processing cluster array 212 can be configured to perform various types of parallel processing operations. For example, processing cluster array 212 is configured to perform general-purpose parallel computing operations. For example, processing cluster array 212 can include logic for performing processing tasks including filtering video and / or audio data, performing modeling operations including physics operations, and performing data transformations.

[0056] The processing cluster array 212 is configured to perform parallel graphics processing operations. In such embodiments where the parallel processor 200 is configured to perform graphics processing operations, the processing cluster array 212 may include additional logic to support the execution of such graphics processing operations, including but not limited to texture sampling logic to perform texture operations, as well as tessellation logic and other vertex processing logic. In addition, the processing cluster array 212 may be configured to execute graphics processing related shader programs, such as but not limited to vertex shaders, tessellation shaders, geometry shaders, and pixel shaders. The parallel processing unit 202 may transfer data from the system memory via the I / O unit 204 for processing. During processing, the transferred data may be stored in an on-chip memory (e.g., the parallel processor memory 222) during processing and then written back to the system memory.

[0057] In embodiments where parallel processing units 202 are used to perform graphics processing, scheduler 210 may be configured to divide the processing workload into tasks of approximately equal size to better enable distribution of graphics processing operations to multiple clusters 214A-214N in processing cluster array 212. In some of these embodiments, portions of processing cluster array 212 may be configured to perform different types of processing. For example, a first portion may be configured to perform vertex shading and topology generation, a second portion may be configured to perform tessellation and geometry shading, and a third portion may be configured to perform pixel shading or other screen-space operations to produce a rendered image for display. Intermediate data generated by one or more of clusters 214A-214N may be stored in a buffer to allow the intermediate data to be transferred between clusters 214A-214N for further processing.

[0058] During operation, the processing cluster array 212 may receive processing tasks to be executed via the scheduler 210, which receives commands defining the processing tasks from the front end 208. For graphics processing operations, a processing task may include data to be processed (e.g., surface (patch) data, primitive data, vertex data, and / or pixel data) and state parameters and an index of commands that define how the data is to be processed (e.g., what program to execute). The scheduler 210 may be configured to fetch an index corresponding to a task or may receive an index from the front end 208. The front end 208 may be configured to ensure that the processing cluster array 212 is configured in a valid state before initiating a workload specified by an incoming command buffer (e.g., a batch buffer, a push buffer, etc.).

[0059] Each of the one or more instances of parallel processing unit 202 can be coupled to parallel processor memory 222. Parallel processor memory 222 can be accessed via memory crossbar switch 216, which can receive memory requests from processing cluster array 212 and I / O unit 204. Memory crossbar switch 216 can access parallel processor memory 222 via memory interface 218. Memory interface 218 can include a plurality of partition units (e.g., partition unit 220A, partition unit 220B, through partition unit 220N), each of which can be coupled to a portion (e.g., memory cells) of parallel processor memory 222. The number of partition units 220A-220N can be configured to be equal to the number of memory cells, such that each partition unit 220A-220N has a corresponding memory cell 224A-224N. In other embodiments, the number of partition units 220A-220N may not be equal to the number of memory cells.

[0060] Memory units 224A-224N may include various types of memory devices, including dynamic random access memory (DRAM) or graphics random access memory, such as synchronous graphics random access memory (SGRAM), including graphics double data rate (GDDR) memory. Optionally, memory units 224A-224N may also include 3D stacked memory, including but not limited to high bandwidth memory (HBM). Those skilled in the art will appreciate that the specific implementation of memory units 224A-224N may vary and may be selected from among a variety of conventional designs. Render targets, such as frame buffers or texture maps, may be stored across memory units 224A-224N, allowing partition units 220A-220N to write portions of each render target in parallel to efficiently use the available bandwidth of parallel processor memory 222. In some embodiments, local instances of parallel processor memory 222 may be eliminated in favor of a unified memory design for system memory combined with local cache memory.

[0061] Optionally, any of the clusters 214A-214N in the processing cluster array 212 has the capability to process data to be written to any of the memory cells 224A-224N within the parallel processor memory 222. The memory crossbar 216 can be configured to transmit the output of each cluster 214A-214N to any partition unit 220A-220N or to another cluster 214A-214N, which can perform additional processing operations on the output. Each cluster 214A-214N can communicate with a memory interface 218 via the memory crossbar 216 to read from or write to various external memory devices. In one embodiment having a memory crossbar 216, the memory crossbar 216 has connections to a memory interface 218 for communicating with the I / O unit 204, and to a local instance of parallel processor memory 222, thereby enabling processing units within different processing clusters 214A-214N to communicate with system memory or other memory that is not local to the parallel processing unit 202. Typically, the memory crossbar 216 may be capable of separating traffic flows between the clusters 214A-214N and the partition units 220A-220N, for example, using virtual channels.

[0062] Although a single instance of parallel processing unit 202 is illustrated within parallel processor 200, any number of instances of parallel processing unit 202 may be included. For example, multiple instances of parallel processing unit 202 may be provided on a single plug-in card, or multiple plug-in cards may be interconnected. For example, parallel processor 200 may be included in Figure 1 The plug-in device in the (one or more) plug-in device 120 can be a graphics card (such as a discrete graphics card including one or more GPUs, one or more memory devices, and a device-to-device or network or fabric interface). Different instances of parallel processing unit 202 can be configured to interoperate even if different instances have different numbers of processing cores, different amounts of local parallel processor memory, and / or other configuration differences. Optionally, some instances of parallel processing unit 202 can include higher precision floating point units relative to other instances. Systems incorporating one or more instances of parallel processing unit 202 or parallel processor 200 can be implemented in a variety of configurations and form factors, including but not limited to desktop computers, laptop computers, or handheld personal computers, servers, workstations, game consoles, and / or embedded systems. The orchestrator can use one or more of the following to form a composite node for workload execution: decomposed processor resources, cache resources, memory resources, storage resources, and networking resources.

[0063] In one embodiment, the parallel processing unit 202 can be partitioned into multiple instances. Those multiple instances can be configured to execute workloads associated with different clients in an isolated manner, thereby enabling a predetermined quality of service to be provided to each client. For example, each cluster 214A-214N can be partitioned and isolated from the other clusters, allowing the processing cluster array 212 to be divided into multiple computing partitions or instances. In such a configuration, workloads executed on isolated partitions are protected from failures or errors associated with different workloads executed on different partitions. Partition units 220A-220N can be configured to enable dedicated and / or isolated paths to memory for clusters 214A-214N associated with the corresponding computing partition. This data path isolation allows computing resources within a partition to communicate with one or more assigned memory units 224A-224N without being interfered with by the activities of other partitions. In one embodiment, data path isolation can be enhanced by encrypting data in the memory of various partitions using encryption keys unique to the associated partition, so that data is secure on a per-partition basis both at rest and in transit.

[0064] Figure 2B is a block diagram of the partition unit 220. The partition unit 220 may be Figure 2A20N。 As illustrated, the partition unit 220 includes an L2 cache 221, a frame buffer interface 225, and a ROP 226 (raster operation unit). The L2 cache 221 is a read / write cache configured to perform load and store operations received from the memory crossbar 216 and the ROP 226. Read misses and urgent writeback requests are output by the L2 cache 221 to the frame buffer interface 225 for processing. Updates can also be sent to the frame buffer via the frame buffer interface 225 for processing. In one embodiment, the frame buffer interface 225 is connected to the L2 cache 221. Figure 2A The memory unit 224 in the memory units 224A-224N of the parallel processor memory 222 is connected by an interface. The partition unit 220 may also be additionally or alternatively connected to one of the memory units in the parallel processor memory by an interface via a memory controller (not shown).

[0065] In graphics applications, ROP 226 is a processing unit that performs raster operations such as stenciling, z-testing, blending, and the like. ROP 226 then outputs processed graphics data, which is stored in graphics memory. In some embodiments, ROP 226 includes or is coupled to CODEC 227, which includes compression logic for compressing depth or color data written to memory or L2 cache 221 and decompressing depth or color data read from memory or L2 cache 221. The compression logic may be lossless compression logic that utilizes one or more of a variety of compression algorithms. The type of compression performed by CODEC 227 may vary based on the statistical characteristics of the data to be compressed. For example, in one embodiment, delta color compression is performed on depth and color data on a per-slice basis. In one embodiment, CODEC 227 includes compression and decompression logic that can compress and decompress computational data associated with machine learning operations. CODEC 227 can for example compress sparse matrix data for sparse machine learning operation.CODEC 227 can also compress the sparse matrix data encoded in sparse matrix format (such as coordinate list encoding (COO), compressed sparse row (CSR), compressed sparse column (CSC) etc.) to generate compressed and encoded sparse matrix data.Compressed and encoded sparse matrix data can be decompressed and / or decoded before being processed by processing element, or processing element can be configured to consume compressed, encoded or compressed and encoded data for processing.In one embodiment, CODEC 227 can be configured as a general data compression engine, for GPU database acceleration and large-capacity data analysis use.

[0066] ROP 226 may be included in each processing cluster (e.g. Figure 2A Rather than being included in partition unit 220, the processed graphics data is stored in clusters 214A-214N of FIG. 2 . In such an embodiment, read requests and write requests for pixel data, rather than pixel fragment data, are routed through memory crossbar 216. The processed graphics data may be displayed on a display device such as a Figure 1 10B), is routed for further processing by the processor(s) 102, or is routed for processing by the processor(s) 102. Figure 2A The processing is further processed by one of the processing entities within the parallel processor 200.

[0067] Figure 2C is a block diagram of a processing cluster 214 within a parallel processing unit. For example, a processing cluster represents Figure 2A An instance of a processing cluster in one of the processing clusters 214A-214N. Processing cluster 214 can be configured to execute many threads in parallel, where the term "thread" refers to an instance of a specific program executed on a specific set of input data. Alternatively, single instruction multiple data (SIMD) instruction issuance technology can be used to support the parallel execution of a large number of threads without providing multiple independent instruction units. Alternatively, single instruction multiple thread (SIMT) technology can be used to support the parallel execution of a large number of generally synchronized threads using a common instruction unit that is configured to issue instructions to a set of processing engines within each processing cluster in the processing cluster. Unlike the SIMD execution mechanism in which all processing engines generally execute the same instruction, SIMT execution allows different threads to more easily follow divergent execution paths through a given thread program. Those skilled in the art will understand that the SIMD processing mechanism represents a functional subset of the SIMT processing mechanism.

[0068] The operation of the processing cluster 214 may be controlled by means of a pipeline manager 232 that distributes processing tasks to SIMT parallel processors. The pipeline manager 232 receives data from Figure 2A The graphics multiprocessor 234 is an exemplary instance of a SIMT parallel processor. However, various types of SIMT parallel processors with different architectures may be included within the processing cluster 214. One or more instances of the graphics multiprocessor 234 may be included within the processing cluster 214. The graphics multiprocessor 234 may also be referred to as a streaming multiprocessor (SM) and may be capable of executing a large number of execution threads simultaneously.

[0069] The graphics multiprocessor 234 can process data, and the data crossbar 240 can be used to distribute the processed data to one of multiple possible destinations, including instances of the graphics multiprocessor 234 within the processing cluster 214. The pipeline manager 232 can facilitate the distribution of processed data by specifying the destination for the processed data to be distributed via the data crossbar 240. Each graphics multiprocessor 234 within the processing cluster 214 can include the same set of function execution logic (e.g., arithmetic logic unit, load-store unit, etc.). The function execution logic can be configured in a pipelined manner, where new instructions can be issued before previous instructions have completed. The function execution logic supports a wide variety of operations, including integer and floating-point arithmetic, comparison operations, Boolean operations, bit shifts, and the calculation of various algebraic functions. The same functional unit hardware can be utilized to perform different operations, and any combination of functional units can exist.

[0070] Instructions transmitted to processing cluster 214 constitute threads. A collection of threads executed across a collection of parallel processing engines is a thread group. Thread groups execute the same program on different input data. Each thread within a thread group can be assigned to a different processing engine within graphics multiprocessor 234. A thread group can include fewer threads than the number of processing engines within graphics multiprocessor 234. When a thread group includes fewer threads than the number of processing engines, one or more of the processing engines can be idle during cycles in which that thread group is being processed. A thread group can also include more threads than the number of processing engines within graphics multiprocessor 234. When a thread group includes more threads than the number of processing engines within graphics multiprocessor 234, processing can be performed during consecutive clock cycles. Optionally, multiple thread groups can be executed concurrently on graphics multiprocessor 234.

[0071] The graphics multiprocessor 234 may include an internal cache memory to perform load and store operations. Alternatively, the graphics multiprocessor 234 may forgo the internal cache memory and use the cache memory within the processing cluster 214 (e.g., level 1 (L1) cache 248). Each graphics multiprocessor 234 also has access to a partition unit (e.g., Figure 2A20N) is shared between all instances of processing cluster 214 and can be used to transfer data between threads. Graphics multiprocessor 234 can also access off-chip global memory, which can include one or more of local parallel processor memory and / or system memory. Any memory external to parallel processing unit 202 can be used as global memory. Embodiments in which processing cluster 214 includes multiple instances of graphics multiprocessor 234 can share common instructions and data, which can be stored in L1 cache 248.

[0072] Each processing cluster 214 may include an MMU 245 (memory management unit) configured to map virtual addresses into physical addresses. In other embodiments, one or more instances of the MMU 245 may reside in Figure 2A The MMU 245 includes a set of page table entries (PTEs) used to map virtual addresses to physical addresses of the slice and optionally includes a cache line index. The MMU 245 may include an address translation lookaside buffer (TLB) or cache that may reside within the graphics multiprocessor 234 or L1 cache 248 of the processing cluster 214. Physical addresses are processed to distribute data access locality, thereby allowing efficient request interleaving between partition units. The cache line index can be used to determine whether a request for a cache line is a hit or a miss.

[0073] In graphics and compute applications, the processing clusters 214 can be configured such that each graphics multiprocessor 234 is coupled to a texture unit 236 for performing texture mapping operations, such as determining texture sample locations, reading texture data, and filtering texture data. Texture data is read from an internal texture L1 cache (not shown), or in some embodiments, from an L1 cache within the graphics multiprocessor 234, and retrieved from an L2 cache, local parallel processor memory, or system memory as needed. Each graphics multiprocessor 234 outputs processed tasks to a data crossbar 240 to provide the processed tasks to another processing cluster 214 for further processing or to store the processed tasks in an L2 cache, local parallel processor memory, or system memory via a memory crossbar 216. A preROP 242 (pre-raster operation unit) is configured to receive data from the graphics multiprocessor 234, direct the data to a ROP unit, which can communicate with a partition unit (e.g., a partition unit) as described herein. Figure 2AThe preROP 242 unit may perform optimizations for color blending, organize pixel color data, and perform address translation.

[0074] It will be appreciated that the core architecture described herein is illustrative and that variations and modifications are possible. Any number of processing units (e.g., graphics multiprocessor 234, texture unit 236, preROP 242, etc.) may be included within processing cluster 214. Parallel processing units as described herein may include any number of instances of processing cluster 214. Optionally, each processing cluster 214 may be configured to operate independently of other instances of processing cluster 214 using, for example, separate and distinct processing units, L1 cache, L2 cache to facilitate data and fault isolation.

[0075] Figure 2D An example of a graphics multiprocessor 234 is shown, wherein the graphics multiprocessor 234 is coupled to the pipeline manager 232 of the processing cluster 214. The graphics multiprocessor 234 has an execution pipeline, including, but not limited to, an instruction cache 252, an instruction unit 254, an address mapping unit 256, a register file 258, one or more general-purpose graphics processing unit cores (GPGPU cores 262), and one or more load / store units 266. The GPGPU cores 262 and the load / store units 266 are coupled to a cache memory 272 and a shared memory 270 via a memory and cache interconnect 268. The graphics multiprocessor 234 may also include a ray tracing core 263, which includes hardware logic for accelerating ray tracing operations, and a tensor core 264, which includes hardware logic for accelerating tensor (e.g., matrix) operations. The instruction cache 252 may receive a stream of instructions to be executed from the pipeline manager 232. Instructions are cached in the instruction cache 252 and dispatched for execution by the instruction unit 254. The instruction unit 254 can dispatch instructions as thread groups (e.g., warps), where each thread in the thread group is assigned to a different execution unit within the GPGPU core 262. Instructions can access any of the local, shared, or global address spaces by specifying an address within the unified address space. An address mapping unit 256 can be used to convert addresses in the unified address space into different memory addresses that can be accessed by the load / store unit 266.

[0076] The register file 258 provides a collection of registers for the functional units of the graphics multiprocessor 234. The register file 258 provides temporary storage for operands for the data paths of the functional units connected to the graphics multiprocessor 234 (e.g., the GPGPU core 262, the load / store unit 266). The register file 258 can be divided among each of the functional units so that each functional unit is allocated a dedicated portion of the register file 258. For example, the register file 258 can be divided among different warps being executed by the graphics multiprocessor 234.

[0077] GPGPU cores 262 may each include a floating point unit (FPU) and / or an integer arithmetic logic unit (ALU) used to execute instructions of graphics multiprocessor 234. In some implementations, GPGPU cores 262 may include hardware logic that may otherwise reside within tensor core 264 and / or ray tracing core 263. GPGPU cores 262 may be architecturally similar or architecturally different. For example, and in one embodiment, a first portion of GPGPU core 262 includes a single-precision FPU and integer ALU, while a second portion of the GPGPU core includes a double-precision FPU. Optionally, the FPU may implement the IEEE 754-2008 standard for floating-point arithmetic or enable variable-precision floating-point arithmetic. In one embodiment, a single-precision FPU or a separate set of FPUs may be configured to perform operations on 16-bit floating-point operands, such as operands in half-precision format or bfloat16 format (e.g., brain floating point), which is a 16-bit floating-point format with one sign bit, eight exponent bits, and eight significand bits (seven of which are explicitly stored). The FPUs within one or more GPGPU cores 262 may also support one or more 8-bit floating-point formats. Supported 8-bit floating-point formats include the E4M3 format with a 4-bit exponent and a 3-bit mantissa, and the E5M2 format with a 5-bit exponent and a 2-bit mantissa. The graphics multiprocessor 234 may additionally include one or more fixed-function or special-function units for performing specific functions such as copying rectangles or pixel blending operations. One or more of the GPGPU cores may also include fixed-function or special-function logic.

[0078] The GPGPU core 262 may include SIMD logic capable of executing a single instruction on multiple sets of data. Optionally, the GPGPU core 262 may physically execute SIMD4, SIMD8, and SIMD16 instructions and logically execute SIMD1, SIMD2, and SIMD32 instructions. SIMD instructions for the GPGPU core may be generated by a shader compiler at compile time, or SIMD instructions for the GPGPU core may be automatically generated when executing a program written and compiled for a single program multiple data (SPMD) or SIMT architecture. Multiple threads of a program configured for a SIMT execution model may be executed with the aid of a single SIMD instruction. For example, and in one embodiment, eight SIMT threads may be executed in parallel as SIMD8 instructions, the eight SIMT threads performing the same or similar operations. In one embodiment, a warp of 32 SIMT threads may be executed as a single SIMD32 instruction. Warp divergence may be handled with the aid of multiple SIMD instructions.

[0079] The memory and cache interconnect 268 is an interconnect network that connects each of the functional units of the graphics multiprocessor 234 to the register file 258 and to the shared memory 270. For example, the memory and cache interconnect 268 is a crossbar interconnect that allows the load / store unit 266 to perform load and store operations between the shared memory 270 and the register file 258. The register file 258 can operate at the same frequency as the GPGPU core 262, so data transfers between the GPGPU core 262 and the register file 258 are very low latency. The shared memory 270 can be used to enable communication between threads executing on the functional units within the graphics multiprocessor 234. The shared memory 270 can also be used as a program-managed cache. The cache memory 272 can be used, for example, as an automatically managed data cache to cache texture data transferred between the functional units and the texture unit 236. Shared memory 270 and cache memory 272 can be coupled to data crossbar switch 240 to enable communication with other components of the processing cluster, thereby facilitating collaborative execution of cluster workgroups across multiple graphics multiprocessors within the processing cluster. Threads executing on GPGPU core 262 can programmatically store data in shared memory in addition to automatically cached data stored in cache memory 272. In one embodiment, shared memory 270 and cache memory 272 can be combined into a single configurable memory unit that can be selectively configured as either cache memory 272 or shared memory 270.

[0080] Figure 2E Explained relative to Figure 2D The graphics multiprocessor 234 has an alternative configuration of the graphics multiprocessor 235. The disclosure of any features in combination with the graphics multiprocessor 235 described herein also discloses the same. Figure 2D The corresponding combination of the graphics multiprocessor 234 is not limited to such a combination. Figure 2D Graphics multiprocessor 234, Figure 2E The graphics multiprocessor 235 includes multiple additional instances of execution resources 286A-286D. For example, the graphics multiprocessor 235 may include multiple instruction units 254A-254D, register files 258A-258D, and texture units 280A-280D. The graphics multiprocessor 235 also includes multiple sets of graphics or compute execution units (e.g., GPGPU cores 262A-262D, ray tracing cores 263A-263D, and tensor cores 264A-264D) and multiple sets of load / store units 266A-266D. The execution resources 286A-286D work in conjunction with the texture unit(s) 280A-280D for texture operations, while sharing the instruction cache 252, shared memory 270, and cache memories 272A-272B. In one embodiment, execution resources 286A-286D additionally include multi-function units (MUFUs) and / or special function units (SFUs) (e.g., MFUs 267A-267D) used to perform specialized mathematical operations such as transcendental operations including exponential, logarithmic, and trigonometric functions.

[0081] The various components can communicate via interconnect structure 290. Interconnect structure 290 may include one or more crossbar switches to enable communication between the various components of graphics multiprocessor 235. GPGPU cores 262A-262D, ray tracing cores 263A-263B, and tensor cores 264A-264D can each communicate with shared memory 270 via interconnect structure 290. Interconnect structure 290 can arbitrate communications within graphics multiprocessor 235 to ensure fair bandwidth allocation between components. In one embodiment, interconnect structure 290 can be an independent high-speed network fabric layer on which each component of graphics multiprocessor 235 is stacked. Components of graphics multiprocessor 235 can also communicate with remote components via interconnect structure 290.

[0082] In one embodiment, the graphics multiprocessor 235 includes a tensor transfer engine 292, which is a copy engine that can be configured to accelerate the movement of tensor data into and out of the graphics multiprocessor 235. The tensor transfer engine 292 can accelerate tensor memory operations by asynchronously performing address generation and data movement operations for N-dimensional blocks of tensor data, which offloads operations that would otherwise be performed manually by program code executed by the graphics multiprocessor 235. The tensor transfer engine 292 can be configured to copy data between, for example, the shared memory 270 and / or cache memories 272A-272B of the graphics multiprocessor 235 and memory external to the graphics multiprocessor 235, such as graphics processor global memory (e.g., parallel processor memory 222). In one embodiment, data transfers performed by the tensor transfer engine 292 can be configured to selectively bypass various levels of intermediate data storage between source and destination memory. For example, transfers between global memory and shared memory 270 can bypass register files 258A-258D. In one embodiment, threads can synchronize on asynchronous tensor transfers with the help of a non-blocking barrier synchronization mechanism.

[0083] In various embodiments, graphics multiprocessor 235 can be customized for specific use cases by including or excluding certain components, thereby allowing for various implementations of graphics multiprocessor 235 that are tailored to target power, performance, and area requirements. For example, a compute-oriented variant of graphics multiprocessor 235 that will not perform graphics operations can exclude ray tracing cores 263A-263D. A fully graphics-oriented variant can exclude tensor transfer engine 292, while a graphics-oriented variant that is additionally configured to accelerate neural network inference can include at least one version of tensor transfer engine 292.

[0084] Those skilled in the art will understand that Figure 1 and Figures 2A-2E The architecture described in is illustrative and not limiting with respect to the scope of the present embodiments. Thus, the techniques described herein may be implemented on any suitably configured processing unit, including but not limited to: one or more mobile application processors, one or more desktop or server central processing units (CPUs) (including multi-core CPUs), one or more parallel processing units (such as Figure 2A parallel processing unit 202), and one or more graphics processors or special processing units.

[0085] The parallel processor or GPGPU described herein can be communicatively coupled to a host / processor core to accelerate graphics operations, machine learning operations, pattern analysis operations, and various general-purpose GPU (GPGPU) functions. The GPU can be communicatively coupled to the host processor / core via a bus or another interconnect (e.g., a high-speed interconnect such as PCIe, NVLink, or other known protocols, standardized protocols, or proprietary protocols). In other embodiments, the GPU can be integrated on the same package or chip as the core, and the GPU can be communicatively coupled to the core via an internal processor bus / interconnect (i.e., inside the package or chip). Regardless of the manner in which the GPU is connected, the processor core can assign work to the GPU in the form of a sequence of commands / instructions contained in a work descriptor. The GPU then uses dedicated circuit modules / logic to efficiently process these commands / instructions.

[0086] Figure 3 A graphics processing unit (GPU 380) is illustrated that includes a dedicated collection of graphics processing resources arranged into multiple core groups 365A-365N. Multiple core groups 365A-365N and Figure 2D Graphics multiprocessor 234 or Figure 2E 234 or 235. While details are provided for a single example of multi-core groups 365A-365N (e.g., multi-core group 365A), it will be appreciated that other multi-core groups 365B-365N may be equipped with the same or similar collection of graphics processing resources. As described herein, the details described with respect to multi-core groups 365A-365N may also apply to either graphics multiprocessor 234 or graphics multiprocessor 235.

[0087] As illustrated, the multi-core group 365A may include a graphics core 370, a tensor core 371, and a ray tracing core 372. The graphics core 370 is similar to the GPGPU cores 262A-262D and may be configured to execute instructions to perform graphics and / or general computing operations. The scheduler / dispatcher 368 schedules and dispatches graphics threads for execution on the various cores within the multi-core group 365A. A register file 369 is included, which stores operand values ​​used by the core when performing graphics or general computing operations for the executed threads. These register files 369 may include, for example, registers that may be configured to store integer values ​​or floating-point values, including vector registers for storing packed integer and / or floating-point data elements and slice registers for storing tensor / matrix values. Slice registers may be implemented as multidimensional registers comprising a combined set of vector registers.

[0088] One or more combined level 1 (L1) caches and shared memory units 373 store graphics data, such as texture data, vertex data, pixel data, ray data, bounding volume data, etc., locally within each multi-core group 365A. One or more texture units 374 may also be used to perform texture operations, such as texture mapping and sampling. A level 2 (L2) cache 375, shared by all or a subset of the multi-core groups 365A-365N, stores graphics data and / or instructions for multiple concurrent graphics threads. As illustrated, the L2 cache 375 can be shared across multiple multi-core groups 365A-365N. One or more memory controllers 367 couple the GPU 380 to memory 366, which can be system memory (e.g., DRAM) and / or dedicated graphics memory (e.g., GDDR6 memory).

[0089] An input / output circuit module (I / O circuit module 363) couples GPU 380 to one or more I / O devices 362, such as a digital signal processor (DSP), a network controller, or a user input device. On-chip interconnects may be used to couple I / O devices 362 to GPU 380 and memory 366. At least one I / O memory management unit (IOMMU 364) of I / O circuit module 363 directly couples I / O devices 362 to system memory 366. Optionally, IOMMU 364 manages multiple sets of page tables used to map virtual addresses to physical addresses in memory 366. I / O devices 362, CPU(s) 361, and GPU 380 can then share the same virtual address space.

[0090] In one implementation of IOMMU 364, IOMMU 364 supports virtualization. In this case, it may manage a first set of page tables used to map guest / graphics virtual addresses to guest / graphics physical addresses and a second set of page tables used to map guest / graphics physical addresses to system / host physical addresses (e.g., within memory 366). The base address of each of the first and second sets of page tables may be stored in a control register and swapped out upon context switch (e.g., so that the new context is provided with access to the relevant set of page tables). Although not described in detail in the present disclosure, the first set of page tables may be used to map guest / graphics virtual addresses to guest / graphics physical addresses and a second set of page tables used to map guest / graphics physical addresses to system / host physical addresses (e.g., within memory 366). Figure 3 , but each of the cores within the multi-core group 365A-365N may include a translation lookaside buffer (TLB) for caching guest virtual to guest physical translations, guest physical to host physical translations, and guest virtual to host physical translations.

[0091] (One or more) CPU 361, GPU 380 and I / O device 362 can be integrated on a single semiconductor chip and / or chip package. Memory 366 can be integrated on the same chip or can be coupled to memory controller 367 via an off-chip interface. In one implementation, memory 366 includes GDDR6 memory that shares the same virtual address space as other physical system-level memory, although the basic principles described herein are not limited to this particular implementation.

[0092] Tensor Core 371 may include multiple execution units specifically designed to perform matrix operations, which are the basic computational operations used to perform deep learning operations. For example, simultaneous matrix multiplication operations can be used for neural network training and reasoning. Tensor Core 371 can perform matrix processing using a variety of operand precisions including single-precision floating point (e.g., 32-bit), half-precision floating point (e.g., 16-bit), integer word (16-bit), byte (8-bit), and half-byte (4-bit). For example, a neural network implementation extracts features of each rendered scene, potentially combining details from multiple frames to construct a high-quality final image.

[0093] In deep learning implementations, parallel matrix multiplication work can be scheduled for execution on the Tensor Core 371. Neural network training in particular requires a large number of matrix dot product operations. To handle the inner product formulation of an N×N×N matrix multiplication, the Tensor Core 371 can include at least N dot product processing elements. Before the matrix multiplication begins, a complete matrix is ​​loaded into the slice register, and for each of N cycles, at least one column of the second matrix is ​​loaded. Each cycle, there are N dot products processed.

[0094] Depending on the specific implementation, matrix elements can be stored with different precisions, including 16-bit words, 8-bit bytes (e.g., INT8), and 4-bit nibbles (e.g., INT4). Different precision modes can be specified for the tensor core 371 to ensure that the most efficient precision is used for different workloads (e.g., such as inference workloads that can tolerate quantization to bytes and nibbles). Supported formats also include 64-bit floating point (FP64) and non-IEEE floating point formats (such as bfloat16 format). One embodiment includes support for a reduced precision tensor-floating point (TF32) mode that uses the range of FP32 (8-bit) and the precision of FP16 (10-bit) to perform calculations. Reduced precision TF32 operations can be performed on FP32 inputs and produce FP32 outputs with higher performance relative to FP32 and increased precision relative to FP16. In one embodiment, one or more of 8-bit floating point (FP8) formats, 6-bit floating point (FP6) formats, and 4-bit floating point (FP4) formats are supported, including a floating point format denoted as the microscale (MX) format.

[0095] In one embodiment, the Tensor Core 371 supports a sparse operation mode for matrices in which the majority of values ​​are zero. The Tensor Core 371 includes support for sparse input matrices encoded in a sparse matrix representation (e.g., Coordinate List Encoding (COO), Compressed Sparse Row (CSR), Compressed Sparse Column (CSC), etc.). In cases where the sparse matrix representation can be further compressed, the Tensor Core 371 also includes support for compressed sparse matrix representations. Compressed, encoded, and / or compressed and encoded matrix data can be read by the Tensor Core 371 along with associated compression and / or encoding metadata, and non-zero values ​​can be extracted. For example, for a given input matrix A, non-zero values ​​can be loaded from a compressed and / or encoded representation of at least a portion of matrix A. Based on the position of the non-zero value in matrix A (which can be determined based on the index or coordinate metadata associated with the non-zero value), the corresponding value in the input matrix B can be loaded. Depending on the operation to be performed (e.g., multiplication), if the corresponding value is a zero value, the loading of the value from the input matrix B can be bypassed. In one embodiment, the pairing of values ​​for certain operations (such as multiplication operations) can be pre-scanned by the scheduler logic and only operations between non-zero inputs are scheduled. Depending on the dimensions of matrices A and B and the operations to be performed, the output matrix C can be dense or sparse. Where the output matrix C is sparse and depending on the configuration of the tensor core 371, the output matrix C can be output in a compressed format, sparse coding, or compressed sparse coding.

[0096] The ray tracing core 372 can accelerate ray tracing operations for both real-time ray tracing implementations and non-real-time ray tracing implementations. In particular, the ray tracing core 372 may include a ray traversal / intersection circuit module for performing ray traversals using a bounding volume hierarchy (BVH) and identifying intersections between primitives and rays enclosed within the BVH volume. The ray tracing core 372 may also include a circuit module for performing depth testing and culling (e.g., using a Z-buffer or similar arrangement). In one implementation, the ray tracing core 372 performs traversal and intersection operations in conjunction with the image denoising techniques described herein, at least a portion of which may be executed on the tensor core 371. For example, the tensor core 371 may implement a deep learning neural network to perform denoising on frames generated by the ray tracing core 372. However, the CPU(s) 361, the graphics core 370, and / or the ray tracing core 372 may also implement all or part of the denoising and / or deep learning algorithms.

[0097] Additionally, as described above, a distributed approach to denoising can be employed, wherein GPU 380 is located in a computing device coupled to other computing devices via a network or high-speed interconnect. In this distributed approach, the interconnected computing devices can share neural network learning / training data to increase the speed at which the entire system learns to perform denoising on different types of image frames and / or different graphics applications.

[0098] The ray tracing cores 372 can handle all BVH traversals and / or ray-primitive intersections, thereby freeing the graphics core 370 from being overloaded with thousands of instructions per ray. For example, each ray tracing core 372 includes a first set of specialized circuit modules for performing bounding box tests (e.g., for traversal operations) and / or a second set of specialized circuit modules for performing ray-triangle intersection tests (e.g., intersecting rays that have already been traversed). Thus, for example, the multi-core group 365A can simply start ray probing, and the ray tracing cores 372 independently perform ray traversals and intersections and return hit data (e.g., hits, misses, multiple hits, etc.) to the thread context. While the ray tracing cores 370 perform the traversal and intersection operations, the graphics core 370 and the tensor cores 371 are then freed to perform other graphics or computational work. Optionally, each ray tracing core 372 can include a traversal unit for performing BVH test operations and / or an intersection unit for performing ray-primitive intersection tests. The intersection unit generates a "hit," "miss," or "multiple hits" response that it provides to the appropriate thread. During the traversal and intersection operations, execution resources of other cores (e.g., graphics core 370 and tensor core 371) are freed up to perform other forms of graphics work. In an alternative embodiment described below, a hybrid rasterization / ray tracing approach is used in which rendering operations are distributed between graphics core 370 and ray tracing core 372.

[0099] The ray tracing core 372 may include hardware support for a ray tracing instruction set such as Microsoft's DirectX Ray Tracing (DXR) including the DispatchRays command and ray generation shaders, nearest hit shaders, any hit shaders, and miss shaders that enable assigning a set of shaders and textures to each object. Another ray tracing platform that may be supported by the ray tracing core 372, graphics core 370, and tensor core 371 is the Vulkan API (e.g., Vulkan version 1.1.85 or later). However, it is noted that the basic principles described herein are not limited to any particular ray tracing ISA. In general, the ray tracing core 372, tensor core 371, and graphics core 370 may support a ray tracing instruction set that includes instructions / functions for one or more of the following: ray generation, nearest hit, any hit, ray-primitive intersection, per-primitive and hierarchy bounding box construction, misses, visits, and exceptions. More specifically, preferred embodiments include ray tracing instructions for performing one or more of the following functions:

[0100] Light Generation - Ray generation instructions can be executed per pixel, sample, or other user-defined work assignments.

[0101] Recent Hits - A nearest hit instruction can be executed to locate the closest intersection of a ray and a primitive within the scene.

[0102] Any hit -Any hit instruction identifies multiple intersections between rays and primitives within the scene to potentially identify a new closest intersection point.

[0103] intersect The -intersect command performs a ray-primitive intersection test and outputs the result.

[0104] Per-primitive bounding box construction - This command builds a bounding box around a given primitive or group of primitives (e.g., when building a new BVH or other acceleration data structure).

[0105] miss - Indicates that the ray missed the scene or all geometry within a specified region of the scene.

[0106] visit - Indicates the subvolume that the ray will traverse.

[0107] abnormal - Includes various types of exception handlers (e.g., called for various error conditions).

[0108] In one embodiment, the ray tracing core 372 may be adapted to accelerate general computational operations using computational techniques similar to ray intersection testing. A computational framework may be provided that enables shader programs to be compiled into low-level instructions and / or primitives that perform general computational operations with the aid of the ray tracing core. Exemplary computational problems that may benefit from computational operations performed on the ray tracing core 372 include computations involving the propagation of beams, waves, rays, or particles within a coordinate space. Interactions associated with that propagation may be computed relative to geometric shapes or meshes within the coordinate space. For example, computations associated with the propagation of electromagnetic signals through an environment may be accelerated using instructions or primitives executed with the aid of the ray tracing core. Diffraction and reflection of the signal by objects in the environment may be computed as direct ray tracing simulations.

[0109] The ray tracing core 372 can also be used to perform calculations that are not directly similar to ray tracing. For example, the ray tracing core 372 can be used to accelerate mesh projection, mesh refinement, and volume sampling calculations. General coordinate space calculations such as nearest neighbor calculations can also be performed. For example, a set of points near a given point can be found by defining a bounding box around the point in coordinate space. The BVH and ray detection logic within the ray tracing core 372 can then be used to determine the set of intersections of the points within the bounding box. The intersection constitutes the origin and the nearest neighbors of that origin. The calculations performed using the ray tracing core 372 can be performed in parallel with the calculations performed on the graphics core 370 and the tensor core 371. The shader compiler can be configured to compile a compute shader or other general graphics processing program into low-level primitives that can be parallelized across the graphics core 370, the tensor core 371, and the ray tracing core 372. Technologies for GPU to host processor interconnect

[0110] Figure 4A The plurality of GPUs 410-413 (e.g., such as Figure 2A ) is communicatively coupled to a plurality of multi-core processors 405-406 via high-speed links 440A-440D (e.g., buses, point-to-point interconnects, etc.). Depending on the implementation, the high-speed links 440A-440D may support 4 GB / s, 30 GB / s, 80 GB / s, or higher communication throughput. Various interconnect protocols may be used, including but not limited to PCIe 4.0, PCIe 5.0, PCIe 6.0, and various NVLink and NVLink-C2C (chip-to-chip) interconnect protocols (e.g., NVLink v5). However, the underlying principles described herein are not limited to any particular communication protocol or throughput.

[0111] Two or more of the GPUs 410-413 may be interconnected via high-speed links 442A-442B, which may be implemented using the same or different protocols / links as those used for high-speed links 440A-440D. Similarly, two or more of the multi-core processors 405-406 may be connected via high-speed link 443, which may be a symmetric multiprocessor (SMP) bus operating at 20 GB / s, 30 GB / s, 120 GB / s, or lower or higher speeds. Alternatively, the same protocols / links may be implemented (e.g., via a common interconnect fabric). Figure 4A All communications between the various system components shown in FIG. However, as mentioned, the basic principles described herein are not limited to any particular type of interconnection technology.

[0112] Each of the multi-core processors 405 and 406 can be communicatively coupled to processor memories 401-402 via memory interconnects 430A-430B, respectively, and each GPU 410-413 can be communicatively coupled to GPU memories 420-423 via GPU memory interconnects 450A-450D, respectively. Memory interconnects 430A-430B and 450A-450D can utilize the same or different memory access technologies. By way of example and not limitation, processor memories 401-402 and GPU memories 420-423 can be volatile memories such as dynamic random access memory (DRAM) (including stacked DRAM), graphics DDR SDRAM (GDDR) (e.g., GDDR5, GDDR6), or high bandwidth memory (HBM), and / or can be non-volatile memories such as 3D Xpoint / Optane or Nano-Ram. For example, a portion of the memory can be volatile memory and another portion can be non-volatile memory (e.g., using a two-level memory (2LM) hierarchy). The memory subsystem as described herein is compatible with a variety of memory technologies such as the double data rate versions promulgated by JEDEC (Joint Electron Device Engineering Council).

[0113] As described below, although the various processors 405-406 and GPUs 410-413 may be physically coupled to specific processor memories 401-402 and GPU memories 420-423, respectively, a unified memory architecture may be implemented in which the same virtual system address space (also referred to as an "effective address" space) is distributed across all of the various physical memories. For example, the processor memories 401-402 may each include 64GB of system memory address space, and the GPUs 420-423 may each include 32GB of system memory address space (resulting in a total of 256GB of addressable memory in this example).

[0114] Figure 4BAdditional optional details of the interconnection between processor 407 and graphics accelerator 446 are illustrated. Graphics accelerator 446 may include one or more GPU chips integrated on a line card that is coupled to processor 407 via high-speed link 440. Alternatively, graphics accelerator 446 may be integrated on the same package or chip as processor 407. Processor 407 includes multiple cores 460A-460D, each having a translation lookaside buffer 461A-461D and one or more caches 462A-462D. The cores may include various other components for executing instructions and processing data, which are not illustrated to avoid obscuring the basic principles of the components described herein (e.g., instruction fetch units, branch prediction units, decoders, execution units, reorder buffers, etc.). Caches 462A-462D may include a level 1 (L1) cache and a level 2 (L2) cache. In addition, one or more shared caches 456 may be included in the cache hierarchy and shared by the set of cores 460A-460D. For example, one embodiment of processor 407 includes 24 cores, each with its own L1 cache, twelve shared L2 caches, and twelve shared L3 caches. In this embodiment, one of the L2 cache and the L3 cache is shared by two adjacent cores. Processor 407 is connected to system memory 441, which may include processor memories 401-402.

[0115] Coherence is maintained for data and instructions stored in the various caches 462A-462D, the shared cache(s) 456, and system memory 441 via inter-core communication via a coherence bus 464. For example, each cache may have associated cache coherence logic / circuitry modules that communicate via the coherence bus 464 in response to detected reads or writes to a particular cache line. In one implementation, a cache snooping protocol is implemented via the coherence bus 464 to snoop cache accesses. Cache snooping / coherence techniques are well understood by those skilled in the art and will not be described in detail herein to avoid obscuring the underlying principles described herein. Proxy circuitry 425 may be provided that communicatively couples the graphics accelerator 446 to the coherence bus 464, thereby allowing the graphics accelerator 446 to participate in the cache coherence protocol as a peer of a core. In particular, an interface 435 provides connectivity to the proxy circuitry 425 via a high-speed link 440 (e.g., a PCIe bus, NVLink, etc.), and an interface 437 connects the graphics accelerator 446 to the high-speed link 440.

[0116] In one implementation, interface 437 is coupled to accelerator integrated circuit 436, which provides cache management, memory access, context management, and interrupt management services on behalf of graphics processing engines 431, 432, ..., N of graphics accelerator 446. Graphics processing engines 431, 432, ..., N can each include a separate graphics processing unit (GPU). Alternatively, graphics processing engines 431, 432, ..., N can include different types of graphics processing engines within a GPU, such as a graphics execution unit, a media processing engine (e.g., a video encoder / decoder), a sampler, and a block image transfer (BLIT) engine. In other words, the graphics accelerator can include graphics processing engines 431-432, ..., N of a single GPU, or graphics processing engines 431-432, ..., N can be associated with multiple GPUs integrated on a common package, line card, or chip. Graphics processing engines 431-432, ..., N can be configured with any graphics processor or computing accelerator architecture described herein. The work to be performed by graphics processing engines 431 , 432 may be specified by means of work descriptors that provide an indication of the work to be done by graphics accelerator 446 .

[0117] The accelerator integrated circuit 436 may include a memory management unit (MMU) 439 for performing various memory management functions such as virtual to physical memory translation (also known as effective to real memory translation) and a memory access protocol for accessing system memory 441. The MMU 439 may also include a translation lookaside buffer (TLB) (not shown) for caching virtual / effective to physical / real address translations. In one implementation, the cache 438 stores commands and data for efficient access by the graphics processing engines 431, 432, ..., N. Data stored in the cache 438 and graphics memories 433-434, ..., M may be kept consistent with the core caches 462A-462D, the shared cache(s) 456, and the system memory 441. As mentioned, this may be accomplished with the aid of proxy circuitry 425 that participates in cache coherence mechanisms on behalf of cache 438 and graphics memory 433-434, ..., M (e.g., sending updates to cache 438 relating to modifications / accesses of cache lines on processor caches 462A-462D, shared cache(s) 456, and receiving updates from cache 438).

[0118] Registers 445 store context data for threads executed by graphics processing engines 431-432, ..., N, and context management circuitry 448 manages thread contexts. For example, context management circuitry 448 can perform save and restore operations to save and restore the contexts of various threads during context switches (e.g., where a first thread is saved and a second thread is restored so that the second thread can be executed by the graphics processing engine). For example, upon context switching, context management circuitry 448 can store current register values ​​to a designated area in memory (e.g., identified by a context pointer). It can then restore the register values ​​when returning to the context. Interrupt management circuitry 447 can, for example, receive and process interrupts received from system devices.

[0119] In one implementation, the virtual / effective address from the graphics processing engine is converted into a real / physical address in the system memory 441 by the MMU 439. Optionally, the accelerator integrated circuit 436 supports multiple (e.g., 4, 8, 16) graphics accelerators 446 and / or other accelerator devices. The graphics accelerator 446 can be dedicated to a single application executed on the processor 407, or the graphics accelerator 446 can be shared between multiple applications. Optionally, a virtualized graphics execution environment is provided in which the resources of the graphics processing engines 431-432, ..., N are shared with multiple applications, virtual machines (VMs), or containers. Resources can be subdivided into "slices" that are allocated to different VMs and / or applications based on the processing requirements and priorities associated with the VMs and / or applications or based on a predetermined partition profile for the graphics accelerator 446. VM and container can be used interchangeably herein.

[0120] A virtual machine (VM) can be software that runs an operating system and one or more applications. A VM can be defined by specifications, configuration files, virtual disk files, non-volatile random access memory (NVRAM) settings files, and log files, and is supported by the physical resources of the host computing platform. A VM can include an operating system (OS) or application environment installed on software that emulates dedicated hardware. The experience that end users have on a virtual machine is the same as they would have on dedicated hardware. Specialized software called a hypervisor completely emulates the CPU, memory, hard disk, network, and other hardware resources of a PC client or server, allowing virtual machines to share resources. A hypervisor can emulate multiple virtual hardware platforms that are isolated from each other, allowing virtual machines to run on the same underlying physical host. Servers, VMware ESXi and other operating systems.

[0121] A container is a package of applications, configurations, and dependencies so that the application runs reliably from one computing environment to another. Containers can share the operating system installed on a server platform and run as isolated processes. A container is a package of software that contains everything the software needs to run, such as system tools, libraries, and settings. Containers are not installed like traditional software programs, which allows them to be isolated from other software and the operating system itself. The isolated nature of containers provides several benefits. First, the software in the container will run the same in different environments. For example, a container containing PHP and MySQL can be installed in Computers and Second, containers offer increased security because the software will not affect the host operating system. While installed applications can change system settings and modify resources (such as the Windows Registry), containers can only modify settings within the container.

[0122] Thus, the accelerator integrated circuit 436 acts as a bridge to the system for the graphics accelerator 446 and provides address translation and system memory caching services. In one embodiment, to facilitate the bridge functionality, the accelerator integrated circuit 436 may also include shared I / O 497 (e.g., PCIe, USB, or other) and hardware to enable system control of voltage, clocking, performance, thermal, and security. The shared I / O 497 may utilize a separate physical connection or may traverse the high-speed link 440. In addition, the accelerator integrated circuit 436 may provide virtualization facilities for the host processor to manage virtualization of the graphics processing engine, interrupts, and memory management.

[0123] Because the hardware resources of graphics processing engines 431-432, ..., N are explicitly mapped to the real address space seen by processor 407, any host processor can directly address these resources using effective address values. An optional feature of accelerator integrated circuit 436 is the physical separation of graphics processing engines 431-432, ..., N, so that they appear to the system as independent units. In one embodiment, accelerator integrated circuit 436 includes a security circuit module 444 that enables configurable cryptographic isolation of data associated with each slice of resources. Different slices can be associated with different security domains, so that different cryptographic keys are used to encrypt data associated with various security domains. In one embodiment, the security domains of graphics accelerator 446 can be integrated into a trusted execution environment supported by processor 407. In one embodiment, secure I / O capabilities can be enabled that allow each security domain to be presented as a separate trusted I / O device, which enables support for trusted DMA and MMIO operations.

[0124] One or more graphics memories 433-434, ..., M may be respectively coupled to each of the graphics processing engines 431-432, ..., N. The graphics memories 433-434, ..., M store instructions and data being processed by each of the graphics processing engines 431-432, ..., N. The graphics memories 433-434, ..., M may be volatile memories such as DRAM (including stacked DRAM), GDDR memories (e.g., GDDR5, GDDR6), or HBM, and / or may be non-volatile memories such as 3D Xpoint / Optane, Samsung Z-NAND, or Nano-Ram.

[0125] To reduce data traffic on high-speed link 440, biasing techniques may be used to ensure that the data stored in graphics memories 433-434, ..., M is the data that will be used most frequently by graphics processing engines 431-432, ..., N and preferably not used (at least less frequently) by cores 460A-460D. Similarly, the biasing mechanism attempts to keep data needed by the cores (and preferably not graphics processing engines 431-432, ..., N) within caches 462A-462D, shared cache(s) 456, and system memory 441.

[0126] In an alternative variation, an accelerator integrated circuit 436 is integrated within the processor 407, and the graphics processing engines 431-432, ..., N communicate to the accelerator integrated circuit 436 over a high-speed link 440, via an interface 437 and an interface 435 (which again may utilize any form of bus, fabric, or interface protocol). In this variation, the accelerator integrated circuit 436 performs the same operations as those described above.

[0127] The described embodiments can support different programming models, including a dedicated process programming model (without graphics accelerator virtualization) and a shared programming model (with virtualization). The latter can include programming models controlled by accelerator integrated circuit 436 and programming models controlled by graphics accelerator 446. In a dedicated process model embodiment, graphics processing engines 431, 432, ..., N can be dedicated to a single application or process under a single operating system. A single application can aggregate other application requests to graphics engines 431, 432, ..., N, thereby providing virtualization within a VM / partition. In a dedicated process programming model, graphics processing engines 431, 432, ..., N can be shared by multiple VMs / application partitions. The shared model requires a hypervisor to virtualize graphics processing engines 431-432, ..., N to allow access by each operating system. For a single-partition system without a hypervisor, graphics processing engines 431-432, ..., N are owned by the operating system. In both cases, the operating system can virtualize graphics processing engines 431-432, ..., N to provide access to each process or application. For a shared programming model, graphics accelerator 446 or individual graphics processing engines 431-432, ..., N use process handles to select process elements. Process elements can be stored in system memory 441 and can be addressable using the effective address to real address translation techniques described herein. The process handle can be an implementation-specific value provided to the host process when registering its context with the graphics processing engines 431-432, ..., N. The host process can perform this by calling system software to add the process element to a process element linked list. The lower 16 bits of the process handle can be the offset of the process element within the process element linked list.

[0128] Figure 4C An accelerator integrated slice 490 is illustrated. As used herein, a "slice" comprises a designated portion of the processing resources of an accelerator integrated circuit 436. The address space 482 of the application within the system memory 441 stores a process element 483. The process element 483 may be stored in response to a GPU call 481 from an application 480 executing on the processor 407. The process element 483 contains the process state of the application 480. The work descriptor (WD 484) contained in the process element 483 may be a single job requested by the application or may contain a pointer to a queue of jobs. In the latter case, the WD 484 is a pointer to a job request queue in the address space 482 of the application.

[0129] Graphics accelerator 446 and / or individual graphics processing engines 431-432, ..., N can be shared by all or a subset of processes in the system. For example, the techniques described herein may include infrastructure for establishing process state and sending WD 484 to graphics accelerator 446 to start a job in a virtualized environment.

[0130] In one implementation, a dedicated process programming model is implementation-specific. In this model, a single process owns either the graphics accelerator 446 or a separate graphics processing engine 431. Because the graphics accelerator 446 is owned by a single process, when the graphics accelerator 446 is assigned, the hypervisor initializes the accelerator integrated circuit 436 for the owning partition and the operating system initializes the accelerator integrated circuit 436 for the owning process.

[0131] In operation, a fetch unit 491 in the accelerator integrated slice 490 fetches a WD 484 to be processed. WD 484 includes an indication of work to be performed by a graphics processing engine of the graphics accelerator 446. As illustrated, data from WD 484 may be stored in registers 445 and may be used by the MMU 439, interrupt management circuitry 447, and / or context management circuitry 448. For example, the MMU 439 may include a segment / page walk circuit module for accessing segment tables / page tables 486 within the OS virtual address space 485. The interrupt management circuitry 447 may process interrupt events 492 received from the graphics accelerator 446. When executing graphics operations, effective addresses 493 generated by the graphics processing engines 431-432, ..., N are converted into real addresses by the MMU 439.

[0132] Registers 445 may be replicated for each graphics processing engine 431-432, ..., N and / or graphics accelerator 446 and may be initialized by a hypervisor or operating system. Each of these replicated registers may be included in an accelerator integration slice 490. In one embodiment, each graphics processing engine 431-432, ..., N may be presented to a hypervisor 496 as a distinct graphics processor device. Quality of service (QoS) settings may be configured for clients of a particular graphics processing engine 431-432, ..., N. Cryptographic and physical data isolation between clients of each engine may be enabled by means of isolated memory access paths and automatic data encryption for each client. Exemplary registers that may be initialized by a hypervisor are shown in Table 1. Table 1 - Registers initialized by the hypervisor 1 Slice Control Register 2 Real address (RA) dispatched process area pointer 3 Permission Mask Override Register 4 Interrupt vector table entry offset 5 Interrupt vector table entry limit 6 Status Register 7 Logical partition ID 8 Real Address (RA) hypervisor accelerator utilizing record pointers 9 Storage Description Register

[0133] Example registers that may be initialized by the operating system are shown in Table 2. Table 2 - Registers initialized by the operating system

[0134] Each WD 484 may be specific to a particular graphics accelerator 446 and / or graphics processing engine 431-432, ..., N. It contains all the information the graphics processing engine requires to do its work, or it may be a pointer to a memory location in a command queue where an application has set up work to be done.

[0135] Figure 4D Additional optional details of the sharing model are illustrated. It includes a hypervisor real address space 498 in which a process element list 499 is stored. The hypervisor real address space 498 is accessible via a hypervisor 496 that virtualizes the graphics processing engine to the operating system 495.

[0136] The shared programming model allows for all or a subset of processes from all or a subset of partitions in the system to use the graphics accelerator 446. There are two programming models in which the graphics accelerator 446 is shared by multiple processes and partitions: time-sliced ​​sharing and graphics-directed sharing.

[0137] In this model, hypervisor 496 owns graphics accelerator 446 and makes its functionality available to all operating systems 495. In order for graphics accelerator 446 to support virtualization by hypervisor 496, graphics accelerator 446 must adhere to the following requirements: 1) Application job requests must be autonomous (i.e., no state needs to be maintained between jobs), or graphics accelerator 446 must provide a context save and restore mechanism. 2) Application job requests must be guaranteed by graphics accelerator 446 to complete within a specified amount of time, including any transition failures, or graphics accelerator 446 must provide the ability to preempt the processing of jobs. 3) Graphics accelerator 446 must be guaranteed fairness between processes when operating in a directed-share programming model.

[0138] For the directed sharing model, application 480 may be required to make an operating system 495 system call using a graphics accelerator 446 type, a work descriptor (WD), an authority mask register (AMR) value, and a context save / restore region pointer (CSRP). The graphics accelerator 446 type describes the targeted acceleration functionality for the system call. The graphics accelerator 446 type can be a system-specific value. The WD is formatted specifically for the graphics accelerator 446 and can be in the form of a graphics accelerator 446 command, an effective address pointer to a user-defined structure, an effective address pointer to a command queue, or any other data structure used to describe the work to be performed by the graphics accelerator 446. In one embodiment, the AMR value is the AMR state to be used for the current process. The value is passed to the operating system similarly to an application setting an AMR. If the accelerator integrated circuit 436 and graphics accelerator 446 implementation do not support the User Authority Mask Override Register (UAMOR), the operating system may apply the current UAMOR value to the AMR value before passing the AMR in the hypervisor call. The hypervisor 496 may optionally apply the current permission mask override register (AMOR) value before placing the AMR into the process element 483. The CSRP may be one of the registers 445 that contains the effective address of an area in the application's address space 482 for the graphics accelerator 446 to save and restore context state. This pointer is optional if no state is required to be saved between jobs or when a job is preempted. The context save / restore area may be pinned system memory.

[0139] Upon receiving the system call, operating system 495 can verify that application 480 has been registered and has been given permission to use graphics accelerator 446. Operating system 492 then calls hypervisor 496 with the information shown in Table 3. Table 3 - OS to Hypervisor Call Parameters

[0140] Upon receiving the hypervisor call, the hypervisor 496 verifies that the operating system 495 has registered and has been given permission to use the graphics accelerator 446. The hypervisor 496 then places the process element 483 in the process element linked list for the corresponding type of graphics accelerator 446. The process element may include the information shown in Table 4. Table 4 - Process element information 1 Work Descriptor (WD) 2 The authority mask register (AMR) value (potentially masked). 3 Effective Address (EA) Context Save / Restore Region Pointer (CSRP) 4 Process ID (PID) and optional thread ID (TID) 5 Virtual Address (VA) Accelerator Utilization Record Pointer (AURP) 6 Virtual address of the storage segment table pointer (SSTP) 7 Logical Interrupt Service Number (LISN) 8 Interrupt vector table, derived from the hypervisor call parameters. 9 Status register (SR) value 10 Logical Partition ID (LPID) 11 Real Address (RA) hypervisor accelerator utilizing record pointers 12 Storage Descriptor Register (SDR)

[0141] The hypervisor may initialize registers 445 of the plurality of accelerator integrated slices 490 .

[0142] like Figure 4E As described in , in an optional implementation, a unified memory addressable via a common virtual memory address space used to access physical processor memories 401-402 and GPU memories 420-423 is employed. In this implementation, operations executed on GPUs 410-413 utilize the same virtual / effective memory address space to access processor memories 401-402 and vice versa, thereby simplifying programmability. A first portion of the virtual / effective address space can be allocated to processor memory 401, a second portion can be allocated to second processor memory 402, a third portion can be allocated to GPU memory 420, and so on. The entire virtual / effective memory space (sometimes referred to as the effective address space) can thus be distributed across each of processor memories 401-402 and GPU memories 420-423, allowing any processor or GPU to access any physical memory using a virtual address mapped to that memory.

[0143] Bias / coherency management circuit modules 494A-494E within one or more of the MMUs 439A-439E may be provided that ensure cache coherency between the caches of the host processor (e.g., multi-core processor 405) and the caches of the GPUs 410-413 and implement biasing techniques that indicate the physical memory where certain types of data should be stored. Figure 4E Multiple instances of bias / coherence management circuit modules 494A-494E are illustrated in FIG, but bias / coherence management circuit modules may be implemented within an MMU of one or more host processors and / or within an accelerator integrated circuit 436.

[0144] The GPU-attached memory 420-423 can be mapped as part of the system memory and can be accessed using shared virtual memory (SVM) technology, but without suffering the typical performance drawbacks associated with full system cache coherence. The ability of the GPU-attached memory 420-423 to be accessed as system memory without the heavy cache coherence overhead provides a beneficial operating environment for GPU offloading. This arrangement allows host processor software to set operands and access computation results without the overhead of traditional I / O DMA data copying. Such traditional copying involves driver calls, interrupts, and memory-mapped I / O (MMIO) accesses, all of which are inefficient relative to simple memory accesses. At the same time, the ability to access the GPU-attached memory 420-423 without cache coherence overhead can be critical to the execution time of the offloaded computation. For example, in situations with a large amount of streaming write memory traffic, cache coherence overhead can significantly reduce the effective write bandwidth seen by the GPUs 410-413. The efficiency of operand setup, the efficiency of result access, and the efficiency of GPU computation all play a role in determining the effectiveness of GPU offloading.

[0145] The selection between GPU bias and host processor bias can be driven by a bias tracker data structure. For example, a bias table can be used, which can be a page-granular structure (i.e., controlled at the granularity of a memory page) that includes 1 or 2 bits per GPU-attached memory page. With or without a bias cache in the GPUs 410-413 (e.g., to cache frequently / recently used entries of the bias table), the bias table can be implemented in a stolen memory range of one or more GPU-attached memories 420-423. Alternatively, the entire bias table can be maintained within the GPU.

[0146] In one implementation, the bias table entry associated with each access to the GPU-attached memory 420-423 is accessed before the actual access to the GPU memory, resulting in the following operations. First, local requests from the GPU 410-413 for pages that find their location in the GPU bias are forwarded directly to the corresponding GPU memory 420-423. Local requests from the GPU for pages that find their location in the host bias are forwarded to the processor (e.g., as discussed above, over a high-speed link). Optionally, requests from the processor for pages that find their location in the host processor bias complete the request like normal memory reads. Alternatively, requests directed to pages in the GPU bias can be forwarded to the GPU 410-413. If the GPU is not currently using the page, the GPU can then transfer the page to the host processor bias. The bias state of a page can be changed either by a software-based mechanism, a hardware-assisted software-based mechanism, or by a purely hardware-based mechanism for a limited set of situations. One mechanism for changing the bias state employs an API call (e.g., OpenCL), which in turn calls the GPU's device driver, which in turn sends a message (or queues a command descriptor) to the GPU directing it to change the bias state and, for some transitions, perform a cache flush operation in the host. The cache flush operation is required for transitions from host processor bias to GPU bias, but not for the reverse transition.

[0147] Cache coherency can be maintained by temporarily rendering GPU-biased pages that are not cacheable by the host processor. To access these pages, the host processor may request access from GPU 410, which may or may not grant immediate access, depending on the implementation. Therefore, to reduce communication between the host processor and GPU 410, it is beneficial to ensure that GPU-biased pages are those required by the GPU but not utilized by the host processor, and vice versa. Graphics processing pipeline

[0148] Figure 5 A graphics processing pipeline 500 is illustrated. Figure 2D Graphics multiprocessor 234 or Figure 2E The graphics multiprocessor 235 of the graphics multiprocessor may implement the graphics processing pipeline 500. The graphics multiprocessor may be included in a graphics processing system such as the graphics processing system described herein. Figure 2A In the parallel processing subsystem of the parallel processor 200, the parallel processing subsystem can be Figure 1 The parallel processor(s) 112 are related and may be used to replace one of those. Various parallel processor systems may be implemented with parallel processing units (e.g. Figure 2A The graphics processing pipeline 500 may be implemented using one or more instances of the parallel processing unit 202 of FIG. For example, a shader unit (e.g., Figure 2C The graphics multiprocessor 234 of FIG may be configured to perform the functions of one or more of the following: a vertex processing unit 504, a tessellation control processing unit 508, a tessellation evaluation processing unit 512, a geometry processing unit 516, and a fragment / pixel processing unit 524. Figure 2A Other processing engines within the processing cluster 214) and corresponding partition units (e.g. Figure 2A The functions of the data assembler 502, the primitive assemblers 506, 514, 518, the tessellation unit 510, the rasterizer 522, and the raster operation unit 526 may be performed by the partitioning units 220A-220N of the graphics processing pipeline 500. The graphics processing pipeline 500 may also be implemented using dedicated processing units for one or more functions. It is also possible that one or more portions of the graphics processing pipeline 500 are executed by parallel processing logic within a general-purpose processor (e.g., a CPU). Optionally, one or more portions of the graphics processing pipeline 500 may access on-chip memory (e.g., memory interface 528) via a memory interface 528. Figure 2A The parallel processor memory 222 in the memory interface 528 may be Figure 2A An example of a memory interface 218 can also be used. Figure 3 The multi-core group 365A in the GPU is used to implement the graphics processor pipeline 500.

[0149] The data assembler 502 is a processing unit that can collect vertex data for surfaces and primitives. The data assembler 502 then outputs the vertex data, including vertex attributes, to the vertex processing unit 504. The vertex processing unit 504 is a programmable execution unit that executes the vertex shader program, thereby lighting and transforming the vertex data as specified by the vertex shader program. The vertex processing unit 504 reads data stored in cache, local or system memory for use in processing vertex data and can be programmed to transform vertex data from an object-based coordinate representation to a world space coordinate space or a normalized device coordinate space.

[0150] A first instance of primitive assembler 506 receives vertex attributes from vertex processing unit 504. Primitive assembler 506 reads the stored vertex attributes as needed and constructs graphics primitives for processing by tessellation control processing unit 508. Graphics primitives include triangles, line segments, points, patches, etc. as supported by various graphics processing application programming interfaces (APIs).

[0151] The tessellation control processing unit 508 treats the input vertices as control points of a geometry patch. The control points are transformed from an input representation from the patch (e.g., a basis for the patch) into a representation suitable for use in surface estimation by the tessellation evaluation processing unit 512. The tessellation control processing unit 508 may also calculate tessellation factors for the edges of the geometry patch. The tessellation factors are applied to individual edges and quantify the view-dependent level of detail associated with the edge. The tessellation unit 510 is configured to receive the tessellation factors for the edges of the patch and tessellate the patch into a plurality of geometric primitives, such as lines, triangles, or quadrilaterals, which are passed to the tessellation evaluation processing unit 512. The tessellation evaluation processing unit 512 operates on the parameterized coordinates of the tessellated patch to generate a surface representation and vertex attributes for each vertex associated with the geometric primitive.

[0152] A second instance of primitive assembler 514 receives vertex attributes from tessellation evaluation processing unit 512, reads stored vertex attributes as needed, and builds graphics primitives for processing by geometry processing unit 516. Geometry processing unit 516 is a programmable execution unit that executes geometry shader programs to transform the graphics primitives received from primitive assembler 514 as specified by the geometry shader programs. Geometry processing unit 516 can be programmed to tessellate the graphics primitives into one or more new graphics primitives and calculate parameters used to rasterize the new graphics primitives.

[0153] The geometry processing unit 516 may be capable of adding or removing elements from the geometry stream. The geometry processing unit 516 outputs parameters and vertices specifying new graphics primitives to the primitive assembler 518. The primitive assembler 518 receives the parameters and vertices from the geometry processing unit 516 and constructs the graphics primitives for processing by the viewport scaling, culling, and clipping unit 520. The geometry processing unit 516 reads data stored in parallel processor memory or system memory for use in processing the geometry data. The viewport scaling, culling, and clipping unit 520 performs clipping, culling, and viewport scaling and outputs the processed graphics primitives to the rasterizer 522.

[0154] The rasterizer 522 can perform depth culling and other depth-based optimizations. The rasterizer 522 also performs scan conversion on new graphics primitives to generate fragments and outputs those fragments and associated coverage data to the fragment / pixel processing unit 524. The fragment / pixel processing unit 524 is a programmable execution unit configured to execute a fragment shader program or a pixel shader program. The fragment / pixel processing unit 524 transforms the fragments or pixels received from the rasterizer 522 as specified by the fragment or pixel shader program. For example, the fragment / pixel processing unit 524 can be programmed to perform operations including, but not limited to, texture mapping, shading, blending, texture correction, and perspective correction to produce shaded fragments or pixels that are output to the raster operation unit 526. The fragment / pixel processing unit 524 can read data stored in either parallel processor memory or system memory for use in processing the fragment data. The fragment or pixel shader program can be configured to shade at a sample, pixel, slice, or other granularity depending on the sampling rate configured for the processing unit.

[0155] Raster operations unit 526 is a processing unit that performs raster operations including but not limited to stenciling, z-testing, blending, etc. and outputs pixel data to be stored in graphics memory (e.g., Figure 2A The parallel processor memory 222 and / or Figure 1 The raster operations unit 526 may be configured to compress z or color data written to memory and decompress z or color data read from memory. Machine Learning Overview

[0156] The architecture described above can be applied to perform training and reasoning operations using machine learning models. Machine learning has been successful in solving many types of tasks. The calculations generated when training and using machine learning algorithms (e.g., neural networks) are naturally suitable for efficient parallel implementation. Accordingly, parallel processors such as general-purpose graphics processing units (GPGPUs) have played an important role in the actual implementation of deep neural networks. Parallel graphics processors with single instruction multiple threads (SIMT) architectures are designed to maximize the amount of parallel processing in the graphics pipeline. In the SIMT architecture, groups of parallel threads attempt to execute program instructions together synchronously as frequently as possible to improve processing efficiency. The efficiency provided by the parallel machine learning algorithm implementation allows the use of high-capacity networks and enables those networks to be trained on larger data sets.

[0157] Machine learning algorithms are algorithms that can learn from a collection of data. For example, machine learning algorithms can be designed to model high-level abstractions within a collection of data. For example, image recognition algorithms can be used to determine which of several categories a given input belongs to; regression algorithms can output a numerical value given an input; and pattern recognition algorithms can be used to generate converted text or perform text-to-speech and / or speech recognition.

[0158] An exemplary type of machine learning algorithm is a neural network. There are many types of neural networks; a simple type of neural network is a feedforward network. A feedforward network can be implemented as an acyclic graph in which nodes are arranged in layers. Typically, a feedforward network topology includes an input layer and an output layer separated by at least one hidden layer. The hidden layer transforms the input received by the input layer into a representation useful for generating output in the output layer. The nodes of the network are fully connected to the nodes in the adjacent layers via edges, but there are no edges between the nodes within each layer. Data received at the nodes of the input layer of the feedforward network is propagated (i.e., "fed forward") to the nodes of the output layer using an activation function that calculates the state of the nodes of each successive layer in the network based on coefficients ("weights") associated with each of the edges in the connecting layers. The output from the neural network algorithm can take various forms, depending on the specific model being represented by the algorithm being executed.

[0159] Before a machine learning algorithm can be used to model a particular problem, the algorithm is trained using a training data set. Training a neural network involves: selecting a network topology; using a set of training data that represents the problem being modeled by the network; and adjusting weights until the network model performs with minimum error for all instances of the training data set. For example, during a supervised learning training process for a neural network, the output generated by the network in response to an input representing an instance in the training data set is compared to the "correct" labeled output for that instance, an error signal representing the difference between the output and the labeled output is calculated, and as the error signal is propagated backward through the layers of the network, the weights associated with the connections are adjusted to minimize that error. The network is considered "trained" when the error for each of the outputs generated from the instances of the training data set is minimized.

[0160] The accuracy of a machine learning algorithm can be significantly affected by the quality of the data set used to train the algorithm. The training process can be computationally intensive and can require a significant amount of time on conventional general-purpose processors. Accordingly, many types of machine learning algorithms are trained using parallel processing hardware. This is particularly useful for optimizing the training of neural networks, as the calculations performed when adjusting the coefficients in a neural network are naturally suited to parallel implementation. In particular, many machine learning algorithms and software applications have been adapted to take advantage of the parallel processing hardware within general-purpose graphics processing devices.

[0161] Figure 6 is a generalized diagram of a machine learning software stack 600. A machine learning application 602 is any logic that can be configured to train a neural network using a training data set or to implement machine intelligence using a trained deep neural network. The machine learning application 602 may include training and inference functionality for the neural network and / or specialized software that can be used to train the neural network prior to deployment. The machine learning application 602 may implement any type of machine intelligence, including but not limited to: image recognition, mapping and localization, autonomous navigation, speech synthesis, medical imaging, or language translation. Example machine learning applications 602 include but are not limited to voice-based virtual assistants, image or facial recognition algorithms, autonomous navigation, and software tools used to train the machine learning models used by the machine learning application 602.

[0162] Hardware acceleration for machine learning applications 602 can be enabled via a machine learning framework 604. The machine learning framework 604 can provide a library of machine learning primitives. Machine learning primitives are basic operations typically performed by machine learning algorithms. Without the machine learning framework 604, developers of machine learning algorithms would be required to create and optimize the main computing logic associated with the machine learning algorithm, and then re-optimize the computing logic when developing new parallel processors. Instead, machine learning applications can be configured to use the primitives provided by the machine learning framework 604 to perform necessary calculations. Exemplary primitives include tensor convolution, activation functions, and pooling, which are computing operations performed when training convolutional neural networks (CNNs). The machine learning framework 604 can also provide primitives used to implement basic linear algebra subroutines performed by many machine learning algorithms, such as matrix and vector operations. Examples of machine learning frameworks 604 include, but are not limited to, TensorFlow, TensorRT, PyTorch, MXNet, Caffee, and other advanced machine learning frameworks.

[0163] The machine learning framework 604 can process the input data received from the machine learning application 602 and generate appropriate input to the computing framework 606. The computing framework 606 can abstract the underlying instructions provided to the GPGPU driver 608 so that the machine learning framework 604 can take advantage of hardware acceleration with the GPGPU hardware 610 without requiring the machine learning framework 604 to be very familiar with the architecture of the GPGPU hardware 610. In addition, the computing framework 606 can enable hardware acceleration for the machine learning framework 604 across a wide variety of types and generations of GPGPU hardware 610. In one example, the computing framework 606 can include the CUDA computing framework and associated machine learning libraries, such as the CUDA Deep Neural Network (cuDNN) library. The machine learning software stack 600 can also include a communication library or framework to facilitate multi-GPU and multi-node computing. GPGPU machine learning acceleration

[0164] Figure 7 A general purpose graphics processing unit (GPGPU 700) is illustrated, which may be Figure 2A Parallel processor 200 or Figure 1 The general processing unit can be configured to provide hardware acceleration support for primitives provided by machine learning frameworks to accelerate the processing of the type of computational workload associated with training deep neural networks. In addition, the GPGPU 700 can be directly linked to other instances of the GPGPU to create a multi-GPU cluster, thereby increasing the training speed of deep neural networks in particular. Primitives are also supported to accelerate inference operations for deployed neural networks.

[0165] GPGPU 700 includes a host interface 702 for enabling connection to a host processor. Host interface 702 may be a PCI Express interface. However, the host interface may also be a vendor-specific communication interface or communication structure. GPGPU 700 receives commands from the host processor and uses a global scheduler 704 to distribute execution threads associated with those commands to a collection of processing clusters 706A-706H. The processing clusters 706A-706H share a cache memory 708. The cache memory 708 may act as a higher level cache for the cache memory within the processing clusters 706A-706H. The illustrated processing clusters 706A-706H may be used with Figure 2A Corresponding to the processing clusters 214A-214N in.

[0166] GPGPU 700 includes memory 714A-714B coupled to processing clusters 706A-706H via a set of memory controllers 712A-712B. Memories 714A-714B may include various types of memory devices, including dynamic random access memory (DRAM) or graphics random access memory, such as synchronous graphics random access memory (SGRAM), including graphics double data rate (GDDR) memory. Memories 714A-714B may also include 3D stacked memory, including but not limited to high bandwidth memory (HBM).

[0167] Each of processing clusters 706A-706H may include a collection of graphics multiprocessors (such as Figure 2D Graphics multiprocessor 234, Figure 2E Graphics multiprocessor 325), or may include Figure 3 The graphics multiprocessors of the computing cluster include multiple types of integer and floating-point logic units that can perform computational operations within a range of precision suitable for machine learning computations. For example, at least a subset of the floating-point units in each of the processing clusters 706A-706H can be configured to perform 16-bit or 32-bit floating-point operations, while a different subset of the floating-point units can be configured to perform 64-bit floating-point operations.

[0168] Multiple instances of GPGPU 700 can be configured to operate as a computing cluster. The communication mechanisms used by the computing cluster for synchronization and data exchange vary across embodiments. For example, multiple instances of GPGPU 700 communicate via a host interface 702. In one embodiment, GPGPU 700 includes an I / O hub 709 that couples GPGPU 710 to a GPU link 710, which enables direct connections to other instances of the GPGPU. GPU link 710 can be coupled to a dedicated GPU-to-GPU bridge that enables communication and synchronization between multiple instances of GPGPU 700. Optionally, GPU link 710 is coupled to a high-speed interconnect to transmit and receive data to and from other GPGPUs or parallel processors. Multiple instances of GPGPU 700 can be located in separate data processing systems and communicate via a network device accessible via host interface 702. In addition to or as an alternative to host interface 702 , GPU link 710 may also be configured to enable connection to a host processor.

[0169] While the illustrated configuration of GPGPU 700 can be configured to train neural networks, alternative configurations of GPGPU 700 can be configured for deployment within a high-performance or low-power inference platform. In the inference configuration, GPGPU 700 includes fewer of processing clusters 706A-706H relative to the training configuration. Additionally, the memory technology associated with memory 714A-714B can differ between the inference configuration and the training configuration. In one embodiment, the inference configuration of GPGPU 700 can support inference-specific instructions. For example, the inference configuration can provide support for one or more 8-bit integer or floating point dot product instructions commonly used during inference operations for deployed neural networks.

[0170] Figure 8 A multi-GPU computing system 800 is illustrated. The multi-GPU computing system 800 may include a processor 802 coupled to a plurality of GPGPUs 806A-806D via a host interface switch 804. The host interface switch 804 may be a PCI Express switching device that couples the processor 802 to a PCI Express bus, over which the processor 802 may communicate with the set of GPGPUs 806A-806D. Each of the plurality of GPGPUs 806A-806D may be Figure 7 GPGPU 700. GPGPUs 806A-806D may be interconnected via a collection of high-speed point-to-point GPU-to-GPU links 816 (P2P GPU links 816). High-speed GPU-to-GPU links may be connected via dedicated GPU links such as Figure 7 P2P GPU link 816 enables direct communication between each of GPGPUs 806A-806D without requiring communication over the host interface bus to which processor 802 is connected. With GPU-to-GPU traffic directed to the P2P GPU link, the host interface bus remains available for system memory access or communication with other instances of multi-GPU computing system 800, for example, via one or more network devices. Figure 8 GPGPUs 806A-806D are connected to processor 802 via host interface switch 804, but processor 802 may alternatively include direct support for P2P GPU link 816 and connect directly to GPGPUs 806A-806D. In one embodiment, P2P GPU link 816 enables multi-GPU computing system 800 to operate as a single logical GPU. Machine Learning Neural Network Implementation

[0171] The computing architecture described herein can be configured to perform a type of parallel processing that is particularly well-suited for training and deploying neural networks for machine learning. A neural network can be generalized as a network of functions having a graph relationship. As is well known in the art, there are a wide variety of types of neural network implementations used in machine learning. One exemplary type of neural network is a feedforward network as previously described. A second exemplary type of neural network is a convolutional neural network (CNN), while a third exemplary type of neural network is a recurrent neural network (RNN).

[0172] CNNs are specialized feedforward neural networks for processing data with a known, grid-like topology, such as image data. Accordingly, CNNs are commonly used in computational vision and image recognition applications, but they can also be used for other types of pattern recognition, such as speech and language processing. The nodes in the input layer of a CNN are organized into a collection of "filters" (feature detectors inspired by receptive fields similar to those found in the retina), and the output of each collection of filters is propagated to nodes in successive layers of the network. The computations used in a CNN involve applying the mathematical operation of convolution to each filter to produce the output of that filter. Convolution is a specialized mathematical operation that is performed on two functions to produce a third function that is a modified version of one of the two original functions. In convolutional network terminology, the first function for the convolution can be called the input, while the second function can be called the convolution kernel. The output can be called a feature map. For example, the input to a convolutional layer can be a multidimensional array of data defining the various color components of the input image. The convolution kernel can be a multidimensional array of parameters, where the parameters are adapted through the training process used for the neural network.

[0173] An RNN is a series of feedforward neural networks that include feedback connections between layers. RNNs enable modeling of sequential data by sharing parameter data across different parts of the neural network. The architecture for an RNN includes loops that represent the effect of the current value of a variable on its own value at future times, as at least a portion of the output data from the RNN is used as feedback for processing subsequent inputs in the sequence. This feature makes RNNs particularly useful for language processing due to the mutable nature of the language data that can be composed therein.

[0174] The figures described below present exemplary feedforward networks, CNN networks, and RNN networks, and describe the general process for training and deploying each of those types of networks, respectively. It will be understood that these descriptions are exemplary and non-limiting with respect to any specific embodiment described herein and that the concepts described are generally applicable to deep neural networks and machine learning techniques.

[0175] The deep neural networks used in deep learning typically include a front-end network for performing feature recognition, which is coupled to a back-end network that represents a mathematical model that can perform operations (such as object classification, speech recognition, etc.) based on the feature representation provided to the mathematical model. Deep learning enables machine learning to be performed without requiring manual feature engineering to be performed on the model. Instead, deep neural networks can learn features based on statistical structures or correlations within the input data. The learned features can be provided to a mathematical model that can map the detected features to outputs. The mathematical model used by the network is typically dedicated to the specific task to be performed, and different models will be used to perform different tasks.

[0176] Once the neural network is structured, a learning model can be applied to the network to train the network to perform a specific task. The learning model describes how to adjust the weights within the model to reduce the output error of the network. Backpropagation of errors is a common method used to train neural networks. An input vector is presented to the network for processing. The output of the network is compared with the desired output using a loss function, and an error value is calculated for each neuron in the output layer. The error values ​​are then propagated backward until each neuron has an associated error value that roughly represents its contribution to the original output. The network can then use an algorithm (such as a stochastic gradient descent algorithm) to learn from those errors to update the weights of the neural network.

[0177] Figures 9A-9B An exemplary convolutional neural network is illustrated. Figure 9A The various layers within a CNN are described. Figure 9A As shown in , an exemplary CNN for modeling image processing can receive an input 902 describing the red, green, and blue (RGB) components of an input image. The input 902 can be processed by multiple convolutional layers (e.g., a first convolutional layer 904, a second convolutional layer 906). The outputs from the multiple convolutional layers can optionally be processed by a fully connected layer 908. As previously described for feedforward networks, the neurons in a fully connected layer have full connections to all activations in the previous layer. The outputs from the fully connected layer 908 can be used to generate output results from the network. Matrix multiplication can be used instead of convolution to calculate the activations within the fully connected layer 908. Not all CNN implementations utilize a fully connected layer 908. For example, in some implementations, the second convolutional layer 906 can generate outputs for the CNN.

[0178] The convolutional layers are sparsely connected, unlike traditional neural network configurations found in fully connected layers 908. Traditional neural network layers are fully connected so that every output unit interacts with every input unit. However, as illustrated, the convolutional layers are sparsely connected because the output of the convolution of the field is the input to the nodes of the subsequent layer (rather than the corresponding state value of each node among the nodes in the field). The kernel associated with the convolutional layer performs a convolution operation, the output of which is sent to the next layer. The dimensionality reduction performed within the convolutional layer is one aspect that enables CNNs to scale to process large images.

[0179] Figure 9B 1 illustrates exemplary computational stages within a convolutional layer of a CNN. Input to a convolutional layer 912 of the CNN may be processed in three stages of a convolutional layer 914. The three stages may include a convolution stage 916, a detector stage 918, and a pooling stage 920. The convolutional layer 914 may then output data to a subsequent convolutional layer. The final convolutional layer of the network may generate output feature map data or provide input to a fully connected layer, for example, to generate a classification value for input to the CNN.

[0180] Several convolutions are performed in parallel in the convolution stage 916 to produce a set of linear activations. The convolution stage 916 may include an affine transformation, which is any transformation that can be specified as a linear transformation plus a translation. Affine transformations include rotations, translations, scaling, and combinations of these transformations. The convolution stage calculates the output of a function (e.g., a neuron) connected to a specific region in the input, which can be determined as a local region associated with the neuron. The neuron calculates the dot product between the weight of the neuron and the weight of the region in the local input to which the neuron is connected. The output from the convolution stage 916 defines a set of linear activations processed by successive stages of the convolution layer 914.

[0181] The linear activations may be processed by the detector stage 918. In the detector stage 918, each linear activation is processed by a nonlinear activation function. The nonlinear activation function increases the nonlinear nature of the overall network without affecting the receptive field of the convolutional layer. Several types of nonlinear activation functions may be used. One particular type is the rectified linear unit (ReLU), which uses an activation function defined as f(x) = max(0, x) such that the activation is thresholded at zero.

[0182] The pooling stage 920 uses a pooling function that replaces the output of the second convolutional layer 906 with summary statistics of nearby outputs. The pooling function can be used to introduce translation invariance into the neural network so that small translations to the input do not change the pooled output. The invariance of local translation can be useful in scenarios where the exact location of the feature in the input data is more important. Various types of pooling functions can be used during the pooling stage 920, including maximum pooling, average pooling, and l2-norm pooling. In addition, some CNN implementations do not include a pooling stage. Instead, such an implementation replaces an additional convolution stage with an increased stride relative to the previous convolution stage.

[0183] The output from the convolutional layer 914 may then be processed by the next layer 922. The next layer 922 may be an additional convolutional layer or one of the fully connected layers 908. For example, Figure 9A The first convolutional layer 904 may output to the second convolutional layer 906 , and the second convolutional layer may output to the first layer in the fully connected layer 908 .

[0184] A variant on CNN is the convolutional deep belief network, which has a structure similar to that of a CNN and is trained in a manner similar to that of a deep belief network. A deep belief network (DBN) is a generative neural network composed of multiple layers of uncertain (random) variables. A DBN can be trained layer by layer using greedy unsupervised learning. The learned weights of the DBN can then be used to provide a pre-trained neural network by determining the optimal initial set of weights for the neural network.

[0185] Figures 10A-10B An exemplary language model is described. Figure 10AA recurrent neural network (RNN 1000) is illustrated. In a recurrent neural network (RNN), the previous state of the network influences the output of the current state of the network. RNNs can be built in a variety of ways and using a variety of functions. The use of RNNs generally revolves around using mathematical models to predict the future based on previous sequences of inputs. For example, RNNs can be used to perform statistical language modeling to predict upcoming words given a previous sequence of words. RNN 1000 can be described as having an input layer 1002 that receives an input vector, a hidden layer 1004 that implements a recurrent function, a feedback mechanism 1005 that enables 'memory' of previous states, and an output layer 1006 that outputs a result. RNN 1000 operates based on time steps. The state of the RNN at a given time step is influenced by the previous time step via the feedback mechanism 1005. For a given time step, the state of the hidden layer 1004 is defined by the previous state and the input at the current time step. An initial input (x1) at the first time step can be processed by the hidden layer 1004. The state information determined during the processing of the initial input (x1) can be used by the hidden layer 1004 to process the second input (x2). A given state can be calculated as s t =f(Ux t +Ws t-1 ), where U and W are parameter matrices. The function f is typically nonlinear, such as a variant of the hyperbolic tangent function (Tanh) or a modified function f(x) = max(0, x). However, the specific mathematical function used in the hidden layer 1004 may vary depending on the specific implementation details of the RNN 1000. This may also enable acceleration of changes on the RNN network. An example RNN variant is the long short-term memory (LSTM) RNN. LSTM RNNs are capable of learning long-term dependencies that may be necessary for processing longer language sequences.

[0186] Figure 10B The baseline components of the Transformer model 1010 are described. The Transformer model 1010 addresses issues found in RNN models with long input sequences and enables increased parallelization. The Transformer model 1010 and its variants are used to build language models (LLMs) that perform various tasks such as machine translation, automatic summarization, and dialogue management. The Transformer model 1010 can also be configured to perform image generation tasks such as text-to-image generation.

[0187] The Transformer model 1010 includes multiple instances of an encoder 1016 and a decoder 1026. The encoders are stacked end-to-end, with the output of the final encoder being routed as input to the multi-head attention layer of each decoder 1026. The input to the bottom-most instance of the encoder 1016 is processed by the input embedding 1012, which converts the input token into a vector that can be processed by the encoder 1016. The output dictionary is vectorized by the output embedding 1022 before entering the bottom-most instance of the decoder 1026. Positional encodings 1014 and 1024 are added to the input and output vectors at the bottom of the encoder 1016 and decoder 1026 stack. The positional encodings 1014 and 1024 inject information about the relative or absolute position of the token in the sequence of tokens to be processed, as the Transformer model 1010 does not naturally encode the order of tokens.

[0188] The encoder 1016 of the transformer model 1010 analyzes the input text and creates multiple hidden states that preserve the context and meaning of the text data. The encoder 1016 layer forms part of the core of the transformer architecture, although decoder-only 1026 variants of the transformer model 1010 are also possible. The encoder 1016 includes two sublayers: a multi-head attention (MHA) sublayer and a feed-forward network (FFN) sublayer. The MHA sublayer performs multiple concurrent self-attention operations to calculate attention scores, which enables the transformer model 1010 to measure the importance and relative relationships of different tokens in the input sequence in a context-aware manner. The FFN sublayer is a fully connected feed-forward neural network that is position-by-position. The output of each sublayer is processed by an addition and normalization (A&N) operation defined as LayerNorm(x+Sublayer(x)), where the output of the sublayer is added to the input of the sublayer and normalized. In some implementations, the normalization operation can be performed before the sublayer rather than after.

[0189] Decoder 1026 includes three sublayers: a masked MHA sublayer, an MHA sublayer, and an FFN sublayer. The masked MHA sublayer is similar to the MHA layer, except that masking is applied to prevent query positions from focusing on keywords in future positions. The MHA sublayer of decoder 1026 is similar to the MHA sublayer of the encoder, with additional input from encoder 1016. The FFN of decoder 1026 is identical to the FFN of encoder 1016. The linear and softmax blocks take the output of the last instance of decoder 1026 in the decoder stack and generate a probability distribution representing the output probabilities.

[0190] GPU acceleration can also be used for variants of the transformer model 1010 that replace some or all FFN sublayers with sparse mixture of experts (MoE) layers. Each MoE layer includes multiple experts, where each expert is a neural network. The MoE layer can itself be an FFN or can also be an MoE, allowing for hierarchical MoE layers.

[0191] Training of the transformer model 1010 can be optimized with the use of adaptive precision logic that adjusts the precision of calculations applied during training. The adaptive precision logic can attempt to use the smallest possible data types during training without significantly reducing training accuracy. For example, to minimize data loss due to the use of 16-bit, 8-bit, and 4-bit floating point formats, dynamic scaling and casting can be applied during training based on statistical analysis of the tensor data generated during training. A variety of analysis techniques can be used to statistically analyze the tensor data generated during training to determine a set of scaling factors to be applied to data blocks, such as inputs to each layer of the transformer model 1010. For example, absolute minimum and / or maximum values ​​can be used to determine scaling factors that can prevent underflow or overflow of low-precision floating point data types.

[0192] Figure 11 The training and deployment of deep neural networks is described. Once a given network has been structured for a task, the neural network is trained using a training dataset 1102. A training framework 1104 has been developed to enable hardware acceleration of the training process. For example, Figure 6 The machine learning framework 604 can be configured as a training framework 1104. The training framework 1104 can access an untrained neural network 1106 and enable the untrained neural network to be trained using the parallel processing resources described herein to generate a trained neural network 1108. To begin the training process, initial weights can be randomly selected or selected by pre-training using a deep belief network. The training cycle is then performed in either a supervised or unsupervised manner.

[0193] Supervised learning is a learning method in which training is performed as a mediated operation, such as when a training dataset 1102 includes inputs paired with expected outputs for the inputs, or when the training dataset includes inputs with known outputs and the outputs of the neural network are manually graded. The network processes the inputs and compares the resulting outputs to a set of expected or desired outputs. Errors are then propagated back through the system. The training framework 1104 can be adjusted to adjust the weights controlling the untrained neural network 1106. The training framework 1104 can provide tools to monitor how well the untrained neural network 1106 is converging on a model suitable for generating the correct answer based on the known input data. The training process occurs iteratively as the network's weights are adjusted to improve the outputs generated by the neural network. The training process can continue until the neural network reaches a statistically expected accuracy associated with the trained neural network 1108. The trained neural network 1108 can then be deployed to implement any number of machine learning operations to generate inference results 1114 based on the input of new data 1112.

[0194] Unsupervised learning is a learning method in which a network attempts to train itself using unlabeled data. Thus, for unsupervised learning, the training data set 1102 will include input data without any associated output data. The untrained neural network 1106 can learn groupings within the unlabeled inputs and can determine how individual inputs relate to the entire data set. Unsupervised training can be used to generate self-organizing maps, which are a type of trained neural network 1108 that can perform operations useful in reducing the dimensionality of data. Unsupervised training can also be used to perform anomaly detection, which allows for the identification of data points in the input data set that deviate from the normal pattern of the data.

[0195] Variations of supervised and unsupervised training can also be employed. Semi-supervised learning is a technique in which a mixture of labeled and unlabeled data of the same distribution is included in the training dataset 1102. Incremental learning is a variant of supervised learning in which input data is continuously used to further train the model. Incremental learning enables a trained neural network 1108 to adapt to new data 1112 without forgetting the knowledge instilled into the network during initial training. Whether supervised or unsupervised, the training process for particularly deep neural networks can be too computationally intensive for a single computing node. Instead of using a single computing node, a distributed network of computing nodes can be used to accelerate the training process.

[0196] Figure 12Ais a block diagram illustrating distributed learning. Distributed learning is a training model that uses multiple distributed computing nodes to perform supervised or unsupervised training of a neural network. Each of the distributed computing nodes may include one or more host processors and one or more general processing nodes, such as Figure 7 As illustrated, distributed learning can be performed using model parallelism 1202, data parallelism 1204, or a combination of model and data parallelism 1206.

[0197] In model parallelism 1202, different computing nodes in a distributed system can perform training computations for different parts of a single network. For example, each layer of a neural network can be trained by a different processing node of the distributed system. Benefits of model parallelism include the ability to scale to extremely large models. Splitting the computations associated with different layers of a neural network enables training very large neural networks where the weights of all layers would not fit in the memory of a single node. In some instances, model parallelism can be particularly useful when performing unsupervised training of large neural networks. Some implementations of model parallelism 1202 may also be referred to as tensor parallelism.

[0198] In data parallelism 1204, different nodes of the distributed network have complete instances of the model, and each node receives a different portion of the data. The results from the different nodes are then combined. Although different approaches to data parallelism are possible, data parallel training methods all require techniques for combining results and synchronizing model parameters between each node. Exemplary methods for combining data include parameter averaging and update-based data parallelism. Parameter averaging trains each node on a subset of the training data and sets global parameters (e.g., weights, biases) to the average of the parameters from each node. Parameter averaging uses a central parameter server that maintains parameter data. Update-based data parallelism is similar to parameter averaging, except that instead of transmitting parameters from the nodes to the parameter server, updates to the model are transmitted. In addition, update-based data parallelism can be performed in a decentralized manner, where updates are compressed and transmitted between nodes.

[0199] Combined model and data parallelism 1206 can be implemented, for example, in a distributed system where each compute node includes multiple GPUs. Combined model and data parallelism 1206 can also be referred to as hybrid parallelism. Each node can have a complete instance of a model, with separate GPUs within each node being used to train different parts of the model. Distributed training already increases overhead relative to training on a single machine. However, the parallel processors and GPGPUs described herein can each implement various techniques to reduce the overhead of distributed training, including techniques for enabling high-bandwidth GPU-to-GPU data transfer and accelerated remote data synchronization. Pipeline parallelism is a variation of combined model and data parallelism 1206, in which different nodes contain less than the entire model, but more than a single layer of the model. In pipeline parallelism, different groups of layers or sub-models are distributed across different processing nodes. Another variation is expert parallelism, which routes requests for specific experts within the model to different GPUs. Expert parallelism can be used, for example, with MoE transformer models.

[0200] Figure 12B 1 is a block diagram illustrating a programmable network interface 1210 and a data processing unit. The programmable network interface 1210 is a programmable network engine that can be used to accelerate network-based computing tasks within a distributed environment. The programmable network interface 1210 can be coupled to a host system via a host interface 1270. The programmable network interface 1210 can be used to accelerate network or storage operations for a CPU or GPU of a host system. For example, Figure 12A As shown in , the host system can be, for example, a node of a distributed learning system used to perform distributed training. The host system can also be a data center node in a data center.

[0201] In one embodiment, access to remote storage devices containing model data can be accelerated by the programmable network interface 1210. For example, the programmable network interface 1210 can be configured to present the remote storage device as a local storage device to the host system. The programmable network interface 1210 can also accelerate remote direct memory access (RDMA) operations performed between the GPU of the host system and the GPU of the remote system. In one embodiment, the programmable network interface 1210 can enable storage functionality such as, but not limited to, NVME-oF. The programmable network interface 1210 can also accelerate encryption, data integrity, compression, and other operations for the remote storage device on behalf of the host system, thereby allowing the remote storage device to approach the latency of a storage device directly attached to the host system.

[0202] The programmable network interface 1210 can also perform resource allocation and management on behalf of the host system. Storage security operations can be offloaded to the programmable network interface 1210 and can be performed in conjunction with the allocation and management of remote storage resources. Network-based operations for managing access to remote storage devices that would otherwise be performed by the host system's processor can alternatively be performed by the programmable network interface 1210.

[0203] In one embodiment, network and / or data security operations can be offloaded from the host system to the programmable network interface 1210. Data center security policies for data center nodes can be handled by the programmable network interface 1210 rather than by the host system's processor. For example, the programmable network interface 1210 can detect and mitigate attempted network-based attacks (e.g., DDoS) on the host system, thereby preventing the attack from compromising the availability of the host system.

[0204] The programmable network interface 1210 may include a system on chip (SoC 1220) that executes an operating system with the aid of multiple processor cores 1222. The processor core 1222 may include a general-purpose processor (e.g., a CPU) core. In one embodiment, the processor core 1222 may also include one or more GPU cores. The SoC 1220 may execute instructions stored in the memory device 1240. The storage device 1250 may store local operating system data. The storage device 1250 and the memory device 1240 may also be used to cache remote data for the host system. Network ports 1260A-1260B enable connection to a network or structure and facilitate network access to the SoC 1220 and network access to the host system via the host interface 1270. The programmable network interface 1210 may also include an I / O interface 1275, such as a USB interface. The I / O interface 1275 may be used to couple an external device to the programmable network interface 1210 or may be used as a debug interface. Programmable network interface 1210 also includes a management interface 1230 that enables software on a host device to manage and configure programmable network interface 1210 and / or SoC 1220. In one embodiment, programmable network interface 1210 may also include one or more accelerators or GPUs 1245 to offload parallel computing tasks from SoC 1220, a host system, or a remote system coupled via network ports 1260A-1260B. Example Machine Learning Applications

[0205] Machine learning can be applied to solve a wide variety of technical problems, including but not limited to computer vision, autonomous driving and navigation, speech recognition, and language processing. Computer vision has traditionally been one of the most active research areas for machine learning applications. The application range of computer vision ranges from reproducing human visual capabilities (such as recognizing faces) to creating new categories of visual capabilities. For example, a computer vision application can be configured to identify sound waves from vibrations induced in objects visible in a video. Parallel processor-accelerated machine learning enables the use of significantly larger training data sets than previously feasible to train computer vision applications and enables the use of low-power parallel processors to deploy inference systems.

[0206] Parallel processor-accelerated machine learning has autonomous driving applications, including lane and road sign recognition, obstacle avoidance, navigation, and driving control. Accelerated machine learning techniques can be used to train driving models based on datasets that define appropriate responses to specific training inputs. The parallel processors described herein can enable rapid training of increasingly complex neural networks for autonomous driving solutions and enable the deployment of low-power inference processors in mobile platforms suitable for integration into autonomous vehicles.

[0207] Parallel processor-accelerated deep neural networks have enabled machine learning approaches to automatic speech recognition (ASR). ASR involves creating a function that computes the most likely speech sequence given a sequence of input sounds. Accelerated machine learning using deep neural networks has enabled the replacement of hidden Markov models (HMMs) and Gaussian mixture models (GMMs) previously used for ASR.

[0208] Parallel processor-accelerated machine learning can also be used to accelerate natural language processing. The automated learning process can utilize statistical inference algorithms to generate models that are robust to erroneous or unfamiliar inputs. Exemplary natural language processor applications include automatic machine translation between human languages.

[0209] Parallel processing platforms for machine learning can be divided into training platforms and deployment platforms. Training platforms are typically highly parallel and include optimizations to accelerate multi-GPU single-node training and multi-node multi-GPU training. Exemplary parallel processors suitable for training include Figure 7 GPGPU 700 and Figure 8 800. In contrast, deployed machine learning platforms typically include lower-power parallel processors suitable for use in products such as cameras, autonomous robots, and autonomous vehicles.

[0210] In addition, machine learning techniques can be applied to accelerate or enhance graphics processing activities. For example, a machine learning model can be trained to recognize the output generated by a GPU-accelerated application and generate an upgraded version of that output. Such techniques can be applied to accelerate the generation of high-resolution images for gaming applications. Various other graphics pipeline activities can benefit from the use of machine learning. For example, a machine learning model can be trained to perform tessellation operations on geometric data to increase the complexity of the geometric model, thereby allowing fine-detailed geometric graphics to be automatically generated from relatively low-detailed geometric graphics.

[0211] Figure 13 An exemplary inference system-on-chip (SOC 1300) suitable for performing inference using a trained model is illustrated. SOC 1300 can integrate processing components including a media processor 1302, a vision processor 1304, a GPGPU 1306, and a multi-core processor 1308. GPGPU 1306 can be a GPGPU as described herein (such as GPGPU 700), and multi-core processor 1308 can be a multi-core processor as described herein (such as multi-core processors 405-406). SOC 1300 can additionally include on-chip memory 1305 that can enable a shared on-chip data pool accessible by each of the processing components. The processing components can be optimized for low-power operation to enable deployment of a variety of machine learning platforms including autonomous vehicles and autonomous robots. For example, an implementation of SOC 1300 can be used as part of a master control system for an autonomous vehicle. Where the SOC 1300 is configured for use with an autonomous vehicle, the SOC is designed and configured to comply with relevant functional safety standards of the deployment jurisdiction.

[0212] During operation, the media processor 1302 and the vision processor 1304 can work together to accelerate computer vision operations. The media processor 1302 can enable low-latency decoding of multiple high-resolution (e.g., 4K, 8K) video streams. The decoded video streams can be written to a buffer in the on-chip memory 1305. The vision processor 1304 can then parse the decoded video and perform preliminary processing operations on the frames of the decoded video to prepare them for processing the frames using the trained image recognition model. For example, the vision processor 1304 can accelerate the convolution operations of a CNN used to perform image recognition on high-resolution video data, while the back-end model calculations are performed by the GPGPU 1306.

[0213] The multi-core processor 1308 may include control logic to assist in sequencing and synchronizing data transfers and shared memory operations performed by the media processor 1302 and the vision processor 1304. The multi-core processor 1308 may also function as an application processor to execute software applications that can utilize the inference computing capabilities of the GPGPU 1306. For example, at least a portion of navigation and driving logic may be implemented in software executing on the multi-core processor 1308. Such software may issue computational workloads directly to the GPGPU 1306, or may issue computational workloads to the multi-core processor 1308, which may offload at least a portion of those operations to the GPGPU 1306.

[0214] GPGPU 1306 may include a computational cluster, such as a low-power configuration of processing clusters 706A-706H within GPGPU 700. The computational cluster within GPGPU 1306 may support instructions specifically optimized to perform inference computations on trained neural networks. For example, GPGPU 1306 may support instructions for performing low-precision computations such as 8-bit and 4-bit integer vector operations. Additional System Overview

[0215] Figure 14 is a block diagram of processing system 1400 . Figure 14 Elements with the same or similar names as elements of any other figure herein describe the same elements in the other figure, may operate or function in a similar manner as that, may include the same components, and may be linked to other entities, as those described elsewhere in this document, but are not limited to such. The processing system 1400 can be used in a single-processor desktop system, a multi-processor workstation system, or a server system having (one or more) processors 1402 or processor cores 1407. The processing system 1400 can be a processing platform incorporated within a system-on-chip (SoC) integrated circuit for use with a mobile, handheld, or embedded device, such as within an Internet of Things (IoT) device having wired or wireless connectivity to a local area network or a wide area network.

[0216] The processing system 1400 may be a system having Figure 1 For example, in different configurations, the processor(s) 1402 or processor core 1407 may be associated with the processor 1402 or processor core 1407. Figure 1 The graphics processor(s) 1408 may correspond to the processor(s) 102 of FIG. Figure 1 The external graphics processor 1418 may be Figure 1One of the (one or more) plug-in devices 120.

[0217] The processing system 1400 may include, be coupled with, or be integrated into a server-based gaming platform; a gaming console, including a gaming and media console; a mobile gaming console, a handheld gaming console, or an online gaming console. The processing system 1400 may be part of a mobile phone, a smartphone, a tablet computing device, or a mobile internet-connected device, such as a laptop computer with low internal storage capacity. The processing system 1400 may also include, be coupled with, or be integrated into a wearable device, such as a smartwatch wearable device; smart glasses or clothing that is enhanced with augmented reality (AR) or virtual reality (VR) features to provide visual, audio, or tactile output to supplement the real-world visual, audio, or tactile experience or otherwise provide text, audio, graphics, video, holographic images or video, or tactile feedback. The processing system 1400 may include or be part of a television or set-top box device. The processing system 1400 may include, be coupled to, or be integrated into an autonomous vehicle such as a bus, a tractor-trailer, a sedan, an automobile, or an electric bicycle, an airplane, or a glider (or any combination thereof). The autonomous vehicle may use the processing system 1400 to process the environment sensed around the vehicle.

[0218] The processor(s) 1402 may include one or more instances of a processor core 1407 for processing instructions that, when executed, perform operations for system and user software. At least one of the processor cores 1407 may be configured to process a specific instruction set 1409. The instruction set 1409 may facilitate complex instruction set computing (CISC), reduced instruction set computing (RISC), or computing with very long instruction words (VLIW). In one embodiment, one of the processor cores 1407 may process a different instruction set 1409 that may include instructions for facilitating emulation of other instruction sets. The processor core 1407 may also include other processing devices such as a digital signal processor (DSP).

[0219] The processor(s) 1402 may include a cache memory 1404. Depending on the architecture, the processor(s) 1402 may have a single internal cache or multiple levels of internal cache. In some embodiments, the cache memory is shared between the various components of the processor(s) 1402. In some embodiments, the processor(s) 1402 also uses an external cache (e.g., a level 3 (L3) cache or a last level cache (LLC)) (not shown) that may be shared between the processor cores 1407 using known cache coherence techniques. A register file 1406 may additionally be included in the processor(s) 1402 and may include different types of registers (e.g., integer registers, floating point registers, status registers, and an instruction pointer register) for storing different types of data. Some registers may be general purpose registers, while other registers may be specific to the design of the processor(s) 1402.

[0220] The processor(s) 1402 may be coupled to one or more interface buses 1410 to transmit communication signals, such as addresses, data, or control signals, between the processor(s) 1402 and other components in the processing system 1400. In one of these embodiments, the interface bus(es) 1410 may be a processor bus, such as a version of a Direct Media Interface (DMI) bus. However, the processor bus is not limited to a DMI bus and may include one or more peripheral component interconnect buses (e.g., PCI, PCI Express), a memory bus, or other types of interface buses. For example, the processor(s) 1402 may include a memory controller 1416 and a platform controller hub 1430. The memory controller 1416 facilitates communication between memory devices and other components of the processing system 1400, while the platform controller hub 1430 provides connections to I / O devices via a local I / O bus.

[0221] Memory device 1420 may be a dynamic random access memory (DRAM) device, a static random access memory (SRAM) device, a flash memory device, a phase change memory device, or another memory device with suitable performance to function as process memory. Memory device 1420 may, for example, operate as system memory for processing system 1400 to store data 1422 and instructions 1421 for use when processor(s) 1402 execute applications or processes. Memory controller 1416 may also optionally be coupled to an external graphics processor 1418, which may communicate with graphics processor(s) 1408 in processor(s) 1402 to perform graphics and media operations. In some embodiments, graphics, media, and / or compute operations may be assisted by accelerator 1412, which is a coprocessor that can be configured to perform a specialized set of graphics, media, or compute operations. For example, accelerator 1412 may be a matrix multiplication accelerator used to optimize machine learning or compute operations. The accelerator 1412 may be a ray tracing accelerator that may be used to perform ray tracing operations in conjunction with the graphics processor(s) 1408. The accelerator 1412 may also be an AI accelerator or NPU to accelerate neural network training or inference operations. In one embodiment, an external accelerator 1419 may be used in place of the accelerator 1412 or in conjunction with the accelerator 1412. The accelerator 1412 and / or the external accelerator 1419 may have the same Figure 1 The functionality of the accelerator device(s) 130 is similar to the functionality of the accelerator device(s) 130 .

[0222] A display device 1411 may be provided that may be connected to the processor(s) 1402. The display device 1411 may be one or more of an internal display device, such as in a mobile electronic device or laptop, or an external display device attached via a display interface, such as a display port, etc. The display device 1411 may be a head mounted display (HMD), such as a stereoscopic display device for use with VR or AR applications.

[0223] The platform controller hub 1430 can enable peripheral devices to connect to the memory device 1420 and (one or more) processors 1402 via a high-speed I / O bus. The I / O peripheral devices include, but are not limited to, an audio controller 1446, a network controller 1434, a firmware interface 1428, a wireless transceiver 1426, a touch sensor 1425, a data storage device 1424 (e.g., non-volatile memory, volatile memory, hard drive, flash memory, NAND, 3D NAND, 3D Xpoint / Optane, etc.). The data storage device 1424 can be connected via a storage interface (e.g., SATA) or via a peripheral bus (e.g., PCI, PCI Express). The touch sensor 1425 can include a touch screen sensor, a pressure sensor, or a fingerprint sensor. The wireless transceiver 1426 can be a Wi-Fi transceiver, a Bluetooth transceiver, or a mobile network transceiver (such as a 3G, 4G, 5G, or Long Term Evolution (LTE) transceiver). The firmware interface 1428 enables communication with the system firmware and can be, for example, a Unified Extensible Firmware Interface (UEFI). The network controller 1434 can enable network connectivity to a wired network. In some embodiments, a high-performance network controller (not shown) is coupled to the interface bus(es) 1410. The audio controller 1446 can be a multi-channel high-definition audio controller. In some of these embodiments, the processing system 1400 includes an optional legacy I / O controller 1440 for coupling legacy (e.g., Personal System 2 (PS / 2)) devices to the system. The platform controller hub 1430 can also be connected to one or more Universal Serial Bus (USB) controllers 1442 to connect to input devices such as a keyboard and mouse 1443 combination, a camera 1444, or other USB input devices.

[0224] It will be appreciated that the processing system 1400 shown is exemplary and non-limiting, as other types of differently configured data processing systems may also be used. For example, instances of the memory controller 1416 and the platform controller hub 1430 may be integrated into a discrete external graphics processor, such as the external graphics processor 1418. The platform controller hub 1430 and / or the memory controller 1416 may be external to the processor(s) 1402. For example, the memory controller 1416 and the platform controller hub 1430 may be external to the processing system 1400 and configured as a memory controller hub and a peripheral controller hub within a system chipset that communicate with the processor(s) 1402.

[0225] For example, a circuit board ("sled") can be used on which components such as a CPU, memory, and other components are placed and designed for improved thermal performance. Processing components such as a processor can be located on the top side of the sled, while nearby memory such as DIMMs are located on the bottom side of the sled. As a result of the enhanced airflow provided by this design, the components can operate at higher frequencies and power levels than in a typical system, thereby improving performance. In addition, the sled is configured to blind-mate with power and data communication cables in the rack, thereby enhancing their ability to be quickly removed, upgraded, reinstalled, and / or replaced. Similarly, the various components located on the sled (such as processors, accelerators, memory, and data storage drives) are configured to be easily upgraded due to their increased spacing from each other. In an illustrative embodiment, the components additionally include hardware authentication features to prove their authenticity.

[0226] The data center can utilize a single network architecture ("fabric") that supports multiple other network architectures including Ethernet and Omni-Path. The sled can be coupled to the switch via optical fiber, which provides higher bandwidth and lower latency than typical twisted pair cabling (e.g., Category 5, Category 5e, Category 6, Category 7, Category 8, etc.). Due to the high-bandwidth, low-latency interconnect and network architecture, the data center can pool resources such as memory, accelerators (e.g., GPUs, graphics accelerators, FPGAs, ASICs, neural network and / or artificial intelligence accelerators, etc.), and physically dispersed data storage drives, and provide them to computing resources (e.g., processors) on an as-needed basis, thereby enabling the computing resources to access the pooled resources as if they were local.

[0227] The power supply or power source can provide voltage and / or current to the processing system 1400 or any component or system described herein. In one example, the power supply includes an AC to DC (alternating current to direct current) adapter for plugging into a wall socket. Such AC power can be a renewable energy (e.g., solar) power source. In one example, the power source includes a DC power source such as an external AC to DC converter. The power supply or power supply can also include wireless charging hardware for charging by proximity to a charging field. The power source can include an internal battery, an AC power source, a motion-based power supply, a solar power supply, or a fuel cell source.

[0228] Figures 15A-15C A computing system and a graphics processor are described. Figures 15A-15CElements having the same or similar names as elements of any other figure in this document describe the same elements in that other figure, may operate or function in a similar manner as that one, may include the same components, and may be linked to other entities as those described elsewhere in this document, but are not limited to such.

[0229] Figure 15A 1402. The block diagram of processor 1500 is a diagram of a processor 1500, which may be a variant of one of processor(s) 1402 and may be used in place of one of those processors. Thus, any feature disclosed herein in combination with processor 1500 also discloses the corresponding combination with processor(s) 1402, but is not limited thereto. Processor 1500 may have one or more processor cores 1502A-1502N, at least one memory controller 1514, and a graphics processor 1508. Graphics processor 1508 may be integrated within processor 1500, within a system chipset, or coupled via a system bus. Processor 1500 may include additional cores, up to and including additional core 1502N, represented by a dashed box. Each of processor cores 1502A-1502N includes one or more internal cache units 1504A-1504N. In some embodiments, each processor core 1502A-1502N also has access to one or more shared cache units 1506. Internal cache units 1504A-1504N and shared cache unit(s) 1506 represent a cache memory hierarchy within processor 1500. The cache memory hierarchy may include at least one level of instruction and data cache within each processor core and one or more levels of shared mid-level cache, such as level two (L2), level three (L3), level four (L4), or other levels of cache, with the highest level of cache before external memory being categorized as LLC. In some embodiments, cache coherence logic maintains coherence between the various cache units (e.g., shared cache unit(s) 1506 and internal cache unit(s) 1504A-1504N).

[0230] The processor 1500 may also include one or more bus controller units 1516 and a system agent core 1510. The one or more bus controller units 1516 manage a collection of peripheral buses, such as one or more PCI buses or PCI Express buses. The one or more bus controller units 1516 may also manage one or more memory buses to various external memory devices (not shown). The system agent core 1510 provides management functionality for the various processor components and may include at least one memory controller 1514.

[0231] For example, one or more of the processor cores 1502A-1502N may include support for simultaneous multithreading. The system agent core 1510 includes components for coordinating and operating the cores 1502A-1502N during multithreaded processing. The system agent core 1510 may also include a power control unit (PCU) that includes logic and components for regulating the power state of the processor cores 1502A-1502N and the graphics processor 1508.

[0232] The processor 1500 may further include a graphics processor 1508 for performing graphics processing operations. In some of these embodiments, the graphics processor 1508 is coupled to one or more shared cache units 1506 and a system agent core 1510 including at least one memory controller 1514. The system agent core 1510 may also include a display controller 1511 for driving the graphics processor output to one or more coupled displays. The display controller 1511 may also be a separate module coupled to the graphics processor via at least one interconnect or may be integrated within the graphics processor 1508.

[0233] A ring-based or mesh-based interconnect 1512 may be used to couple the internal components of the processor 1500. However, alternative interconnect units may be used, such as point-to-point interconnects, switched interconnects, or other technologies, including those known in the art. In some of these embodiments having a ring-based or mesh-based interconnect 1512, the graphics processor 1508 is coupled to the ring-based or mesh-based interconnect 1512 via an I / O link 1513.

[0234] Exemplary I / O links 1513 represent at least one of a variety of I / O interconnects, including an on-package I / O interconnect that facilitates communication between various processor components and a high-performance memory module 1518, such as an embedded DRAM module (eDRAM) or a high-bandwidth memory (HMB) module. Optionally, when a DRAM memory system is also present, each of the processor cores 1502A-1502N and the graphics processor 1508 can use the high-performance memory module 1518 as a unified memory and / or a shared last-level cache. Optionally, one or more accelerators 1515 can also be included within the processor 1500, including, for example, an NPU to accelerate certain neural network operations. The NPU can enable lower power inference operations relative to the use of the graphics processor 1508, or can operate in conjunction with the graphics processor 1508 to enable higher inference performance relative to the graphics processor 1508 alone. In one embodiment, the NPU within one or more accelerators 1515 can include matrix or tensor acceleration logic and can be used to implement at least some of the computational operations described herein as being achievable with the graphics processor 1508.

[0235] The processor cores 1502A-1502N may, for example, be homogeneous cores that execute the same instruction set architecture. Alternatively, the processor cores 1502A-1502N are heterogeneous in terms of instruction set architecture (ISA), wherein one or more of the processor cores 1502A-1502N execute a first instruction set and at least one of the other cores executes a subset of the first instruction set or a different instruction set. The processor cores 1502A-1502N may be heterogeneous in terms of microarchitecture, wherein one or more cores with relatively higher power consumption are coupled with one or more power cores with lower power consumption. As another example, the processor cores 1502A-1502N are heterogeneous in terms of computing power. In addition, the processor 1500 may be implemented on one or more chips or chiplets, or the processor 1500 may be implemented as a SoC integrated circuit having the illustrated components in addition to other components. A plurality of chiplets may be used to implement the SoC integrated circuit.

[0236] Figure 15B is a block diagram of the hardware logic of the graphics processor core block 1519 according to some embodiments described herein. In some embodiments, Figure 15B Elements having the same reference numerals (or names) as elements of any other figure herein may operate or function in a manner similar to that described elsewhere herein. In one embodiment, graphics processor core block 1519 is an example of a partition of a graphics processor. Graphics processor core block 1519 may be included in Figure 15A1508 or a discrete graphics processor, parallel processor, and / or compute accelerator. A graphics processor as described herein may include multiple graphics core blocks based on a target power and performance envelope. Each graphics processor core block 1519 may include a functional block 1530 coupled to multiple graphics cores 1521A-1521F, each of which includes modular blocks of fixed-function logic and general-purpose programmable logic. The graphics processor core block 1519 also includes a shared / cache memory 1536 accessible by all graphics cores 1521A-1521F, rasterizer logic 1537, and additional fixed-function logic 1538.

[0237] In some embodiments, functional block 1530 includes a geometry / fixed function pipeline 1531 that can be shared by all graphics cores in graphics processor core block 1519. In various embodiments, geometry / fixed function pipeline 1531 includes a 3D geometry pipeline, a video front-end unit, a thread generator and a global thread dispatcher, as well as a unified return buffer manager that manages a unified return buffer. In one embodiment, functional block 1530 also includes a graphics SoC interface 1532, a graphics microcontroller 1533, and a media pipeline 1534. Graphics SoC interface 1532 provides an interface between graphics processor core block 1519 and other core blocks within a graphics processor or computing accelerator SoC. Graphics microcontroller 1533 is a programmable subprocessor that can be configured to manage various functions of graphics processor core block 1519, including thread dispatching, scheduling, and preemption. Media pipeline 1534 includes logic used to facilitate decoding, encoding, pre-processing, and / or post-processing of multimedia data, including image and video data. The media pipeline 1534 implements media operations by means of requests to the computation or sampling logic within the graphics cores 1521A-1521F. One or more pixel backends 1535 may also be included within the functional block 1530. The one or more pixel backends 1535 include a cache for storing pixel color values ​​and may perform blending operations and lossless color compression on the rendered pixel data.

[0238] In one embodiment, graphics SoC interface 1532 enables graphics processor core block 1519 to communicate with a general-purpose application processor core (e.g., a CPU) and / or other components within the SoC or a system host CPU coupled to the SoC via a peripheral interface. Graphics SoC interface 1532 also enables communication with off-chip memory hierarchy elements such as a shared last-level cache, system RAM, and / or embedded on-chip or on-package DRAM. Graphics SoC interface 1532 may also enable communication with fixed-function devices within the SoC, such as a camera imaging pipeline, and enable the use and / or implementation of global memory atomics that may be shared between graphics processor core block 1519 and the CPU within the SoC. Graphics SoC interface 1532 may also implement power management control for graphics processor core block 1519 and enable interfacing between the clock domain of graphics processor core block 1519 and other clock domains within the SoC. In one embodiment, graphics SoC interface 1532 enables receiving command buffers from a command stream converter and a global thread dispatcher configured to provide commands and instructions to each of one or more graphics cores within the graphics processor. Commands and instructions may be dispatched to the media pipeline 1534 when media operations are to be performed, and to the geometry and fixed function pipeline 1531 when graphics processing operations are to be performed. When compute operations are to be performed, the compute dispatch logic may dispatch commands to the graphics cores 1521A-1521F, thereby bypassing the geometry and media pipelines.

[0239] Graphics microcontroller 1533 can be configured to perform various scheduling and management tasks for graphics processor core block 1519. In one embodiment, graphics microcontroller 1533 can execute graphics and / or compute workloads scheduled across the various vector engines 1522A-1522F, 1524A-1524F and matrix engines 1523A-1523F, 1525A-1525F within graphics cores 1521A-1521F. In this scheduling model, host software executing on a CPU core of the SoC that includes graphics processor core block 1519 can submit a workload to one of multiple graphics processor doorbells, which invokes scheduling operations on the appropriate graphics engine. Scheduling operations include determining which workload to run next, submitting the workload to the command stream converter, preempting existing workloads running on the engine, monitoring the progress of the workload, and notifying the host software when the workload is complete. In one embodiment, the graphics microcontroller 1533 may also facilitate a low power or idle state for the graphics processor core block 1519, thereby providing the graphics processor core block 1519 with the ability to save and restore registers within the graphics processor core block 1519 across low power state transitions independent of the operating system and / or graphics driver software on the system.

[0240] Graphics processor core block 1519 may have more or fewer graphics cores 1521A-1521F than illustrated, up to N modular graphics cores. For each set of N graphics cores, graphics processor core block 1519 may also include shared / cache memory 1536, which may be configured as shared memory or cache memory, rasterizer logic 1537, and additional fixed-function logic 1538 to accelerate various graphics and compute processing operations.

[0241] Within each graphics core 1521A-1521F is a collection of execution resources that can be used to perform graphics operations, media operations, and compute operations in response to requests from the graphics pipeline, media pipeline, or shader programs. Graphics cores 1521A-1521F include multiple vector engines 1522A-1522F, 1524A-1524F, matrix acceleration units 1523A-1523F, 1525A-1525D, cache / shared local memory (SLM), samplers 1526A-1526F, and ray tracing units 1527A-1527F.

[0242] The vector engines 1522A-1522F, 1524A-1524F are general-purpose graphics processing units capable of performing floating-point and integer / fixed-point logic operations in service of graphics, media, or compute operations (including graphics, media, or compute / GPGPU programs). The vector engines 1522A-1522F, 1524A-1524F can operate with variable vector widths using SIMD, SIMT, or SIMT+SIMD execution modes. The matrix acceleration units 1523A-1523F, 1525A-1525D include matrix-matrix and matrix-vector acceleration logic that improves the performance of matrix operations, particularly low and mixed precision (e.g., INT8, FP16, BF16, FP8, FP4) matrix operations for machine learning. In one embodiment, the matrix acceleration units 1523A-1523F, 1525A-1525D support the Microscale (MX) format. In one embodiment, each of the matrix acceleration units 1523A-1523F, 1525A-1525D includes one or more systolic arrays of processing elements that can perform concurrent matrix multiplication or dot product operations on matrix elements.

[0243] Samplers 1526A-1526F can read media or texture data into memory and can sample data differently based on the configured sampler state and the texture / media format being read. Threads executing on vector engines 1522A-1522F, 1524A-1524F or matrix acceleration units 1523A-1523F, 1525A-1525D can utilize caches / SLMs 1528A-1528F within each of graphics cores 1521A-1521F. Caches / SLMs 1528A-1528F can be configured as cache memory or as a pool of shared memory local to each of the corresponding graphics cores 1521A-1521F. Ray tracing units 1527A-1527F within graphics cores 1521A-1521F include ray traversal / intersection circuitry for performing ray traversals using a bounding volume hierarchy (BVH) and identifying intersections between rays and primitives enclosed within the BVH volume. In one embodiment, ray tracing units 1527A-1527F include circuitry for performing depth testing and culling (e.g., using a depth buffer or similar arrangement). In one implementation, ray tracing units 1527A-1527F perform traversal and intersection operations in conjunction with image denoising, at least a portion of which may be performed using associated matrix acceleration units 1523A-1523F, 1525A-1525D.

[0244] Figure 15C 15 is a block diagram of a general-purpose graphics processing unit (GPGPU 1570) that can be configured as a graphics processor (e.g., graphics processor 1508) and / or a computing accelerator, according to embodiments described herein. GPGPU 1570 can be interconnected with a host processor (e.g., one or more CPUs 1546) and memories 1571 and 1572 via one or more system and / or memory buses. Memory 1571 can be system memory shared with one or more CPUs 1546, while memory 1572 is device memory dedicated to GPGPU 1570. For example, components within GPGPU 1570 and memory 1572 can be mapped to memory addresses accessible by one or more CPUs 1546. Access to memories 1571 and 1572 can be facilitated by memory controller 1568. Memory controller 1568 can include an internal direct memory access (DMA) controller 1569 or can include logic to perform operations that would otherwise be performed by a DMA controller. In one embodiment, at least one of the one or more CPUs 1546 may include one or more accelerators 1545, including but not limited to a neural network accelerator.

[0245] GPGPU 1570 includes multiple global cache memories, including an L2 cache 1553, an L1 cache 1554, an instruction cache 1555, and a shared memory 1556, at least a portion of which may be partitioned into cache memories. GPGPU 1570 also includes multiple compute units 1560A-1560N. Each compute unit 1560A-1560N includes a set of vector registers 1561, scalar registers 1562, a vector logic unit 1563, a scalar logic unit 1564, and a scheduler 1584. Compute units 1560A-1560N may also include local shared memory 1565 and local cache memories 1566. Compute units 1560A-1560N may be coupled to a constant cache 1567, which may be used to store constant data, which is data that does not change during the execution of a kernel or shader program executed on GPGPU 1570. The constant cache 1567 can be a scalar data cache and can fetch cached data directly into the scalar register 1562. In one embodiment, the compute units 1560A-1560N further include at least one matrix unit 1580 for accelerating matrix, tensor, or artificial intelligence operations and at least one ray tracing unit (RT unit 1582) for accelerating ray tracing operations. The at least one matrix unit 1580 and the RT unit 1582 can include functionality similar to other matrix / tensor accelerators and ray tracing accelerators described herein.

[0246] During operation, one or more CPUs 1546 may write commands to registers or memory in the GPGPU 1570 that have been mapped into the accessible address space. The command processor 1557 may read the commands from the registers or memory and determine how those commands will be processed within the GPGPU 1570. The thread dispatcher 1558 may then be used to dispatch threads to the compute units 1560A-1560N to execute those commands. Each compute unit 1560A-1560N may execute threads independently of other compute units. In addition, each compute unit 1560A-1560N may be independently configured for conditional computation and may conditionally output the results of the computation to memory. When the submitted commands are completed, the command processor 1557 may interrupt one or more CPUs 1546.

[0247] Figure 161 is a block diagram of a graphics processor 1600, which may be a discrete graphics processing unit or may be a graphics processor integrated with multiple processing cores or other semiconductor devices (such as, but not limited to, memory devices or network interfaces). Elements of the graphics processor 1600 with the same or similar names as those of any other figure herein describe the same elements in the other figures, may operate or function in a similar manner to that, may include the same components, and may be linked to other entities, such as those described elsewhere in this article, but are not limited to such. For example, the graphics processor 1600 may be a variant of the graphics processor 1508, and the graphics processor 1600 may be used in place of the graphics processor 1508. The graphics processor may communicate via a memory-mapped I / O interface to registers on the graphics processor and using commands placed in the processor memory. The graphics processor 1600 may include a memory interface 1614 for accessing local memory, one or more internal caches, one or more shared external caches, and / or system memory.

[0248] The graphics processor 1600 may include a display controller 1602 for driving display output data to a display device 1618. The display controller 1602 includes hardware for displaying and compositing multiple layers of video or one or more overlay planes of user interface elements. The display device 1618 may be an internal or external display device. In one embodiment, the display device 1618 is a head-mounted display device, such as a VR or AR display device. The graphics processor 1600 may include a video codec engine 1606 for encoding media into one or more media coding formats, decoding media from one or more media coding formats, or transcoding media between one or more media coding formats, the one or more media coding formats including but not limited to: Moving Picture Experts Group (MPEG) format (such as MPEG-2), Advanced Video Coding (AVC) format (such as H.264 / MPEG-4 AVC, H.265 / HEVC), Alliance for Open Media (AOMedia) VP8, VP9, ​​and Society of Motion Picture and Television Engineers (SMPTE) 421M / VC-1 and Joint Photographic Experts Group (JPEG) format (such as JPEG and Motion JPEG (MJPEG) format).

[0249] The graphics processor 1600 may include a block image transfer (BLIT) engine 1603 for performing two-dimensional (2D) rasterizer operations, including, for example, bit-boundary block transfers, or may use one or more components of a graphics processing engine (GPE 1610) to perform 2D graphics operations. The GPE 1610 may include a 3D pipeline 1612 for performing 3D operations, such as rendering three-dimensional images and scenes using processing functions that operate on 3D primitive shapes (e.g., rectangles, triangles, etc.). The 3D pipeline 1612 includes programmable and fixed-function elements that perform various tasks within the element and / or spawn execution threads to a 3D / media subsystem 1615. While the 3D pipeline 1612 may be used to perform media operations, the GPE 1610 may also include a media pipeline 1616 specifically for performing media operations, such as image or video decoding, encoding, post-processing, and enhancement. The media pipeline 1616 may include fixed-function or programmable logic units that are used to perform one or more specialized media operations (such as video decode acceleration, video deinterlacing, and video encoding acceleration) instead of or on behalf of the video codec engine 1606. The media pipeline 1616 may also include a thread generation unit that is used to generate threads for execution on the 3D / media subsystem 1615. The generated threads perform calculations for the media operations on one or more graphics execution units included in the 3D / media subsystem 1615.

[0250] Figure 17A The graphics processor 1720 is Figure 161720 can be used in place of graphics processor 1600 and vice versa. Therefore, any feature disclosed herein in combination with graphics processor 1600 also discloses the corresponding combination with graphics processor 1720, but is not limited thereto. According to the embodiments described herein, graphics processor 1720 has a sliced ​​architecture. Graphics processor 1720 may include a graphics processing engine cluster 1722 having multiple graphics processing engines within multiple graphics engine slices. Each graphics engine slice 1710A-1710D may be interconnected via a set of slice interconnects 1723A-1723F. Each graphics engine slice 1710A-1710D may also be connected to a memory module or memory device 1726A-1726D via a memory interconnect 1725A-1725D. Memory devices 1726A-1726D may use any graphics memory technology. For example, memory devices 1726A-1726D may be graphics double data rate (GDDR) memory. The memory devices 1726A-1726D may be high bandwidth memory (HBM) modules that may be on-die with their corresponding graphics engine slices 1710A-1710D. The memory devices 1726A-1726D may be stacked memory devices that may be stacked on top of their corresponding graphics engine slices 1710A-1710D. Figures 25A-25B As described in further detail in , each graphics engine slice 1710A- 1710D and associated memory devices 1726A- 1726D can reside on a separate chiplet that is bonded to a base die or base substrate.

[0251] Graphics processor 1720 may be configured with a non-uniform memory access (NUMA) system in which memory devices 1726A-1726D are coupled to associated graphics engine slices 1710A-1710D. A given memory device may be accessed by graphics engine slices other than the slice to which it is directly connected. However, access latency to memory devices 1726A-1726D may be minimized when accessing the local slice. In one embodiment, a cache-coherent NUMA (ccNUMA) system is enabled that uses slice interconnects 1723A-1723F to enable communication between cache controllers within graphics engine slices 1710A-1710D to maintain a consistent memory image when more than one cache stores the same memory location.

[0252] The graphics processing engine cluster 1722 can be connected to an interconnect fabric 1724, which can interconnect on-chip or on-package structures. In one embodiment, the interconnect fabric 1724 includes a network processor, a network-on-chip (NoC), or another switching processor that enables the interconnect fabric 1724 to function as a packet-switched interconnect fabric for exchanging data packets between components of the graphics processor 1720. The interconnect fabric 1724 can enable communication between the graphics engine slices 1710A-1710D and components such as the video codec 1706 and one or more replication engines 1704. The one or more replication engines 1704 can be used to move data to, from, and between memory devices 1726A-1726D and memory external to the graphics processor 1720 (e.g., system memory). The interconnect fabric 1724 can also be used to interconnect the graphics engine slices 1710A-1710D. The graphics processor 1720 may optionally include a display controller 1702 to enable connection to a display device 1718. The graphics processor may also be configured as a graphics or computing accelerator. In an accelerator configuration, the display controller 1702 and the display device 1718 may be omitted.

[0253] The graphics processor 1720 can be connected to the host system via a host interface 1728. The host interface 1728 can enable communication between the graphics processor 1720, system memory, and / or other system components. The host interface 1728 can be, for example, a PCI Express bus or another type of host system interface. For example, the host interface 1728 can be an NVLink or NVSwitch interface. The host interface 1728 and the interconnect structure 1724 can cooperate to enable multiple instances of the graphics processor 1720 to act as a single logical device. The cooperation between the host interface 1728 and the interconnect structure 1724 can also enable the individual graphics engine slices 1710A-1710D to be presented to the host system as different logical graphics devices.

[0254] Figure 17B The computing accelerator 1730 according to the embodiments described herein is illustrated. The computing accelerator 1730 may include Figure 17BThe computing accelerator 1730 may be configured as an AI accelerator or an NPU. In such an embodiment, the execution logic of the computing engine slices 1740A-1740D may be primarily targeted at matrix or tensor operations and include the tensor cores or matrix engines described herein. The computing engine slices 1740A-1740D may not include fixed-function graphics processing logic, but in some embodiments, one or more of the computing engine slices 1740A-1740D may include logic for performing media acceleration. The computing engine slices 1740A-1740D may be connected to the memory devices 1726A-1726D via the memory interconnects 1725A-1725D. Memory devices 1726A-1726D and memory interconnects 1725A-1725D may be similar technologies as in graphics processor 1720 or may be different. Compute engine slices 1740A-1740D may also be interconnected via a set of slice interconnects 1723A-1723F and may be connected to and / or interconnected by interconnect fabric 1724. In one embodiment, compute accelerator 1730 includes a large L3 cache 1736 that may be configured as a device-wide cache. Compute accelerator 1730 may also be configured in a manner similar to that of a processor. Figure 17B The graphics processor 1720 is similarly connected to a host processor and memory via a host interface 1728 .

[0255] The computing accelerator 1730 may also include an integrated network interface 1742. In one embodiment, the integrated network interface 1742 includes a network processor and controller logic that enables the computing engine cluster 1732 to communicate via a physical layer interconnect 1744 without requiring the data to traverse the host system's memory. In one embodiment, one of the computing engine slices 1740A-1740D is replaced by the network processor logic, and data to be transmitted or received via the physical layer interconnect 1744 can be directly transferred to or from the memory devices 1726A-1726D. Multiple instances of the computing accelerator 1730 can be joined to a single logical device via the physical layer interconnect 1744. Alternatively, the various computing engine slices 1740A-1740D can be presented as different network-accessible computing accelerator devices. Graphics processing resources

[0256] Figures 18A-18CExecution logic is illustrated for an array of processing elements including those employed in a graphics processor according to embodiments described herein. Figure 18A A graphics core cluster according to an embodiment is illustrated. Figure 18B A vector engine of a graphics core according to an embodiment is illustrated. Figure 18C A matrix engine of a graphics core according to an embodiment is illustrated. Figures 18A-18C Elements in FIG. 1 having the same reference numeral as elements in any other figure herein may operate or function in any manner similar to that described elsewhere herein. For example, Figure 15B Considered in the context of the graphics processor core block 1519 Figures 18A-18C In one embodiment, Figures 18A-18C The components have Figure 15A Graphics processor 1508 or Figure 15C The equivalent component of the GPGPU 1570 has similar functionality.

[0257] like Figure 18A As shown in FIG, in one embodiment, graphics core cluster 1800 includes graphics core block 1519, which can be any number of graphics cores (e.g., graphics core 1815A, graphics core 1815B, through graphics core 1815N). In one embodiment, the elements of graphics cores 1815A-1815N have the same Figure 15B In such an embodiment, graphics cores 1815A-1815N each include circuit modules including, but not limited to, vector engines 1802A-1802N, matrix engines 1803A-1803N, memory load / store units 1804A-1804N, instruction caches 1805A-1805N, data caches / shared local memories 1806A-1806N, ray tracing units 1808A-1808N, and samplers 1810A-1810N. The circuit modules of graphics cores 1815A-1815N may further include fixed-function logic 1812A-1812N. The number of vector engines 1802A-1802N and matrix engines 1803A-1803N within a design's graphics cores 1815A-1815N may vary based on the workload, performance, and power targets for the design.

[0258] With reference to graphics core 1815A, vector engine 1802A and matrix engine 1803A can be configured to perform parallel computation operations on data in various integer and floating-point data formats based on instructions associated with shader programs. Each vector engine 1802A and matrix engine 1803A can act as a programmable general-purpose computing unit capable of executing multiple simultaneous hardware threads while processing multiple data elements in parallel for each thread. Vector engine 1802A and matrix engine 1803A support processing variable-width vectors with various SIMD widths (including but not limited to SIMD8, SIMD16, and SIMD32). Input data elements can be stored in registers as compact data types, and vector engine 1802A and matrix engine 1803A can process various elements based on the data size of the elements. For example, when operating on a 256-bit wide vector, the bits of the vector are stored in registers and the vector is processed as four separate 64-bit packed data elements (quadword (QW) size data elements), eight separate 32-bit packed data elements (doubleword (DW) size data elements), sixteen separate 16-bit packed data elements (word (W) size data elements), or thirty-two separate 8-bit data elements (byte (B) size data elements). However, different vector widths and register sizes are possible. In one embodiment, the vector engine 1802A and the matrix engine 1803A can also be configured for SIMT operations on warps and thread groups of various sizes (e.g., 8, 16, or 32 threads).

[0259] Continuing with the graphics core 1815A, the memory load / store unit 1804A services memory access requests issued by the vector engine 1802A, the matrix engine 1803A, and / or other components of the graphics core 1815A that have access to memory. The memory access requests may be processed by the memory load / store unit 1804A to load or store the requested data into a cache or memory, or to load or store the requested data from a cache or memory into a register file associated with the vector engine 1802A and / or the matrix engine 1803A. The memory load / store unit 1804A may also perform prefetch operations. Also referring to Figure 19 In one embodiment, the memory load / store unit 1804A is configured to provide SIMT scatter / gather prefetches or block prefetches for data stored in memory 1910, from memory local to other slices via the slice interconnect 1908, or from system memory. Prefetches can be performed for a specific L1 cache (e.g., data cache / shared local memory 1806A), the L2 cache 1904, or the L3 cache 1906. In one embodiment, a prefetch to the L3 cache 1906 automatically causes the data to be stored in the L2 cache 1904.

[0260] Instruction cache 1805A stores instructions to be executed by graphics core 1815A. In one embodiment, graphics core 1815A also includes instruction fetch and prefetch circuitry that fetches or prefetches instructions into instruction cache 1805A. Graphics core 1815A also includes instruction decode logic for decoding instructions within instruction cache 1805A. Data cache / shared local memory 1806A can be configured as a data cache managed by a cache controller that implements a cache replacement policy and / or configured as explicitly managed shared memory. Ray tracing unit 1808A includes circuitry for accelerating ray tracing operations. Sampler 1810A provides texture sampling for 3D operations and media sampling for media operations. Fixed function logic 1812A includes fixed function circuitry shared between various instances of vector engine 1802A and matrix engine 1803A. Graphics cores 1815B-1815N can operate in a manner similar to graphics core 1815A.

[0261] The functionality of the instruction caches 1805A-1805N, data cache / shared local memory 1806A-1806N, ray tracing units 1808A-1808N, samplers 1810A-1810N, and fixed function logic 1812A-1812N corresponds to the equivalent functionality in the graphics processor architecture described herein. For example, the instruction caches 1805A-1805N may be configured to be in the same manner as the ray tracing units 1808A-1808N. Figure 15C The data cache / shared local memory 1806A-1806N, ray tracing units 1808A-1808N and samplers 1810A-1810N can operate in a similar manner to the instruction cache 1555 of FIG. Figure 15B The cache / SLM 1528A-1528F, ray tracing units 1527A-1527F, and samplers 1526A-1526F of the fixed function logic 1812A-1812N may include Figure 15B In one embodiment, ray tracing units 1808A-1808N include components for performing the ray tracing operations performed by Figure 3 The ray tracing core 372 is a circuit module that performs ray tracing acceleration operations.

[0262] like Figure 18BAs shown in FIG, in one embodiment, the vector engine 1802 includes an instruction fetch unit 1837, a general register file array (GRF 1824), an architectural register file array (ARF 1826), a thread arbiter 1822, an issue unit 1830, a branch unit 1832, a SIMD FPU 1834, and, in one embodiment, a SIMD ALU 1835. The GRF 1824 and ARF 1826 include a set of general register files and architectural register files associated with each hardware thread that may be active in the vector engine 1802. In one embodiment, per-thread architectural state is maintained in the ARF 1826, while data used during thread execution is stored in the GRF 1824. The execution state of each thread (including the instruction pointer for each thread) can be saved in thread-specific registers in the ARF 1826. Register renaming can be used to dynamically assign registers to hardware threads.

[0263] In one embodiment, vector engine 1802 has an architecture that is a combination of simultaneous multithreading (SMT) and fine-grained interleaved multithreading (IMT). The architecture has a modular configuration that can be fine-tuned at design time based on the number of registers per graphics core and the target number of simultaneous threads, where graphics core resources are divided across the logic used to execute multiple simultaneous threads. The number of logical threads that can be executed by vector engine 1802 is not limited to the number of hardware threads, and multiple logical threads can be assigned to each hardware thread.

[0264] In one embodiment, vector engine 1802 can collectively issue multiple instructions, each of which can be a different instruction. Thread arbiter 1822 can dispatch instructions to one of issue unit 1830, branch unit 1832, or SIMDFPU 1834 for execution. Each execution thread can access 128 general-purpose registers within GRF 1824, each of which can store 32 bytes accessible as variable-width vectors of 32-bit data elements. In one embodiment, each thread has access to 4 kilobytes within GRF 1824, although embodiments are not limited thereto and more or fewer register resources may be provided in other embodiments. In one embodiment, vector engine 1802 is partitioned into seven hardware threads that can independently execute computational operations, although the number of threads per vector engine 1802 may vary depending on the embodiment. For example, in one embodiment, up to 16 hardware threads are supported. In an embodiment where seven threads have access to 4 kilobytes, GRF 1824 can store a total of 28 kilobytes. In an embodiment where 16 threads have access to 4 kilobytes, GRF 1824 can store a total of 64 kilobytes. Flexible addressing modes allow registers to be addressed together to efficiently create wider registers or to represent strided rectangular block data structures.

[0265] In one embodiment, memory operations, sampler operations, and other higher latency system communications are dispatched via "send" instructions executed by message passing send mechanism 1830. In one embodiment, branch instructions are dispatched to branch unit 1832 to facilitate SIMD divergence and eventual convergence.

[0266] In one embodiment, the SIMD FPU 1834 of the vector engine 1802 performs floating-point operations. In one embodiment, the SIMD FPU 1834 also supports integer computations. In one embodiment, the SIMD FPU 1834 can perform up to M 32-bit floating-point (or integer) operations, or up to 2M 16-bit integer or 16-bit floating-point operations. In one embodiment, at least one of the FPUs provides extended math capabilities to support high-throughput transcendental math functions and double-precision 64-bit floating-point. In some embodiments, a SIMD ALU 1835 configured to perform 8-bit integer operations is also present, and the SIMD ALU 1835 configured to perform 8-bit integer operations can be specifically optimized to perform operations associated with machine learning computations. In one embodiment, the SIMD ALU 1835 is replaced by a SIMD FPU 1834 that can be configured to perform both integer and floating-point operations. In one embodiment, the SIMD FPU 1834 and the SIMD ALU 1835 can be configured to execute SIMT programs. In one embodiment, combined SIMD+SIMT operations are supported.

[0267] In one embodiment, an array of multiple instances of vector engine 1802 can be instantiated within a graphics core. For scalability, product architects can choose the exact number of vector engines grouped per graphics core. In one embodiment, vector engine 1802 can execute instructions across multiple execution lanes. In another embodiment, each thread executing on vector engine 1802 executes on a different lane.

[0268] like Figure 18CAs shown in , in one embodiment, the matrix engine 1803 includes an array of processing elements configured to perform tensor operations including vector / matrix and matrix / matrix operations (such as, but not limited to, matrix multiplication and / or dot product operations). The matrix engine 1803 is configured with M rows and N columns of processing elements 1852AA-1852MN including multiplier and adder circuits organized in a pipelined manner. In one embodiment, the processing elements 1852AA-1852MN form a physical pipeline stage of an N-wide and M-deep systolic array that can be used to perform vector / matrix operations or matrix / matrix operations (including matrix multiplication, fused multiply-add, dot product or other general matrix-matrix multiplication (GEMM) operations) in a data-parallel manner. In one embodiment, the matrix engine 1803 supports 16-bit and 8-bit floating point operations, as well as 8-bit, 4-bit, 2-bit and binary integer operations. The matrix engine 1803 can also be configured to accelerate specific machine learning operations. In such an embodiment, the matrix engine 1803 may be configured with support for a bfloat (brain floating point) 16-bit floating point format or a tensor float 32-bit floating point format (TF32) having a different number of mantissa and exponent bits relative to the Institute of Electrical and Electronics Engineers (IEEE) 754 format.

[0269] In one embodiment, during each cycle, each stage can add the result of the operation performed at that stage to the output of the previous stage. In other embodiments, after a set of computation cycles, the pattern of data movement between processing elements 1852AA-1852MN can vary based on the instruction or macro-operation being executed. For example, in one embodiment, partial sum loopback is enabled, and the processing elements can instead add the output of the current cycle to the output generated in the previous cycle. In one embodiment, the last stage of the systolic array can be configured with a loopback to the initial stage of the systolic array. In such an embodiment, the number of physical pipeline stages can be decoupled from the number of logical pipeline stages supported by the matrix engine 1803. For example, where processing elements 1852AA-1852MN are configured as a systolic array of M physical stages, a loopback from stage M to the initial pipeline stage can enable processing elements 1852AA-1852MN to operate as a systolic array of, for example, 2M, 3M, 4M, etc., logical pipeline stages.

[0270] In one embodiment, the matrix engine 1803 includes memories 1841A-1841N, 1842A-1842M for storing input data in the form of row and column data of an input matrix. The memories 1842A-1842M can be configured to store the row elements (A0-Am) of a first input matrix and the memories 1841A-1841N can be configured to store the column elements (B0-Bn) of a second input matrix. The row elements and column elements are provided as input to the processing elements 1852AA-1852MN for processing. In one embodiment, the row elements and column elements of the input matrix can be stored in a systolic register file 1840 within the matrix engine 1803 before those elements are provided to the memories 1841A-1841N, 1842A-1842M. In one embodiment, the systolic register file 1840 is excluded and the registers (e.g., Figure 18B 1802 of the vector engine 1802) or other memory of the graphics core including the matrix engine 1803 (e.g. Figure 18A The results generated by the processing elements 1852AA-1852MN are then output to output buffers and / or written to register files (e.g., systolic register file 1840, GRF 1824, data cache / shared local memory 1806A-1806N) for further processing by other functional units of the graphics processor or for output to memory.

[0271] In some embodiments, the matrix engine 1803 is configured to support input sparsity, where multiplication operations for sparse regions of input data can be bypassed by skipping multiplication operations with zero-valued operands. In one embodiment, processing elements 1852AA-1852MN are configured to skip the execution of certain operations with zero-valued inputs. In one embodiment, the sparsity within the input matrix can be detected, and operations with known zero output values ​​can be bypassed before being submitted to the processing elements 1852AA-1852MN. Loading zero-valued operands into the processing elements can be bypassed, and processing elements 1852AA-1852MN can be configured to perform multiplication on non-zero-valued input elements. The matrix engine 1803 can also be configured to support output sparsity, so that operations with predetermined zero results are bypassed. For input sparsity and / or output sparsity, in one embodiment, metadata is provided to the processing elements 1852AA-1852MN to indicate which processing elements and / or data channels will be active during a processing cycle.

[0272] In one embodiment, matrix engine 1803 includes hardware for enabling the operation of sparse data with the compressed representation of a sparse matrix, wherein the sparse matrix stores non-zero values ​​and metadata identifying the position of the non-zero values ​​in the matrix. Exemplary compressed representations include, but are not limited to, compressed tensor representations, such as compressed sparse rows (CSR) representations, compressed sparse columns (CSC) representations, compressed sparse fibers (CSF) representations. Support for compressed representations enables operations to be performed on inputs in compressed tensor formats without requiring decompression or decoding of compressed representations. In such embodiments, operations can be performed only on non-zero input values, and the non-zero output values ​​as a result can be mapped into the output matrix. In some embodiments, hardware support for machine-specific lossless data compression formats used when transmitting data within hardware or across a system bus is also provided. Such data can be retained in the compressed format of sparse input data, and matrix engine 1803 can use the compression metadata for compressed data to enable operations to be performed only on non-zero values ​​or to enable blocks of zero data input to be bypassed for multiplication operations.

[0273] In various embodiments, the input data can be provided by the programmer in a compressed tensor representation, or the codec can compress the input data into a compressed tensor representation or another sparse data encoding. In addition to supporting compressed tensor representations, streaming compression of sparse input data can be performed before the data is provided to the processing elements 1852AA-1852MN. In one embodiment, compression is performed on data written to a cache memory associated with the graphics core cluster 1800, where the compression is performed using an encoding supported by the matrix engine 1803. In one embodiment, the matrix engine 1803 includes support for inputs having structured sparsity in which a predetermined level or pattern of sparsity is imposed on the input data. This data can be compressed to a known compression ratio, where the compressed data is processed by the compression elements 1852AA-1852MN based on metadata associated with the compressed data.

[0274] Figure 19 A slice 1900 of a multi-slice processor according to an embodiment is illustrated. In one embodiment, the slice 1900 represents Figure 17A Graphics engine chips 1710A-1710D or Figure 17B Slice 1900 of a multi-slice graphics processor includes an array of graphics core clusters (e.g., graphics core cluster 1800A, graphics core cluster 1800B, through graphics core cluster 1800N), each of which includes an array of graphics cores 515A-515N. Slice 1900 also includes a global dispatcher 1902 for dispatching threads to processing resources of slice 1900.

[0275] Slice 1900 may include or be coupled with an L3 cache 1906 and a memory 1910. In various embodiments, L3 cache 1906 may be excluded, or slice 1900 may include additional levels of cache, such as an L4 cache. In one embodiment, each instance of slice 1900 in a multi-slice graphics processor is associated with a memory 1910, such as in Figure 17A and Figure 17B In one embodiment, the multi-chip processor can be configured as a multi-chip module in which the L3 cache 1906 and / or memory 1910 reside on a separate chiplet compared to the graphics core cluster 1800A-1800N. In this context, a chiplet is an at least partially packaged integrated circuit that includes different logic units that can be assembled into a larger package with other chiplets. For example, the L3 cache 1906 can be included in a dedicated cache chiplet or can reside on the same chiplet as the graphics core cluster 1800A-1800N. In one embodiment, the L3 cache 1906 can be included in an active base die or an active interposer.

[0276] Memory structure 1903 enables communication between graphics core cluster 1800A-1800N, L3 cache 1906, and memory 1910. L2 cache 1904 is coupled to memory structure 1903 and can be configured to cache transactions executed with the aid of memory structure 1903. Slice interconnect 1908 enables communication with other slices on the graphics processor and can be Figure 17A and 17B 1723F. In embodiments where L3 cache 1906 is excluded from slice 1900, L2 cache 1904 may be configured as a combined L2 / L3 cache. Memory structure 1903 may be configured to route data to a memory controller associated with memory 1910 or to L3 cache 1906 based on the presence or absence of L3 cache 1906 in a particular implementation. L3 cache 1906 may be configured as a per-slice cache dedicated to the processing resources of slice 1900 or may be a partition of a GPU-wide L3 cache.

[0277] Figure 20is a block diagram illustrating a graphics processor instruction format 2000. The graphics processor execution unit supports an instruction set having instructions in a variety of formats. Solid-line boxes illustrate components that are typically included in execution unit instructions, while dashed lines include components that are optional or included only in a subset of instructions. In some embodiments, the graphics processor instruction format 2000 depicted and illustrated is a macroinstruction, as it is an instruction supplied to the execution unit, rather than a micro-operation resulting from instruction decoding once the instruction is processed. Therefore, a single instruction can cause the hardware to execute multiple micro-operations.

[0278] As described herein, the graphics processor execution unit can natively support instructions in the 128-bit instruction format 2010. Depending on the selected instruction, instruction options, and the number of operands, a compressed 64-bit instruction format 2030 may be used for some instructions. The native 128-bit instruction format 2010 provides access to all instruction options, while some options and operations are restricted in the 64-bit command format 2030. The native instructions available in the 64-bit command format 2030 vary depending on the embodiment. Instructions are partially compressed using a set of index values ​​in the index field 2013. The execution unit hardware references a set of compression tables based on the index values ​​and uses the compression table output to reconstruct the native instruction in the 128-bit instruction format 2010. Instructions of other sizes and formats may be used.

[0279] For each format, the instruction opcode 2012 defines the operation to be performed by the execution unit. The execution unit executes each instruction in parallel across multiple data elements of each operand. For example, in response to an addition instruction, the execution unit performs a simultaneous addition operation across each color channel representing a texture element or picture element. By default, the execution unit executes each instruction across all data channels of the operand. The instruction control field 2014 can enable control of certain execution options such as channel selection (e.g., predication) and data channel order (e.g., swizzling). For instructions in the 128-bit instruction format 2010, the execution size field 2016 limits the number of data channels that will be executed in parallel. The execution size field 2016 may not be available for use with the compressed 64-bit instruction format 2030.

[0280] Some execution unit instructions have up to three operands, including two source operands src0 2020 and src1 2022 and a destination operand (dest 2018). Other instructions (such as, for example, data manipulation instructions, dot product instructions, multiply-add instructions, or multiply-accumulate instructions) may have a third source operand (e.g., src2 2024). The instruction opcode 2012 determines the number of source operands. The last source operand of an instruction may be an immediate (e.g., hard-coded) value passed with the instruction. The execution unit may also support multiple destination instructions, where one or more of the destinations may be implicit or implicit based on the instruction and / or the specified destination.

[0281] The 128-bit instruction format 2010 may include an access / address mode field 2026 that specifies, for example, whether direct register addressing mode or indirect register addressing mode is used. When direct register addressing mode is used, the register address of one or more operands is provided directly by bits in the instruction.

[0282] The 128-bit instruction format 2010 may also include an access / addressing mode field 2026 that specifies the address mode and / or access mode of the instruction. The access mode may be used to define the data access alignment for the instruction. Access modes including 16-byte aligned access mode and 1-byte aligned access mode may be supported, wherein the byte alignment of the access mode determines the access alignment of the instruction operands. For example, when in the first mode, the instruction may use byte-aligned addressing for source and destination operands, and when in the second mode, the instruction may use 16-byte aligned addressing for all source and destination operands.

[0283] The address mode portion of the access / address mode field 2026 may determine whether the instruction is to use direct or indirect addressing. When direct register addressing mode is used, bits in the instruction directly provide the register address of one or more operands. When indirect register addressing mode is used, the register address of one or more operands may be calculated based on the address register value and the address immediate field in the instruction.

[0284] Instructions can be grouped based on the instruction opcode 2012 bit-fields to simplify opcode decoding 2040. For 8-bit opcodes, positions 4, 5, and 6 allow the execution unit to determine the type of opcode. The precise opcode grouping shown is merely an example. Move and logic opcode group 2042 can include data movement and logic instructions (e.g., move (mov), compare (cmp)). Move and logic opcode group 2042 can share five least significant bits (LSBs), wherein the move (mov) instruction is in the form of 0000xxxxb and the logic instruction is in the form of 0001xxxxb. Flow control instruction group 2044 (e.g., call, jump (jmp)) includes instructions in the form of 0010xxxxb (e.g., 0x20). Miscellaneous instruction group 2046 includes a mixture of instructions, including synchronization instructions (e.g., wait, send) in the form of 0011xxxxb (e.g., 0x30). The parallel math instruction group 2048 includes component-wise arithmetic instructions (e.g., add, multiply (mul)) of the form 0100xxxxb (e.g., 0x40). The parallel math instruction group 2048 performs arithmetic operations in parallel across data lanes. The vector math group 2050 includes arithmetic instructions (e.g., dp4) of the form 0101xxxxb (e.g., 0x50). The vector math group performs arithmetic such as dot product calculations on vector operands. In one embodiment, the opcode decode 2040 can be used to determine which portion of the execution unit will be used to execute the decoded instruction. For example, some instructions can be designated as systolic instructions to be executed by a systolic array. Other instructions, such as ray tracing instructions (not shown), can be routed to a ray tracing core or ray tracing logic within a slice or partition of the execution logic. Graphics pipeline

[0285] Figure 21 is a block diagram of a graphics processor 2100 according to another embodiment. Figure 21 Elements having the same or similar names as elements of any other figure in this document describe the same elements in that other figure, may operate or function in a similar manner as that one, may include the same components, and may be linked to other entities as those described elsewhere in this document, but are not limited to such.

[0286] The graphics processor 2100 may include different types of graphics processing pipelines, such as a geometry pipeline 2120, a media pipeline 2130, a display engine 2140, thread execution logic 2150, and a render output pipeline 2170. The graphics processor 2100 may be a graphics processor within a multi-core processing system that includes one or more general-purpose processing cores. The graphics processor 2100 may be controlled by register writes to one or more control registers (not shown) or by commands issued to the graphics processor 2100 via a ring or mesh interconnect 2102. The ring or mesh interconnect 2102 may couple the graphics processor 2100 to other processing components (such as other graphics processors or general-purpose processors). Commands from the ring or mesh interconnect 2102 are interpreted by a command stream converter 2103, which supplies instructions to various components of the geometry pipeline 2120 or the media pipeline 2130.

[0287] The command stream converter 2103 can direct the operation of the vertex fetcher 2105, which reads vertex data from memory and executes vertex processing commands provided by the command stream converter 2103. The vertex fetcher 2105 can provide the vertex data to the vertex shader 2107, which performs coordinate space transformation and lighting operations on each vertex. The vertex fetcher 2105 and the vertex shader 2107 can execute the vertex processing instructions by dispatching execution threads to graphics cores 2152A-2152B with the help of the thread dispatcher 2131.

[0288] Graphics cores 2152A-2152B may be arrays of vector processors with instruction sets for performing graphics and media operations. Graphics cores 2152A-2152B may have an attached L1 cache 2151 dedicated to each array or shared between arrays. The cache may be configured as a data cache, an instruction cache, or a single cache partitioned into different partitions containing data and instructions.

[0289] The geometry pipeline 2120 may include a tessellation component for performing hardware-accelerated tessellation of 3D objects. The programmable hull shader 2111 may configure tessellation operations. The programmable domain shader 2117 may provide back-end evaluation of tessellation output. The tessellator 2113 may operate under the direction of the programmable hull shader 2111 and contain specialized logic for generating a set of detailed geometric objects based on a coarse geometric model provided as input to the geometry pipeline 2120. Alternatively, if tessellation is not used, the tessellation component (e.g., the programmable hull shader 2111, the tessellator 2113, and the programmable domain shader 2117) may be bypassed. The tessellation component may operate based on data received from the vertex shader 2107.

[0290] The complete geometric object can be processed by the geometry shader 2119 via one or more threads dispatched to the graphics cores 2152A-2152B, or the complete geometric object can proceed directly to the clipper 2129. The geometry shader can operate on the entire geometric object, rather than on vertices or patches of vertices as in previous stages of the graphics pipeline. If tessellation is disabled, the geometry shader 2119 receives input from the vertex shader 2107. Through the geometry shader program, the geometry shader 2119 can be programmable to perform geometry tessellation if the tessellation unit is disabled.

[0291] Before rasterization, the clipper 2129 processes the vertex data. The clipper 2129 can be a fixed-function clipper or a programmable clipper with clipping and geometry shader functionality. The rasterizer and depth test component 2173 in the render output pipeline 2170 can dispatch a pixel shader to convert the geometric objects into a per-pixel representation. The pixel shader logic can be included in the thread execution logic 2150. Optionally, the application can bypass the rasterizer and depth test component 2173 and access the unrasterized vertex data via the outflow unit 2123.

[0292] The graphics processor 2100 has an interconnect bus, interconnect structure, or other interconnect mechanism that allows data and messages to be passed between the main components of the processor. In some embodiments, the graphics cores 2152A-2152B and associated logic units (e.g., L1 cache 2151, samplers 2154, texture cache 2158, etc.) are interconnected via data ports 2156 to perform memory accesses and communicate with the processor's rendering output pipeline components. The samplers 2154, L1 cache 2151, texture cache 2158, and graphics cores 2152A-2152B can each have separate memory access paths. Optionally, the texture cache 2158 can also be configured as a sampler cache.

[0293] The render output pipeline 2170 may include a rasterizer and depth test component 2173 that converts vertex-based objects into associated pixel-based representations. The rasterizer logic may include a windower / masker unit for performing fixed-function triangle and line rasterization. In some embodiments, an associated render cache 2178 and depth cache 2179 may also be available. A pixel operation component 2177 performs pixel-based operations on data, although in some instances, pixel operations associated with 2D operations are performed by the 2D engine 2141 (e.g., using mixed bit-block image transfers) or, when displayed, by the display controller 2143 using an overlay display plane instead of pixel operations associated with 2D operations (e.g., using mixed bit-block image transfers). A shared L3 cache 2175 may be available to all graphics components, allowing data to be shared without using main system memory.

[0294] The media pipeline 2130 may include a media engine 2137 and a video front end 2134. The video front end 2134 may receive pipeline commands from the command streamer 2103. The media pipeline 2130 may include a separate command streamer. The video front end 2134 may process the media commands before sending them to the media engine 2137. The media engine 2137 may include thread generation functionality for generating threads for dispatching to the thread execution logic 2150 via the thread dispatcher 2131.

[0295] The graphics processor 2100 may include a display engine 2140. This display engine 2140 may be external to the graphics processor 2100 and may be coupled to the graphics processor via a ring or mesh interconnect 2102 or other interconnect bus or structure. The display engine 2140 may include a 2D engine 2141 and a display controller 2143. The display engine 2140 may contain dedicated logic capable of operating independently of the 3D pipeline. The display controller 2143 may be coupled to a display device (not shown), which may be a system-integrated display device such as in a laptop computer or an external display device attached via a display device connector.

[0296] The geometry pipeline 2120 and the media pipeline 2130 can be configured to perform operations based on multiple graphics and media programming interfaces and are not dedicated to any one application programming interface (API). Driver software for the graphics processor can convert API calls dedicated to a specific graphics or media library into commands that can be processed by the graphics processor. Support can be provided for the Open Graphics Library (OpenGL), Open Computing Language (OpenCL), and / or Vulkan graphics and computing APIs all from the Khronos Group. Support can also be provided for the Direct3D library from Microsoft. Combinations of these libraries can be supported. Support can also be provided for the open source computer vision library (OpenCV). If a mapping can be performed from the pipeline of a future API to the pipeline of the graphics processor, future APIs with compatible 3D pipelines will also be supported. Graphics pipeline programming

[0297] Figure 22A is used to describe, for example, Figure 16 and Figure 21 A block diagram of a graphics processor command format 2200 for programming a graphics processing pipeline is depicted. Figure 22B is a block diagram illustrating a graphics processor command sequence 2210 according to an embodiment. Figure 22A The solid-line boxes in illustrate components that are typically included in a graphics command, while the dashed lines include components that are optional or included only in a subset of the graphics commands. Figure 22A The graphics processor command format 2200 includes fields for identifying the client 2202 of the command, a command operation code (opcode 2204), and a data field 2206. Some commands also include a sub-opcode 2205 and a command size 2208.

[0298] Client 2202 may specify a client unit of a graphics device that processes command data. A graphics processor command parser may examine the client field of each command to adjust further processing of the command and route the command data to the appropriate client unit. A graphics processor client unit may include a memory interface unit, a rendering unit, a 2D unit, a 3D unit, and a media unit. Each client unit may have a corresponding processing pipeline for processing commands. Once a command is received by a client unit, the client unit reads the opcode 2204 and sub-opcode 2205 (if present) to determine the operation to be performed. The client unit uses the information in the data field 2206 to execute the command. For some commands, command size 2208 is expected to explicitly specify the size of the command. The command parser may automatically determine the size of at least some of the commands based on the command opcode. Commands may be aligned using multiples of double words. Other command formats may also be used.

[0299] Figure 22B The flowchart in FIG2 illustrates a graphics processor command sequence 2210. Software or firmware of a data processing system featuring an exemplary graphics processor may utilize a version of the command sequence shown for establishing, executing, and terminating a set of graphics operations. The sample command sequence is shown and described for illustrative purposes only and is not limited to these specific commands or to such command sequences. Furthermore, commands in the command sequence may be issued as batches so that the graphics processor will process the sequence of commands at least partially concurrently.

[0300] Graphics processor command sequence 2210 may begin with a pipeline flush command 2212 to cause any active graphics pipeline to complete currently pending commands for that pipeline. Optionally, 3D pipeline 2222 and media pipeline 2224 may not operate concurrently. A pipeline flush is performed to cause active graphics pipelines to complete any pending commands. In response to a pipeline flush, the command parser for the graphics processor will suspend command processing until the active drawing engines complete pending operations and the associated read cache is invalidated. Optionally, any data in the render cache that is marked as "dirty" may be flushed to memory. Pipeline flush command 2212 may be used for pipeline synchronization or may be used before placing the graphics processor into a low-power state.

[0301] When a command sequence requires the graphics processor to explicitly switch between pipelines, a pipeline select command 2213 may be used. A pipeline select command 2213 may be requested only once within an execution context before issuing a pipeline command, unless the context is issuing commands for both pipelines. A pipeline flush command 2212 may be requested immediately before a pipeline switch performed with the help of a pipeline select command 2213.

[0302] Pipeline control commands 2214 can configure the graphics pipeline for operation and can be used to program the 3D pipeline 2222 and the media pipeline 2224. Pipeline control commands 2214 can configure the pipeline state for the active pipeline. Pipeline control commands 2214 can be used for pipeline synchronization and can be used to flush data from one or more cache memories within the active pipeline before processing a batch of commands.

[0303] Commands associated with return buffer status 2216 can be used to configure a set of return buffers for a corresponding pipeline to write data. Some pipeline operations require the allocation, selection, or configuration of one or more return buffers to which the operation writes intermediate data during processing. A graphics processor may also use one or more return buffers to store output data and communicate across threads during execution. Return buffer status 2216 may include selecting the size and number of return buffers to be used for a set of pipeline operations.

[0304] The remaining commands in the command sequence differ based on the active pipeline for the operation. Based on pipeline determination 2220 , the command sequence is tailored for either the 3D pipeline 2222 to begin at 3D pipeline state 2230 or the media pipeline 2224 to begin at media pipeline state 2240 .

[0305] The commands used to configure the 3D pipeline state 2230 include 3D state setup commands for vertex buffer state, vertex element state, constant color state, depth buffer state, and other state variables to be configured before processing 3D primitive commands. The values ​​of these commands are determined at least in part based on the specific 3D API in use. The 3D pipeline state 2230 commands can also selectively disable or bypass certain pipeline elements if those elements will not be used.

[0306] 3D primitive 2232 commands can be used to submit 3D primitives for processing by the 3D pipeline. The commands and associated parameters passed to the graphics processor via the 3D primitive 2232 commands are forwarded to the vertex acquisition function in the graphics pipeline. The vertex acquisition function uses the 3D primitive 2232 command data to generate vertex data structures. The vertex data structures are stored in one or more return buffers. The 3D primitive 2232 commands can be used to perform vertex operations on the 3D primitives via vertex shaders. To process the vertex shaders, the 3D pipeline 2222 dispatches shader execution threads to the graphics processor execution units.

[0307] The 3D pipeline 2222 can be triggered by executing 2234 commands or events. Registers can be written to trigger command execution. Execution can be triggered by a 'go' or 'kick' command in a command sequence. Pipeline synchronization commands can be used to trigger command execution to flush the command sequence through the graphics pipeline. The 3D pipeline will perform geometry processing on 3D primitives. Once the operation is completed, the resulting geometry object is rasterized and the pixel engine colors the resulting pixels. For those operations, additional commands for controlling pixel shading and pixel backend operations can also be included.

[0308] When performing media operations, the graphics processor command sequence 2210 can follow the media pipeline 2224 path. Generally, the specific use and manner of programming the media pipeline 2224 depends on the media or compute operation to be performed. During media decoding, certain media decoding operations can be offloaded to the media pipeline. The media pipeline can also be bypassed and the media decoding can be performed in whole or in part using resources provided by one or more general-purpose processing cores. The media pipeline can also include elements for general-purpose graphics processor unit (GPGPU) operations, where the graphics processor is used to perform SIMD vector operations using compute shader programs that are not explicitly related to the rendering of graphics primitives.

[0309] The media pipeline 2224 can be configured in a similar manner to the 3D pipeline 2222. A set of commands used to configure the media pipeline state 2240 is dispatched or placed into a command queue before the media object commands 2242. The commands for the media pipeline state 2240 can include data used to configure the media pipeline elements that will be used to process the media objects. This includes data used to configure the video decoding and video encoding logic within the media pipeline, such as the encoding or decoding format. The commands for the media pipeline state 2240 can also support the use of one or more pointers to "indirect" state elements that contain batches of state settings.

[0310] Media object commands 2242 may supply a pointer to a media object for processing by the media pipeline. A media object includes a memory buffer containing the video data to be processed. Optionally, all media pipeline states must be valid before issuing a media object command 2242. Once the pipeline state is configured and the media object command 2242 is queued, the media pipeline 2224 is triggered by an execute command 2244 or an equivalent execute event (e.g., a register write). The output from the media pipeline 2224 may then be post-processed by operations provided by the 3D pipeline 2222 or the media pipeline 2224. GPGPU operations may be configured and executed in a similar manner to media operations. Graphics software architecture

[0311] Figure 23An exemplary graphics software architecture for data processing system 2300 is illustrated. Such a software architecture may include a 3D graphics application 2310, an operating system 2320, and a processor 2330. Processor 2330 may include a graphics processor 2332 and one or more general-purpose processor cores 2334. Processor 2330 may be a variant of, and may be used in place of, one of processor(s) 1402 or any other of the processors described herein. Thus, the disclosure of any features in combination with processor(s) 1402 or any other of the processors described herein also discloses the corresponding combination with graphics processor 2332, but is not limited to such. Furthermore, Figure 23 Elements having the same or similar names as elements of any other figure herein describe the same elements in that other figure, may operate or function in a similar manner as that one, may include the same components, and may be linked to other entities as those described elsewhere in this document, but are not limited to such. 3D graphics application 2310 and operating system 2320 are each executed in system memory 2350 of the data processing system.

[0312] 3D graphics application 2310 may include one or more shader programs including shader instructions 2312. The shader language instructions may be in a high-level shader language, such as Direct3D's High-Level Shader Language (HLSL), OpenGL Shader Language (GLSL), or the like. The application may also include executable instructions 2314 in a machine language suitable for execution by a general-purpose processor core(s) 2334. The application may also include graphics objects 2316 defined by vertex data.

[0313] The operating system 2320 may be from Microsoft Corporation. Operating system, proprietary UNIX-like operating system or open source UNIX-like operating system using a variant of the Linux kernel. Operating system 2320 can support graphics API 2322, such as Direct3D API, OpenGL API or Vulkan API. When the Direct3D API is in use, operating system 2320 uses front-end shader compiler 2324 to compile any shader instructions 2312 using HLSL into a lower-level shader language. Compilation can be just-in-time (JIT) compilation or the application can perform shader precompilation. During the compilation of 3D graphics application 2310, high-level shaders can be compiled into low-level shaders. Shader instructions 2312 can be provided in an intermediate form such as a version of the Standard Portable Intermediate Representation (SPIR) used by the Vulkan API.

[0314] The user-mode graphics driver 2326 may include a back-end shader compiler 2327 for converting shader instructions 2312 into a hardware-specific representation. When the OpenGL API is in use, shader instructions 2312 in the GLSL high-level language are passed to the user-mode graphics driver 2326 for compilation. The user-mode graphics driver 2326 may use operating system kernel-mode functions 2328 to communicate with a kernel-mode graphics driver 2329. The kernel-mode graphics driver 2329 may communicate with the graphics processor 2332 to dispatch commands and instructions. IP core implementation

[0315] One or more aspects may be implemented by representative code stored on a machine-readable medium that represents and / or defines logic within an integrated circuit (such as a processor). For example, a machine-readable medium may include instructions representing various logic within a processor. When read by a machine, the instructions may cause the machine to manufacture logic for performing the techniques described herein. Such representations (referred to as "IP cores") are reusable units of logic for an integrated circuit that may be stored on a tangible, machine-readable medium as a hardware model describing the structure of the integrated circuit. The hardware model may be supplied to various customers or manufacturing facilities that load the hardware model on a manufacturing machine that manufactures the integrated circuit. The integrated circuit may be manufactured so that the circuit performs the operations described in association with any of the embodiments described herein.

[0316] Figure 242 is a block diagram illustrating an IP core development system 2400 that can be used to manufacture integrated circuits to perform operations according to an embodiment. The IP core development system 2400 can be used to generate modular, reusable designs that can be incorporated into larger designs or used to build entire integrated circuits (e.g., SoC integrated circuits). A design facility 2430 can generate a software simulation 2410 of the IP core design in a high-level programming language (e.g., C / C++). The software simulation 2410 can be used to design, test, and verify the behavior of the IP core using a simulation model 2412. The simulation model 2412 can include functional simulation, behavioral simulation, and / or timing simulation. A register transfer level design (RTL design 2415) can then be created or synthesized from the simulation model 2412. The RTL design 2415 is an abstraction of the behavior of the integrated circuit (including associated logic executed using the modeled digital signals) that models the flow of digital signals between hardware registers. In addition to the RTL design 2415, lower-level designs at the logic level or transistor level can also be created, designed, or synthesized. As a result, the specific details of the initial design and simulation can vary.

[0317] The RTL design 2415 or equivalent can be further synthesized by the design facility into a hardware model 2420, which can be in hardware description language (HDL) or other representation of physical design data. The HDL can be further simulated or tested to verify the IP core design. Non-volatile memory 2440 (e.g., a hard disk, flash memory, or any non-volatile storage medium) can be used to store the IP core design for delivery to a manufacturing facility 2465. The manufacturing facility 2465 can be a third-party manufacturing facility. Alternatively, the IP core design can be transmitted via a wired connection 2450 or a wireless connection 2460 (e.g., via the Internet). The manufacturing facility 2465 can then manufacture an integrated circuit based at least in part on the IP core design. The manufactured integrated circuit can be configured to perform operations according to at least one embodiment described herein.

[0318] Figure 25AA cross-sectional side view of a package assembly 2590 of an integrated circuit comprising multiple units of hardware logic chiplets connected to a substrate 2580 (e.g., a base die) is illustrated. Graphics processing units, parallel processors, and / or computational accelerators as described herein can be composed of different silicon chiplets that are manufactured separately. In this context, a chiplet is an at least partially packaged integrated circuit that includes different units of logic that can be assembled into a larger package with other chiplets. Different combinations of chiplets with different IP core logic can be assembled into a single device. In addition, chiplets can be integrated into a base die or base chiplet using active interposer technology. The concepts described herein enable interconnection and communication between different forms of IP within a GPU. IP cores can be manufactured using different process technologies and can be composed during manufacturing, which avoids the complexity of fusing multiple IP (especially on large SoCs with multiple flavors of IP) into the same manufacturing process. Enabling the use of multiple process technologies improves time to market and provides a cost-effective way to create multiple product SKUs. Additionally, disaggregated IP can be more easily power-gated independently, allowing components not used for a given workload to be shut down, thereby reducing overall power consumption.

[0319] In various embodiments, the package assembly 2590 can include a fewer or greater number of components and chiplets interconnected by an interconnect structure 2585 or a bridge structure 2587. The bridge structure 2587 can be used to facilitate point-to-point interconnection between, for example, a logic or I / O chiplet 2574 and a memory chiplet 2575. In some implementations, the bridge structure 2587 can also be embedded within the substrate 2580. The chiplets within the package assembly 2590 can have a 2.5D arrangement using Chip-on-Wafer-on-Substrate (CoWoS) stacking, where multiple dies are stacked side-by-side on a silicon interposer that includes through-silicon vias (TSVs) to couple the chiplets to the substrate 2580, which includes electrical connections to the package interconnects 2583.

[0320] In one embodiment, the silicon interposer is an active interposer 2589 that includes embedded logic in addition to the TSVs. In such an embodiment, the chiplets within the package assembly 2590 are arranged on top of the active interposer 2589 using 3D face-to-face die stacking. The active interposer 2589 may also include I / O hardware logic 2591, cache memory 2592, and other hardware logic 2593 in addition to the interconnect structure 2585 and the bridge structure 2587. The interconnect structure 2585 enables communication between the various logic chiplets within the active interposer 2589. The interconnect structure 2585 may be a NoC interconnect or another form of packet switching fabric that exchanges data packets between components of the package assembly. For complex assemblies, the interconnect structure 2585 may be a dedicated chiplet that enables communication between the various hardware logic of the package assembly 2590.

[0321] The hardware logic chiplets may include dedicated hardware logic chiplets 2572, logic or I / O chiplets 2574, and / or memory chiplets 2575. The dedicated hardware logic chiplets 2572 and logic or I / O chiplets 2574 may be implemented at least in part in configurable logic or fixed-functionality logic hardware and may include one or more portions of any of the processor core(s), graphics processor(s), parallel processor(s), or other accelerator devices described herein. The memory chiplets 2575 may be DRAM (e.g., GDDR, HBM) memory or cache (SRAM) memory. The cache memory 2592 within the active interposer 2589 (or substrate 2580) may serve as a global cache for the package assembly 2590, as part of a distributed global cache, or as a dedicated cache for the interconnect structure 2585.

[0322] Each chiplet can be fabricated as a separate semiconductor die, and each chiplet can be coupled to a base die embedded within or coupled to a substrate 2580. Coupling to the substrate 2580 can be performed with the aid of an interconnect structure 2573. The interconnect structure 2573 can be configured to route electrical signals between the various chiplets and logic within the substrate 2580. The interconnect structure 2573 can include interconnects such as, but not limited to, bumps or pillars. In some embodiments, the interconnect structure 2573 can be configured to route electrical signals, such as, for example, input / output (I / O) signals and / or power or ground signals associated with the operation of the logic, I / O, and memory chiplets. In one embodiment, an additional interconnect structure couples the active interposer 2589 to the substrate 2580.

[0323] Substrate 2580 may be an epoxy-based laminate substrate and / or may also include other suitable types of substrates. Package assembly 2590 may be connected to other electrical devices via package interconnects 2583. Package interconnects 2583 may be coupled to the surface of substrate 2580 to route electrical signals to other electrical devices such as a motherboard, other chipsets, or multi-chip modules.

[0324] The logic or I / O chiplet 2574 and the memory chiplet 2575 can be electrically coupled via a bridge structure 2587 configured to route electrical signals between the logic or I / O chiplet 2574 and the memory chiplet 2575. The bridge structure 2587 can be a dense interconnect structure that provides routing for electrical signals. The bridge structure 2587 can include a bridge substrate composed of glass or a suitable semiconductor material. Circuit routing features can be formed on the bridge substrate to provide chip-to-chip connections between the logic or I / O chiplet 2574 and the memory chiplet 2575. The bridge structure 2587 can also be referred to as a silicon bridge or an interconnect bridge. For example, the bridge structure 2587 is an embedded multi-die interconnect bridge (EMIB). Alternatively, the bridge structure 2587 can simply be a direct connection from one chiplet to another chiplet.

[0325] Figure 25B A package assembly 2594 including interchangeable chiplets 2595 according to an embodiment is illustrated. The interchangeable chiplets 2595 can be assembled into standardized chiplet sockets or chiplet receptacles on base chiplets 2596, 2598. The base chiplets 2596, 2598 can be coupled via a bridge interconnect 2597, which can be similar to other bridge interconnects described herein and can be, for example, EMIB. Memory chiplets can also be connected to logic or I / O chiplets via the bridge interconnect. The I / O and logic chiplets can communicate via the interconnect structure. The base chiplets can each support one or more sockets in a standardized format for one of logic or I / O or memory / cache.

[0326] SRAM and power delivery circuitry can be fabricated into one or more of the base chiplets 2596, 2598, which can be fabricated using a different process technology relative to the interchangeable chiplets 2595, stacked on top of the base chiplets. For example, the base chiplets 2596, 2598 can be fabricated using a larger process technology, while the interchangeable chiplets can be fabricated using a smaller process technology. One or more of the interchangeable chiplets 2595 can be memory (e.g., DRAM) chiplets. Different memory densities can be selected for the package assembly 2594 based on the power and / or performance targeted for the product using the package assembly 2594. Additionally, logic chiplets with different numbers of functional units of different types can be selected at assembly time based on the power and / or performance targeted for the product. Additionally, chiplets containing different types of IP logic cores can be inserted into the interchangeable chiplet sockets, enabling hybrid processor designs that can mix and match IP blocks of different technologies. Exemplary System-on-Chip Integrated Circuit

[0327] Figure 26 An exemplary integrated circuit that can be fabricated using one or more IP cores is described. Other logic and circuitry beyond that illustrated may be included, including additional graphics processors / cores, peripheral interface controllers, or general-purpose processor cores. Figure 26 Elements having the same or similar names as elements of any other figure in this document describe the same elements in that other figure, may operate or function in a similar manner as that one, may include the same components, and may be linked to other entities as those described elsewhere in this document, but are not limited to such.

[0328] The system-on-chip integrated circuit 2600 includes one or more application processors 2605 (e.g., CPUs), a graphics processor 2610, which may be a variant of the graphics processor(s) 1408 or may be a variant of any of the graphics processors described herein and may be used in place of any of the graphics processors described. Thus, the disclosure of any feature in combination with a graphics processor herein also discloses the corresponding combination with the graphics processor 2610, but is not limited to such. The system-on-chip integrated circuit 2600 may additionally include an image processor 2615 and / or a video processor 2620, any of which may be modular IP cores from the same or multiple different design facilities. The system-on-chip integrated circuit 2600 may include peripheral or bus logic, including a USB controller 2625, a UART controller 2630, an SPI / SDIO controller 2635, and I 2 S / I2 C controller 2640. In addition, the integrated circuit may include a display device 2645, which is coupled to a high-definition multimedia interface (HDMI) controller 2650 and a reliability, availability, and serviceability engine (RAS engine 2655). The RAS engine 2655 is used to identify potential failures that may occur during device operation to minimize the downtime that would result if those potential failures were to occur. Storage can be provided by a flash memory subsystem 2660 including flash memory and a flash memory controller. A memory interface can be provided by means of a memory controller 2665 for accessing SDRAM or SRAM memory devices. Some integrated circuits additionally include an embedded security engine 2670. Adjustments for kernels compiled on the device

[0329] In graphics processing operations, statically or just-in-time (JIT) compiled kernels may not be able to effectively handle dynamic workload characteristics that depend on the computational data, where such characteristics may include, for example, non-inferable dynamic shapes, inherent sparsity, and statistics that can be exploited to improve overall performance.

[0330] In graphics processing, kernels of multiple different related shapes (e.g., shapes associated with multiple different matrix operations) may be received for processing by a graphics processing unit (GPU). When there are kernels that arrive at one or more operational shapes, there is an attempt to typically optimize the compilation for one or more shapes. However, because some shapes are not fixed, there may be dynamic drift in the input. The data may be in CSR (compressed sparse row) format for GEMM (general matrix-matrix multiplication) with a specific data distribution being received. Although the examples described herein relate to matrix multiplication, embodiments are not limited to these operations and may also include, for example, tensor operations where the computation kernels are associated with different tensor shapes in the computation.

[0331] In this example, a core generator may currently be set up for a specific shape, such as a 2K by 8K GEMM (or any other shape), resulting in specific operations for compilation. If a 4K by 4K GEMM is received instead, this will result in different compilation operations. Generally speaking, it is impossible to maintain compilation without modification every time the shape changes.

[0332] In the case of associated shape changes, the system can pre-compile to generate a binary output, where the compiler takes the final shape binary and applies it to the compilation. However, this comes with an overhead price, as it will usually require going back to the general processing unit (such as the CPU) to generate the changes.

[0333] In some embodiments, a device, system, or process provides JIT kernel tuning (also referred to herein as adjustments) for on-device compilation during computational graph execution. In some embodiments, a hardware mechanism is provided to implement compiled code fine-tuning (including pruning) of a scheduled workload command (DAG) based on intermediate data generated during the execution of the workload directed acyclic graph (DAG). In some embodiments, an on-device compiler is utilized in conjunction with kernel tuning. In some embodiments, JIT kernel adjustments are provided on the device during the execution of the computational graph. In some examples, the device may include processing resources, wherein the processing resources may include a low-power processor, a microcontroller, an accelerator, or other processing resources. As used herein, an accelerator refers to a device that accelerates processing operations performed by one or more processors, such as a CPU (central processing unit). An accelerator may include a GPU or other device.

[0334] In some embodiments, to enable the apparatus to make accurate and efficient decisions regarding compilation, adjustments are provided for compilation, where the adjustments may include modifications or adjustments to scheduled workload commands. An apparatus is to determine one or more characteristics of a kernel (which may include a shape associated with the kernel in question) and inform a circuit module about the compilation of the kernel in the generation of binary data. This operation includes the implementation of one or more algorithms to provide instructions for operation on a particular shape, as each shape will generally result in the generation of different code for efficient operation.

[0335] In embodiments, execution of the adjustment algorithm imposes a certain time cost. However, in embodiments, execution of one or more adjustment algorithms allows for the elimination of the need to return to the CPU or other general processing unit in response to each change in the shape of the input kernel, as the relevant compilation decisions are instead made on a device such as a GPU or other accelerator. This contrasts with conventional operations, in which a return to the CPU may occur with each change in the input shape because decisions need to be made regarding the selection of tensors and other factors to apply a matrix multiplication (matmul), which would utilize binary data based on the relevant shapes.

[0336] In some embodiments, processing resources such as GPUs or other accelerators include hardware that implements changes in binary code generation. In a first implementation, certain existing features, such as existing cores, can be utilized in code generation, wherein one or more additional features and instructions are implemented to achieve efficient operation in the processing resources. In a second implementation, additional cores can be included in the processing resources (which may generally be referred to as command processors in this article) used for compiling operations. The additional command processors can be RISC (reduced instruction set computer) processors or other types of cores configured to perform compiling operations. The processing resources can also include a scheduler for scheduling operations, and in an embodiment, the command processor can receive scheduling data from the scheduler to be applied to program the command processor to perform operations.

[0337] Figure 27 is an illustration of a device or system for providing kernel tuning using graphics execution, according to some embodiments. In some embodiments, device or system 2700 includes one or more processing resources 2710, which may include, but are not limited to, one or more general-purpose processors 2715 (e.g., CPUs (central processing units)), one or more graphics processing units (GPUs) or other accelerators 2720, and one or more other processing resources 2740 (e.g., microcontrollers or low-power processors). Device or system 2700 also includes memory 2750, such as DRAM (dynamic random access memory) or other memory technology, for storing data for processing; and storage 2755, such as an SSD (solid-state drive) or other non-volatile storage device, for long-term storage of data.

[0338] In some embodiments, the GPU or other accelerator 2720 includes a compilation circuit module 2730 for providing support for on-device compilation, the compilation circuit module 2730 being used to implement tuning of the kernel using graphics execution, such as fine-tuning of the kernel relative to the GPU or other accelerator 2720. The tuning of the kernel may include tuning the kernel in response to a shape change associated with a received kernel (including a shape change associated with matrix multiplication), which may be used to optimize the compilation of the kernel for execution. The GPU or other accelerator 2720 also includes a scheduler 2735 for providing scheduling of instructions, wherein the scheduler 2735 may provide scheduling data associated with the compute kernel, which is used to support the compilation of the kernel.

[0339] The other elements of the device or system 2700 and the one or more processing resources 2710 may be as described above with respect to Figures 1 to 26 As stated.

[0340] Figure 28is an illustration of a compute kernel receiving changes for compilation at a processor core according to some embodiments. Figure 28 As shown in , a processor core 2800 of a processing resource may receive computation kernels of varying shapes over time (t) for compilation. The illustrated kernels (Compute Kernel-1, Compute Kernel-2, Compute Kernel-3, Compute Kernel-4) may each represent a different matrix multiplication, such as in graphics processing. Therefore, each compute kernel may require a different compilation to optimize the generation of binary output.

[0341] In conventional operation, it may be necessary to provide communication with a general purpose processor upon each change in the associated shape to determine an optimized compilation of the compute kernel. In some embodiments, processing resources instead support on-device compilation, which includes the ability to apply one or more algorithms to identify adjustments to the compilation of each received compute kernel.

[0342] Figure 29 is an illustration of a circuit module adapted on a device supporting a kernel for compilation according to some embodiments. In some embodiments, Figure 27 The device circuit module 2910 of the compilation circuit module 2730 shown in FIG is provided within one or more processing resources 2900 (e.g., a GPU or other accelerator) to provide hardware support for on-device adjustments of compiled kernels to generate compiled binary output 2940.

[0343] In some embodiments, the device circuit module 2910 of one or more processing resources 2900 includes a kernel evaluation circuit module 2915 that evaluates input kernels 2930 (e.g., the kernels shown (Kernel-1 and Kernel-2)) to identify one or more characteristics of the input kernels 2930 for use in adjusting the compilation of the kernels. The one or more characteristics include determining the associated shape of the input kernels, such as the shape associated with the matrix multiplication performed based on the contents of the kernels. The characteristics may also include sparsity information about the matrix, statistics about the matrix, and other characteristics. For example, the kernel evaluation circuit module 2915 may evaluate the input kernels 2930 to determine that Kernel-1 has a first shape (Shape-1 for a first matrix multiplication) and that Kernel-2 has a second shape (Shape-2 for a second matrix multiplication). Depending on the input received, the associated shape of the kernels may potentially vary with each kernel received.

[0344] In some embodiments, the device circuit module 2910 further includes a device compiler circuit module 2920 that performs kernel compilation. The device compiler circuit module 2920 includes the ability to adjust received kernels based at least in part on one or more determined characteristics of the kernels (including the determined shape of such kernels) as established by the kernel evaluation circuit module 2915. In some embodiments, the one or more processing resources 2900 further include a scheduler, e.g. Figure 27 , wherein the device compiler circuit module 2920 is to further coordinate the compilation of the compute kernel based on the scheduling data associated with the compute kernel received from the scheduler. The device compiler circuit module 2920 may include support for running compiler software by one or more processing resources 2900.

[0345] In some embodiments, the device compiler circuit module 2920 adjusts the compilation of the received kernel by applying one or more algorithms that optimize the kernel compilation based at least in part on the determined characteristics of the kernel. In some embodiments, the device circuit module 2910 includes a kernel algorithm circuit module 2925 to facilitate the selection and application of one or more algorithms in adjusting the kernel compilation. For example, the kernel algorithm circuit module 2925 may include storage for one or more algorithms for application in adjusting the kernel compilation.

[0346] Figure 30 3010 is an illustration of an on-device compiled process including a kernel according to some embodiments. In some embodiments, code fine-tuning and / or pruning of scheduled workload commands is implemented based on intermediate data generated during the workload DAG. As illustrated, a JIT (Just-in-Time) compute kernel is processed using an on-device compiler 3015, which can run on a compute core of a processing resource 3000 or a separately reserved compute core (command processor) 3010, wherein the processing resource 3000 can include Figure 27 , wherein the processing resources in one or more processing resources 2710 are shown in FIG. This operation allows the compute kernel 3020 to be adjusted to improve the handling of dynamic workload characteristics that depend on the data being computed.

[0347] In some embodiments, the compilation of the compute kernel 3020 can be based at least in part on characteristic data 3025, which provides information about one or more determined characteristics of the kernel. In particular, the one or more characteristics can include a determined shape associated with the kernel, such as a shape for matrix multiplication or other operations.

[0348] In some embodiments, compilation of the received JIT kernel produces a compiled kernel for execution. Figure 3030. In the figure, the compiled kernel is shown as compiled kernel 1 and compiled kernel 2, which are intended to represent binary data generated in the compilation of the service of the JIT kernel, where the computation kernel may include different characteristics, including different associated shapes. The kernel can then be executed for an associated operation such as matrix multiplication. This is shown as the compiled kernel receiving certain inputs (e.g., input data 3035 (e.g., matrix values) and related metadata 3030) and generating an output of the operation 3040, such as the result of the matrix multiplication or other operation.

[0349] Figure 31 31 is an illustration of a process for on-device tuning of a compiled kernel, according to some embodiments. In process 3100, at 3105, a computing kernel is received at a processing resource in a computing system for on-device compilation. In some embodiments, one or more processing resources of the computing system include a circuit module that supports on-device compilation of the computing kernel. In some embodiments, at 3110, the received computing kernel is evaluated using the circuit module to determine one or more characteristics of the computing kernel, where the determined one or more characteristics may include, but are not limited to, a shape associated with the computing kernel. The associated shape may relate to the shape of a matrix in a computation such as matrix multiplication.

[0350] In some embodiments, data regarding the determined one or more characteristics of the compute kernel is provided to the device compiler circuit module 3115, wherein the determined characteristics may include a shape associated with the compute kernel. In some embodiments, at 3120, the device compiler circuit module adjusts the compilation of the compute kernel based at least in part on the data regarding the determined one or more characteristics of the compute kernel. The adjustment may include modifying or adjusting the scheduling of workload commands associated with the compute kernel. In some embodiments, at 3125, the adjustment to the compilation of the compute kernel may include selecting one or more algorithms based at least on the determined characteristics of the compute kernel, and applying the selected one or more algorithms in the compilation of the compute kernel. The selection and application of the one or more algorithms may include utilizing support from the kernel algorithm circuit module.

[0351] Depending on the implementation and nature of the received kernel, process 3100 may continue to generate a binary output based on the processed compute kernel at 3130 and perform the associated operation of the compute kernel at 3035. The execution of the operation may include performing matrix multiplication as needed. If there are additional compute kernels for processing at 3140, process 3100 may continue to evaluate the next subsequent compute kernel at 3110. If not, the process may continue with other operations at 3150.

[0352] The following examples relate to certain embodiments:

[0353] In Example 1, an apparatus includes computer memory storing data for processing; and processing resources including a graphics processing unit (GPU), the GPU including a compilation circuit module; wherein the compilation circuit module includes: a kernel evaluation circuit module for evaluating a received computing kernel for compilation and determining one or more characteristics of the computing kernel; and an apparatus compiler circuit module for supporting compilation of the computing kernel, wherein the apparatus compiler circuit module is to adjust the compilation of the computing kernel based at least in part on the one or more determined characteristics of the computing kernel.

[0354] In Example 2, for the apparatus as provided in Example 1, the one or more determined characteristics of the compute kernel include a shape associated with the compute kernel.

[0355] In Example 3, for the apparatus as provided in Example 1 or 2, the shape associated with the compute kernel is a shape for a matrix multiplication operation.

[0356] In Example 4, for the apparatus provided in any of Examples 1 to 3, the adjusting of the compilation of the compute kernel is performed using application of one or more algorithms, and wherein selection of the one or more algorithms is based at least in part on one or more determined characteristics of the compute kernel.

[0357] In Example 5, for the apparatus provided in any one of Examples 1 to 4, the GPU further includes a scheduler, and wherein the device compiler circuit module further coordinates compilation of the compute kernel based on scheduling data associated with the compute kernel received from the scheduler.

[0358] In Example 6, for the apparatus as provided in any of Examples 1 to 5, wherein the device compiler circuit module is to perform the adjustment of the computation kernel compilation without communicating with the general processing unit.

[0359] In Example 7, for the apparatus provided in any one of Examples 1 to 6, the compilation circuit module is located in a first core of the plurality of cores of the GPU.

[0360] In Example 8, for the device as provided in Example 7, the first core is a reserved core of the GPU dedicated to supporting the compile operation.

[0361] In Example 9, a method includes receiving a compute kernel to compile for processing at a graphics processing unit (GPU), the GPU including a compile circuit module for supporting compile operations; evaluating the compute kernel with an evaluation circuit module of the GPU; determining one or more characteristics of the compute kernel based on the evaluation of the compute kernel; adjusting the compilation of the compute kernel based at least in part on the determined one or more characteristics of the compute kernel; and performing the compilation of the compute kernel based at least in part on the adjustment of the compilation.

[0362] In Example 10, for the method as provided in Example 9, the one or more determined characteristics of the computation kernel include a shape associated with the computation kernel.

[0363] In Example 11, for the method as provided in Example 9 or 10, the shape associated with the computation kernel is a shape for a matrix multiplication operation.

[0364] In Example 12, for the method as provided in any one of Examples 9 to 11, the method further includes selecting one or more algorithms for compilation of the compute kernel, wherein the adjusting of the compilation of the compute kernel includes application of the one or more algorithms.

[0365] In Example 13, for the method as provided in any one of Examples 9 to 12, the compilation of the compute kernel is further adjusted based on scheduling data associated with the compute kernel received from the scheduler.

[0366] In Example 14, for the method as provided in any one of Examples 9 to 13, the adjusting of the compilation of the compute kernel is performed without communicating with the general processing unit.

[0367] In Example 15, one or more non-transitory computer-readable storage media having executable computer program instructions stored thereon, which, when executed by one or more processors, cause the one or more processors to perform operations including receiving a computing kernel to be compiled for processing at a graphics processing unit (GPU), the GPU including a compilation circuit module for supporting compilation operations; evaluating the computing kernel using an evaluation circuit module of the GPU; determining one or more characteristics of the computing kernel based on the evaluation of the computing kernel; adjusting the compilation of the computing kernel based at least in part on the determined one or more characteristics of the computing kernel; and performing the compilation of the computing kernel based at least in part on the adjustment of the compilation.

[0368] In Example 16, for the storage medium as provided in Example 15, the one or more determined characteristics of the compute kernel include a shape associated with the compute kernel.

[0369] In Example 17, for the storage medium as provided in Example 15 or 16, the shape associated with the computation kernel is a shape for a matrix multiplication operation.

[0370] In Example 18, for the storage medium provided in any one of Examples 15 to 17, the executable computer program instructions further include instructions that, when executed by the one or more processors, cause the one or more processors to perform operations including selecting one or more algorithms for compilation of a computing kernel, wherein the adjustment of the compilation of the computing kernel includes application of the one or more algorithms.

[0371] In Example 19, for the storage medium as provided in any one of Examples 15 to 18, the compilation of the compute kernel is further coordinated based on scheduling data associated with the compute kernel received from the scheduler.

[0372] In Example 20, for the storage medium as provided in any one of Examples 15 to 19, the adjusting of the compilation of the compute kernel is performed without communicating with the general processing unit.

[0373] In Example 21, a device includes components for receiving a compute kernel to compile for processing at a graphics processing unit (GPU), the GPU including a compile circuit module for supporting compile operations; components for evaluating the compute kernel using an evaluation circuit module of the GPU; determining one or more characteristics of the compute kernel based on the evaluation of the compute kernel; components for adjusting the compilation of the compute kernel based at least in part on the determined one or more characteristics of the compute kernel; and components for performing the compilation of the compute kernel based at least in part on the adjustment of the compilation.

[0374] In Example 22, for the apparatus as provided in Example 21, the one or more determined characteristics of the compute kernel include a shape associated with the compute kernel.

[0375] In Example 23, for the apparatus as provided in Example 21 or 22, the shape associated with the compute kernel is a shape for a matrix multiplication operation.

[0376] In Example 24, for the apparatus as provided in any one of Examples 21 to 23, the apparatus further comprises means for selecting one or more algorithms to compile the compute kernel, wherein the adjusting of the compilation of the compute kernel comprises application of the one or more algorithms.

[0377] In Example 25, for the apparatus as provided in any one of Examples 21 to 24, the compilation of the compute kernel is further coordinated based on scheduling data associated with the compute kernel received from the scheduler.

[0378] In Example 26, for the apparatus as provided in any one of Examples 21 to 25, the adjusting of the compilation of the compute kernel is performed without communicating with the general processing unit.

[0379] Although the description and illustrations of the embodiments provided herein describe specific components, those skilled in the art will appreciate that such components may be combined into fewer elements or separated into a greater number of components as desired or convenient for a particular implementation. For example, the components described for providing input classification and buffering may be expressed as a first classification component and a second buffering component.

[0380] In the above description, for the purpose of explanation, numerous specific details have been set forth in order to provide a thorough understanding of the embodiments. However, it will be apparent to those skilled in the art that the embodiments may be practiced without some of these specific details. In other examples, well-known structures and devices are shown in block diagram form. Intermediate structures may exist between the components shown. Components described or illustrated herein may have additional inputs or outputs that are not shown or described.

[0381] Various embodiments may include various processes. These processes may be performed by hardware components or may be embodied in computer programs or machine-executable instructions that may be used to cause a general-purpose or special-purpose processor or logic circuit programmed with the instructions to perform the processes. Alternatively, the processes may be performed by a combination of hardware and software.

[0382] Portions of various embodiments may be provided as a computer program product that may include a computer-readable medium having computer program instructions stored thereon, the computer program instructions being used to program a computer (or other electronic device) for execution by one or more processors to perform processes according to certain embodiments. The computer-readable medium may include, but is not limited to, a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic or optical card, flash memory, or other types of computer-readable media suitable for storing electronic instructions. In addition, embodiments may also be downloaded as a computer program product, wherein the program may be transferred from a remote computer to a requesting computer.

[0383] Many of the methods are described in their most basic form, but procedures may be added to or deleted from any of the methods, and information may be added to or subtracted from any of the described messages, without departing from the basic scope of the present embodiments. It will be apparent to those skilled in the art that many further modifications and adaptations are possible. The specific embodiments are provided not to limit the concepts, but to illustrate them. The scope of the embodiments is to be determined not by the specific examples provided above, but only by the claims that follow.

[0384] If element "A" is said to be coupled to or with element "B," element A may be directly coupled to element B or indirectly coupled through, for example, element C. When the specification or claims state that component, feature, structure, process, or characteristic A "causes" component, feature, structure, process, or characteristic B, it means that "A" is at least partially responsible for "B," but there may also be at least one other component, feature, structure, process, or characteristic that contributes to causing "B." If the specification indicates that a component, feature, structure, process, or characteristic "may," "might," or "could" be included, that particular component, feature, structure, process, or characteristic is not required to be included. If the specification or claims refer to "a" or "an" element, this does not mean that only one of the described elements is present.

[0385] An embodiment is an implementation or example. Reference in the specification to "an embodiment," "one embodiment," "some embodiments," or "other embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least some of the embodiments, but not necessarily in all of the embodiments. The various appearances of "an embodiment," "one embodiment," or "some embodiments" do not necessarily refer to the same embodiment. It should be appreciated that in the foregoing description of exemplary embodiments, various features are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in understanding one or more of the various novel aspects. This method of disclosure, however, should not be interpreted as reflecting an intention that the claimed embodiments require more features than expressly recited in the claims. On the contrary, as reflected in the following claims, the novel aspects are less than all the features of a single foregoing disclosed embodiment. Accordingly, the claims are hereby expressly incorporated into this specification, with the claims themselves serving as separate embodiments.

[0386] The foregoing description and drawings are to be regarded as illustrative rather than restrictive. It will be understood by those skilled in the art that various modifications and changes may be made to the embodiments described herein without departing from the broader spirit and scope of the features set forth in the appended claims.

Claims

1. A device comprising: Computer memory used to store data for processing; as well as Processing resources including a graphics processing unit (GPU), the GPU including a compilation circuit module; The compiling circuit module includes: a kernel evaluation circuit module for evaluating a computing kernel received for compilation and determining one or more characteristics of the computing kernel, and A device compiler circuit module is configured to support compilation of the compute kernel, wherein the device compiler circuit module is to adjust the compilation of the compute kernel based at least in part on the one or more characteristics of the compute kernel.

2. The device according to claim 1, wherein The one or more characteristics of the computational kernel include a shape associated with the computational kernel.

3. The device according to claim 2, wherein The shape associated with the computation kernel is a shape for matrix multiplication operations.

4. The apparatus according to claim 1, wherein The compiled adjustment of the compute kernel is performed utilizing application of one or more algorithms, and wherein selection of the one or more algorithms is based at least in part on the one or more characteristics of the compute kernel.

5. The apparatus according to claim 1, wherein The GPU also includes a scheduler, and wherein the device compiler circuitry module is to coordinate the compilation of the compute kernel further based on scheduling data associated with the compute kernel received from the scheduler.

6. The apparatus according to claim 1, wherein The device compiler circuit module is to perform the compiled adjustment of the compute kernel without communicating with a general purpose processing unit.

7. The apparatus according to claim 1, wherein The compiling circuit module is located in a first core among multiple cores of the GPU.

8. The apparatus according to claim 7, wherein The first core is a reserved core of the GPU dedicated to supporting compile operations.

9. A method comprising: receiving a computation kernel to compile for processing at a graphics processing unit (GPU), the GPU including a compile circuit module for supporting the compile operation; evaluating the computation kernel using an evaluation circuit module of the GPU; determining one or more characteristics of the computing kernel based on the evaluation of the computing kernel; adjusting compilation of the compute kernel based at least in part on the one or more characteristics of the compute kernel; as well as Compilation of the compute kernel is performed based at least in part on the compiled adjustments.

10. The method according to claim 9, wherein: The one or more characteristics of the computational kernel include a shape associated with the computational kernel.

11. The method according to claim 10, wherein: The shape associated with the computation kernel is a shape for matrix multiplication operations.

12. The method according to claim 9, further comprising: One or more algorithms are selected for use in compilation of the compute kernel, wherein adjustment of the compilation of the compute kernel includes application of the one or more algorithms.

13. The method according to claim 9, wherein: Adjusting the compilation of the compute kernel is also based on scheduling data associated with the compute kernel received from a scheduler.

14. The method according to claim 9, wherein The compiling and adjusting of the computing kernel is performed without communicating with a general purpose processing unit.

15. One or more non-transitory computer-readable storage media having executable computer program instructions stored thereon that, when executed by one or more processors, cause the one or more processors to perform operations comprising: receiving a computation kernel to compile for processing at a graphics processing unit (GPU), the GPU including a compile circuit module for supporting the compile operation; evaluating the computation kernel using an evaluation circuit module of the GPU; determining one or more characteristics of the computing kernel based on the evaluation of the computing kernel; adjusting compilation of the compute kernel based at least in part on the one or more characteristics of the compute kernel; as well as Compilation of the compute kernel is performed based at least in part on the compiled adjustments.

16. The one or more non-transitory computer-readable storage media of claim 15, wherein: The one or more characteristics of the computational kernel include a shape associated with the computational kernel.

17. The one or more non-transitory computer-readable storage media of claim 16, wherein: The shape associated with the computation kernel is a shape for matrix multiplication operations.

18. The one or more non-transitory computer-readable storage media of claim 15, wherein: The executable computer program instructions also include instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising: One or more algorithms are selected for use in compilation of the compute kernel, wherein adjustment of the compilation of the compute kernel includes application of the one or more algorithms.

19. The one or more non-transitory computer-readable storage media of claim 15, wherein: Adjusting the compilation of the compute kernel is also based on scheduling data associated with the compute kernel received from a scheduler.

20. The one or more non-transitory computer-readable storage media of claim 15, wherein: The compiled adjustments of the computing kernel are performed without communicating with a general purpose processing unit.

21. A device comprising: means for receiving a computation kernel for compilation at a graphics processing unit (GPU) for processing, the GPU including a compilation circuit module for supporting compilation operations; means for evaluating the computational kernel using an evaluation circuit module of the GPU; means for determining one or more characteristics of the computing kernel based on the evaluation of the computing kernel; means for adjusting compilation of the compute kernel based at least in part on the one or more characteristics of the compute kernel; as well as Means for performing compilation of the compute kernel based at least in part on the compiled adjustment.

22. The apparatus according to claim 21, wherein The one or more characteristics of the computational kernel include a shape associated with the computational kernel.

23. The apparatus of claim 21, wherein: The shape associated with the computation kernel is a shape for matrix multiplication operations.

24. The apparatus of claim 21, further comprising: Means for selecting one or more algorithms for use in compiling the compute kernel, wherein adjusting the compilation of the compute kernel comprises applying the one or more algorithms.

25. The apparatus of claim 21, wherein Adjusting the compilation of the compute kernel is further based on scheduling data associated with the compute kernel received from a scheduler.

Images