1. A processor, comprising:

ray tracing circuitry to execute a first set of one or more commands to traverse a ray through a bounding volume level BVH to identify BVH nodes and/or primitives intersected by the ray;

shader execution circuitry to execute one or more shaders, in response to a second set of one or more commands, to render a sequence of image frames based on the BVH nodes and/or primitives intersected by the ray; and

a media processor including motion estimation circuitry to execute a third set of one or more commands to perform non-local mean filtering based on average pixel values collected across the image frame sequence to remove noise from the image frame sequence.

2. The processor of claim 1, further comprising:

a command stream transmitter to stream the first set of commands to the ray tracing circuitry, to stream the second set of commands to the shader execution circuitry, and to stream the third set of commands to the media processing block.

3. The processor of claim 2, wherein the third set of commands comprises an extension to an Application Programming Interface (API) associated with the media processing block.

4. The processor of any of claims 1 to 3, wherein the non-local mean filtering of a target pixel comprises determining an average associated with all pixels across an image sequence weighted by similarity to the target pixel.

5. The processor of any one of claims 1 to 4, wherein the shader execution circuitry comprises a plurality of Execution Units (EU) to execute a plurality of different shaders to render the sequence of image frames.

6. The processor of any of claims 1 to 5, wherein the average pixel value is determined based on an evaluation of pixel blocks spread across the sequence of image frames.

7. The processor of claim 6, wherein the block of pixels comprises a macroblock.

8. The processor of claim 7, wherein a macroblock comprises one or more of: a 16 × 16 pixel block, an 8 × 8 pixel block, a 4 × 4 pixel block, a 16 × 8 pixel block, an 8 × 4 pixel block, or a 16 × 4 pixel block.

9. A method, comprising:

executing a first set of one or more commands on the ray tracing circuitry to traverse the ray through the bounding volume level BVH to identify BVH nodes and/or primitives intersected by the ray;

Executing one or more shaders, in response to a second set of one or more commands, to render a sequence of image frames based on the BVH nodes and/or primitives that intersect the ray; and

executing, on a media processor comprising a motion estimation circuit, a third set of one or more commands to perform non-local mean filtering based on average pixel values collected across the image frame sequence to remove noise from the image frame sequence.

10. The method of claim 9, further comprising:

the first set of commands is streamed to the ray tracing circuitry, the second set of commands is streamed to the shader execution circuitry, and the third set of commands is streamed to the media processing block.

11. The method of claim 10, wherein the third set of commands comprises an extension to an Application Programming Interface (API) associated with the media processing block.

12. The method of any of claims 9 to 11, wherein the non-local mean filtering of a target pixel comprises determining an average associated with all pixels across an image sequence weighted by similarity to the target pixel.

13. The method of any of claims 9 to 11, wherein the one or more shaders are to be executed over a plurality of Execution Units (EUs) to render the sequence of image frames.

14. The method of any of claims 9 to 11, wherein the average pixel value is determined based on an evaluation of pixel blocks spread across the sequence of image frames.

15. The method of claim 14, wherein the block of pixels comprises a macroblock.

16. The method of claim 15, wherein a macroblock comprises one or more of: a 16 × 16 pixel block, an 8 × 8 pixel block, a 4 × 4 pixel block, a 16 × 8 pixel block, an 8 × 4 pixel block, or a 16 × 4 pixel block.

17. A machine-readable medium having program code stored thereon, which when executed by a machine, causes the machine to perform operations comprising:

executing a first set of one or more commands on the ray tracing circuitry to traverse the ray through the bounding volume level BVH to identify BVH nodes and/or primitives intersected by the ray;

executing one or more shaders, in response to a second set of one or more commands, to render a sequence of image frames based on the BVH nodes and/or primitives that intersect the ray; and

Executing, on a media processor comprising a motion estimation circuit, a third set of one or more commands to perform non-local mean filtering based on average pixel values collected across the image frame sequence to remove noise from the image frame sequence.

18. The machine-readable medium of claim 17, further comprising:

the first set of commands is streamed to the ray tracing circuitry, the second set of commands is streamed to the shader execution circuitry, and the third set of commands is streamed to the media processing block.

19. The machine-readable medium of claim 18, wherein the third set of commands comprises an extension to an Application Programming Interface (API) associated with the media processing block.

20. The machine-readable medium of any of claims 17 to 19, wherein the non-local mean filtering of a target pixel comprises determining an average associated with all pixels across an image sequence weighted by similarity to the target pixel.

21. The machine readable medium of any of claims 17 to 20, wherein the one or more shaders execute on a plurality of Execution Units (EUs) to render the sequence of image frames.

22. The machine readable medium of any of claims 17 to 21, wherein the average pixel value is determined based on an evaluation of pixel blocks spread across the sequence of image frames.

23. The machine-readable medium of claim 22, wherein the block of pixels comprises a macroblock.

24. The machine-readable medium of claim 23, wherein a macroblock comprises one or more of: a 16 × 16 pixel block, an 8 × 8 pixel block, a 4 × 4 pixel block, a 16 × 8 pixel block, an 8 × 4 pixel block, or a 16 × 4 pixel block.

25. An apparatus, comprising:

means for executing a first set of one or more commands on the ray tracing circuitry to traverse a ray through the bounding volume level BVH, identifying BVH nodes and/or primitives intersected by the ray;

means for executing one or more shaders, in response to a second set of one or more commands, to render a sequence of image frames based on the BVH nodes and/or primitives that intersect the ray; and

means for executing a third set of one or more commands on a media processor including a motion estimation circuit to perform non-local mean filtering based on average pixel values collected across the sequence of image frames to remove noise from the sequence of image frames.

Background

Ray tracing is a technique in which light transmission is emulated by physics-based rendering. It has been widely used in movie rendering, and until a few years ago it was considered too resource intensive for real-time performance. One of the key operations in ray tracing is processing visibility queries for ray-scene intersections (referred to as "ray traversals") that compute ray-scene intersections by traversing and intersecting nodes in a bounding volume level (BVH).

Drawings

A better understanding of the present invention can be obtained from the following detailed description in conjunction with the following drawings, in which:

FIG. 1 is a block diagram of a computer system with a processor having one or more processor cores and a graphics processor;

2A-2D illustrate a computing system and graphics processor provided by embodiments described herein;

3A-3C illustrate block diagrams of additional graphics processor and compute accelerator architectures provided by embodiments described herein;

FIG. 4 is a block diagram of a graphics processing engine for a graphics processor;

5A-5B illustrate thread execution logic 500 including an array of processing elements employed in a graphics processor core according to embodiments described herein;

Fig. 6 illustrates a further execution unit 600 according to an embodiment;

FIG. 7 illustrates a graphics processor execution unit instruction format;

FIG. 8 is a block diagram of a graphics processor including a graphics pipeline, a media pipeline, a display engine, thread execution logic, and a render output pipeline;

FIG. 9A is a block diagram illustrating a graphics processor command format;

FIG. 9B is a block diagram illustrating a graphics processor command sequence;

FIG. 10 illustrates an exemplary graphical software architecture for a data processing system;

FIG. 11A illustrates an example IP core development feature of an embodiment;

11B-D illustrate various packaging features of different embodiments;

FIG. 12 illustrates an exemplary system-on-chip integrated circuit that may be fabricated using one or more IP cores;

FIG. 13 illustrates an exemplary graphics processor of a system-on-a-chip integrated circuit that may be fabricated using one or more IP cores;

the graphics processor 1340 of FIG. 14 includes the one or more MMUs 1320A-1320B, cache(s) 1325A-1325B, and circuit interconnect(s) 1330A-1330B of the graphics processor 1310 of FIG. 13;

FIG. 15 illustrates an architecture for performing initial training of a machine learning architecture;

FIG. 16 illustrates how the machine learning engine is continually trained and updated during runtime;

FIG. 17 illustrates how the machine learning engine is continually trained and updated during runtime;

18A-B illustrate how machine learning data is shared over a network; and

FIG. 19 illustrates a method for training a machine learning engine;

FIG. 20 illustrates how nodes exchange ghost region (ghost region) data to perform a distributed denoising operation;

FIG. 21 illustrates an architecture in which image rendering and denoising operations are distributed across multiple nodes;

FIG. 22 illustrates additional details of an architecture for distributed rendering and denoising;

FIG. 23 illustrates a method for performing distributed rendering and denoising;

FIG. 24 illustrates a machine learning method;

FIG. 25 illustrates a plurality of interconnected general purpose graphics processors;

FIG. 26 illustrates a set of convolutional layers and fully-connected layers for a machine-learning implementation;

FIG. 27 illustrates one example of a convolutional layer;

FIG. 28 illustrates an example of a set of interconnected nodes in a machine learning implementation;

FIG. 29 illustrates a training framework within which a neural network learns using a training data set;

FIG. 30A illustrates examples of model parallelism (model parallelism) and data parallelism (data parallelism);

FIG. 30B illustrates a system on a chip (SoC);

FIG. 31 illustrates a processing architecture including a ray tracing core and a tensor core;

FIG. 32 illustrates an example of a light beam;

FIG. 33 illustrates an apparatus for performing beam tracking;

FIG. 34 illustrates an example of a beam hierarchy;

FIG. 35 illustrates a method for performing beam tracking;

FIG. 36 illustrates an example of a distributed ray trace engine;

FIGS. 37-38 illustrate compression performed in a ray tracing system;

FIG. 39 illustrates a method implemented on a ray tracing architecture;

FIG. 40 illustrates an exemplary hybrid ray tracing device;

FIG. 41 illustrates a stack for ray tracing operations;

FIG. 42 illustrates additional details of the hybrid ray tracing device;

FIG. 43 illustrates a bounding volume hierarchy;

FIG. 44 illustrates call stack and walk state storage;

FIG. 45 illustrates a method for traversal and intersection;

46A-B illustrate how multiple dispatch cycles are required to execute certain shaders;

FIG. 47 illustrates how multiple shaders execute in a single dispatch cycle;

FIG. 48 illustrates how a single dispatch cycle executes multiple shaders;

FIG. 49 illustrates an architecture for executing ray tracing instructions;

FIG. 50 illustrates a method for executing a ray trace instruction within a thread;

FIG. 51 illustrates one embodiment of an architecture for asynchronous ray tracing;

FIG. 52 illustrates one embodiment of a ray traversal circuit;

FIG. 53 illustrates one embodiment of the present invention for performing non-local mean filtering; and

FIG. 54 illustrates a method according to one embodiment of the invention.

Detailed Description

In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the invention described below. It will be apparent, however, to one skilled in the art that embodiments of the invention may be practiced without some of these specific details. In other instances, well-known structures and devices are shown in block diagram form in order to avoid obscuring the underlying principles of embodiments of the present invention.

Exemplary graphics processor architecture and data types

Overview of the System

Fig. 1 is a block diagram of a processing system 100 according to an embodiment. The system 100 may be used in a single processor desktop system, a multi-processor workstation system, or a server system having a large number of processors 102 or processor cores 107. In one embodiment, system 100 is a processing platform incorporated within a system-on-a-chip (SoC) integrated circuit for use in a mobile, handheld, or embedded device, such as within an internet of things (IoT) device with wired or wireless connectivity to a local or wide area network.

In one embodiment, the system 100 can include, be coupled with, or be integrated within: a server-based game platform; game consoles, including games and media consoles; a mobile game console, a handheld game console, or an online game console. In some embodiments, system 100 is part of: a mobile phone, a smart phone, a tablet computing device, or a mobile internet connected device, such as a laptop computer with low internal storage capacity. The processing system 100 can also include, be coupled with, or be integrated within: wearable devices, such as smart watch wearable devices; smart glasses or apparel augmented with Augmented Reality (AR) or Virtual Reality (VR) features to provide visual, audio, or haptic output to supplement a real-world visual, audio, or haptic experience or to otherwise provide text, audio, graphics, video, holographic images or video, or haptic feedback; other Augmented Reality (AR) devices; or other Virtual Reality (VR) device. In some embodiments, the processing system 100 comprises or is part of a television or set-top box device. In one embodiment, the system 100 can include, be coupled with, or be integrated within: an autonomous vehicle such as a bus, tractor-trailer, automobile, motorcycle or electric bicycle, airplane or glider (or any combination thereof). An autonomous vehicle may use the system 100 to process the environment sensed around the vehicle.

In some embodiments, the one or more processors 102 each include one or more processor cores 107 to process instructions that, when executed, perform operations for system or user software. In some embodiments, at least one of the one or more processor cores 107 is configured to process a particular instruction set 109. In some embodiments, the instruction set 109 may facilitate Complex Instruction Set Computing (CISC), Reduced Instruction Set Computing (RISC), or computing via Very Long Instruction Words (VLIW). One or more processor cores 107 may process different instruction sets 109, which instruction sets 109 may include instructions to facilitate emulation of other instruction sets. Processor core 107 may also include other processing devices, such as a Digital Signal Processor (DSP).

In some embodiments, processor 102 includes cache memory 104. Depending on the architecture, the processor 102 can have a single internal cache or multiple levels of internal cache. In some embodiments, the cache memory is shared among various components of the processor 102. In some embodiments, the processor 102 also uses an external cache (e.g., a level 3 (L3) cache or Last Level Cache (LLC)) (not shown), which may be shared among the processor cores 107 using known cache coherency techniques. The register file 106 can additionally be included in the processor 102 and may include different types of registers (e.g., integer registers, floating point registers, status registers, and instruction pointer registers) 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 102.

In some embodiments, one or more processors 102 are coupled with one or more interface buses 110 to transmit communication signals, such as address, data, or control signals, between the processors 102 and other components in the system 100. Interface bus 110 can be a processor bus in one embodiment, 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 express), a memory bus, or other types of interface buses. In one embodiment, processor(s) 102 include an integrated memory controller 116 and a platform controller hub 130. The memory controller 116 facilitates communication between the memory devices and other components of the system 100, while the Platform Controller Hub (PCH) 130 provides a connection to I/O devices via a local I/O bus.

Memory device 120 can 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 some other memory device having suitable performance to act as a process memory. In one embodiment, memory device 120 is capable of operating as system memory for system 100 to store data 122 and instructions 121 for use when one or more processors 102 execute an application or process. The memory controller 116 is also coupled with an optional external graphics processor 118, which external graphics processor 118 may communicate with one or more graphics processors 108 in the processor 102 to perform graphics and media operations. In some embodiments, graphics, media, and/or computing operations may be assisted by an accelerator 112, which accelerator 112 is a co-processor that can be configured to perform a specialized set of graphics, media, or computing operations. For example, in one embodiment, the accelerator 112 is a matrix multiplication accelerator used to optimize machine learning or computational operations. In one embodiment, the accelerator 112 is a ray tracing accelerator that can be used to perform ray tracing operations in conjunction with the graphics processor 108. In one embodiment, an external accelerator 119 may be used in place of accelerator 112 or in conjunction with accelerator 112.

In some embodiments, a display device 111 can be connected to the processor(s) 102. The display device 111 can be one or more of an internal display device as in a mobile electronic device or laptop device or an external display device attached via a display interface (e.g., DisplayPort, etc.). In one embodiment, display device 111 can be a Head Mounted Display (HMD), such as a stereoscopic display device for use in Virtual Reality (VR) applications or Augmented Reality (AR) applications.

In some embodiments, platform controller hub 130 enables peripherals to connect to memory device 120 and processor 102 via a high speed I/O bus. The I/O peripherals include, but are not limited to, an audio controller 146, a network controller 134, a firmware interface 128, a wireless transceiver 126, a touch sensor 125, a data storage 124 (e.g., non-volatile memory, hard drive, flash memory, NAND, 3D XPoint, etc.). The data storage device 124 can be connected via a storage interface (e.g., SATA) or via a peripheral bus such as a peripheral component interconnect bus (e.g., PCI express). The touch sensor 125 can include a touch screen sensor, a pressure sensor, or a fingerprint sensor. The wireless transceiver 126 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 128 enables communication with system firmware and can be, for example, a Unified Extensible Firmware Interface (UEFI). Network controller 134 may implement a network connection to a wired network. In some embodiments, a high performance network controller (not shown) is coupled to interface bus 110. The audio controller 146 in one embodiment is a multi-channel high definition audio controller. In one embodiment, the system 100 includes an optional legacy I/O controller 140 for coupling legacy (e.g., personal System 2 (PS/2)) devices to the system. The platform controller hub 130 can also be connected to one or more Universal Serial Bus (USB) controllers 142 to connect input devices, such as a keyboard and mouse 143 combination, a camera 144, or other USB input devices.

It will be appreciated that the illustrated system 100 is exemplary and not limiting, as other types of data processing systems configured in different ways may also be used. For example, the instances of memory controller 116 and platform controller hub 130 may be integrated into a separate external graphics processor, such as external graphics processor 118. In one embodiment, platform controller hub 130 and/or memory controller 116 may be external to one or more processors 102. For example, system 100 can include an external memory controller 116 and a platform controller hub 130, which can be configured as a memory controller hub and a peripheral controller hub within a system chipset in communication with processor(s) 102.

For example, a circuit board ("sled") can be used on which components such as a CPU, memory, and other components are placed, which is designed for increased thermal performance. In some embodiments, a processing component, such as a processor, is located on the top side of the sled, while a near memory, such as a DIMM, is located on the bottom side of the sled. As a result of the enhanced air flow provided by this design, the assembly can operate at higher frequencies and power levels than in typical systems, thereby increasing performance. Further, the skillets are configured to blindly mate with the power and data communication cables in the racks, thereby enhancing their ability to be quickly removed, upgraded, reinstalled, and/or replaced. Similarly, 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 one another. In an illustrative embodiment, a component additionally includes a hardware attestation feature to verify its authenticity.

A data center can utilize a single network architecture ("fabric") that supports multiple other network architectures including ethernet and full Path (Omni-Path). The sled can be coupled to the switch via optical fibers that provide higher bandwidth and lower time delay than typical twisted pair cables (e.g., category 5e, category 6, etc.). Due to the high bandwidth, low latency interconnect and network architecture, the data center may use physically disaggregated pool resources, such as memory, accelerators (e.g., GPUs, graphics accelerators, FPGAs, ASICs, neural networks, and/or artificial intelligence accelerators, etc.), and data storage drives, and provide them to computing resources (e.g., processors) on an as-needed basis, so that the computing resources can access the pooled resources as if the pooled resources were local.

The power supply or power source can provide voltage and/or current to the system 100 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 outlet. 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. In one example, a power source or power supply includes wireless charging hardware to charge via a proximity charging field. In one example, the power source can include an internal battery, an alternating current supply, a motion-based power supply, a solar power supply, or a fuel cell source.

2A-2D illustrate a computing system and graphics processor provided by embodiments described herein. The elements of fig. 2A-2D having the same reference numbers (or names) as the elements of any other figure herein can operate or function in any manner similar to that described elsewhere herein, but are not limited to such.

FIG. 2A is a block diagram of an embodiment of a processor 200 having one or more processor cores 202A-202N, an integrated memory controller 214, and an integrated graphics processor 208. Processor 200 can include additional cores up to and including additional core 202N, represented by a dashed box. Each of the processor cores 202A-202N includes one or more internal cache units 204A-204N. In some embodiments, each processor core may also access one or more shared cache units 206. Internal cache units 204A-204N and shared cache unit 206 represent cache levels within processor 200. The cache memory hierarchy may include at least one level of instruction and data cache within each processor core, as well as one or more levels of shared mid-level cache, such as level 2 (L2), level 3 (L3), level 4 (L4), or other levels of cache, with the highest level of cache preceding external memory being classified as an LLC. In some embodiments, cache coherency logic maintains coherency between the various cache units 206 and 204A-204N.

In some embodiments, processor 200 may also include a set of one or more bus controller units 216 and a system agent core 210. One or more bus controller units 216 manage a set of peripheral buses, such as one or more PCI or PCI express buses. The system agent core 210 provides management functionality for various processor components. In some embodiments, the system proxy core 210 includes one or more integrated memory controllers 214 to manage access to various external memory devices (not shown).

In some embodiments, one or more of the processor cores 202A-202N include support for simultaneous multithreading. In such embodiments, the system proxy core 210 includes components for coordinating and operating the cores 202A-202N during processing of multiple threads. The system proxy core 210 may additionally include a Power Control Unit (PCU) that includes logic and components to regulate the power states of the processor cores 202A-202N and the graphics processor 208.

In some embodiments, the processor 200 additionally includes a graphics processor 208 to perform graphics processing operations. In some embodiments, the graphics processor 208 is coupled to a set of shared cache units 206 and a system proxy core 210 that includes one or more integrated memory controllers 214. In some embodiments, the system proxy core 210 also includes a display controller 211 to drive graphics processor output to one or more coupled displays. In some embodiments, the display controller 211 may also be a separate module coupled with the graphics processor via at least one interconnect, or may be integrated within the graphics processor 208.

In some embodiments, ring-based interconnect unit 212 is used to couple internal components of processor 200. However, alternative interconnect elements may be used, such as point-to-point interconnects, switched interconnects, or other techniques, including techniques known in the art. In some embodiments, the graphics processor 208 is coupled with the ring interconnect 212 via an I/O link 213.

Exemplary I/O link 213 represents at least one of a plurality of kinds of I/O interconnects, including on-package I/O interconnects that facilitate communication between various processor components and a high performance embedded memory module 218, such as an eDRAM module. In some embodiments, each of the processor cores 202A-202N and the graphics processor 208 can use the embedded memory module 218 as a shared last level cache.

In some embodiments, processor cores 202A-202N are homogeneous cores that execute the same instruction set architecture. In another embodiment, the processor cores 202A-202N are heterogeneous in Instruction Set Architecture (ISA), wherein one or more of the processor cores 202A-202N 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. In one embodiment, the processor cores 202A-202N are heterogeneous in micro-architecture, with one or more cores having relatively higher power consumption coupled with one or more power cores having lower power consumption. In one embodiment, the processor cores 202A-202N are heterogeneous in computing power. Additionally, processor 200 can be implemented on one or more chips or as an SoC integrated circuit having the illustrated components, among other components.

Fig. 2B is a block diagram of hardware logic of graphics processor core 219 according to some embodiments described herein. Elements of fig. 2B having the same reference numbers (or names) as elements of any other figure herein can operate or function in any manner similar to that described elsewhere herein, but are not limited to such. The graphics processor cores 219, sometimes referred to as core slices, can be one or more graphics cores within a modular graphics processor. Graphics processor core 219 is an example of one graphics core slice, and a graphics processor as described herein may include multiple graphics core slices based on a target power and performance envelope. Each graphics processor core 219 can include a fixed function block 230 coupled to a plurality of sub-cores 221A-221F, also referred to as sub-slices, the plurality of sub-cores 221A-221F comprising modular blocks of general purpose and fixed function logic.

In some embodiments, the fixed function block 230 includes a geometry/fixed function pipeline 231, the geometry/fixed function pipeline 231 capable of being shared by all of the sub-cores in the graphics processor core 219, for example, in a lower performance/or lower power graphics processor implementation. In various embodiments, geometry/fixed function pipeline 231 includes a 3D fixed function pipeline (e.g., 3D pipeline 312 in fig. 3 and 4, described below), a video front end unit, a thread spawner (thread spawner), and a thread dispatcher (thread dispatcher), and a unified return buffer manager that manages a unified return buffer (e.g., unified return buffer 418 in fig. 4, described below).

In one embodiment, fixed function block 230 also includes a graphics SoC interface 232, a graphics microcontroller 233, and a media pipeline 234. Graphics SoC interface 232 provides an interface between graphics processor core 219 and other processor cores within the system-on-a-chip integrated circuit. The graphics microcontroller 233 is a programmable sub-processor that may be configured to manage various functions of the graphics processor core 219, including thread dispatch, scheduling, and preemption (pre-preemption). Media pipeline 234 (e.g., media pipeline 316 of fig. 3 and 4) includes logic to facilitate decoding, encoding, pre-processing, and/or post-processing of multimedia data, including image and video data. Media pipeline 234 implements media operations via requests to compute or sample logic within sub-cores 221A-221F.

In one embodiment, SoC interface 232 enables graphics processor core 219 to communicate with a general-purpose application processor core (e.g., CPU) and/or other components within the SoC, including memory hierarchy elements such as a shared last level cache, system RAM, and/or embedded on-chip or on-package DRAM. SoC interface 232 may also enable communication with fixed-function devices within the SoC (such as a camera imaging pipeline) and enable the use of and/or implement global memory atoms that may be shared between graphics processor core 219 and CPUs within the SoC. SoC interface 232 may also implement power management control for graphics processor core 219 and interface between the clock domain of graphics core 219 and other clock domains within the SoC. In one embodiment, SoC interface 232 enables receipt of command buffers from a command stream dispatcher (command stream) and a global thread dispatcher configured to provide commands and instructions to each of one or more graphics cores within a graphics processor. Commands and instructions can be dispatched to the media pipeline 234 when media operations are to be performed or to geometry and fixed function pipelines (e.g., geometry and fixed function pipeline 231, geometry and fixed function pipeline 237) when graphics processing operations are to be performed.

The graphics microcontroller 233 can be configured to perform various scheduling and management tasks for the graphics processor core 219. In one embodiment, the graphics microcontroller 233 is capable of performing graphics and/or computational workload scheduling on various graphics parallel engines within the Execution Unit (EU) arrays 222A-222F, 224A-224F within the sub-cores 221A-221F. In this scheduling model, host software executing on a CPU core of an SoC that includes graphics processor core 219 is able to submit a workload to one of a plurality of graphics processor doorbells (graphics processor doorbells), which invokes a scheduling operation on the appropriate graphics engine. Scheduling operations include determining which workload to run next, submitting the workload to a command streamer, preempting an existing workload running on an engine, monitoring the progress of the workload, and notifying the host software when the workload is complete. In one embodiment, graphics microcontroller 233 can also facilitate a low power or idle state of graphics processor core 219, providing graphics processor core 219 with the ability to save and restore registers within graphics processor core 219 across low power state transitions independent of the operating system and/or graphics driver software on the system.

Graphics processor core 219 may have more or fewer sub-cores 221A-221F than illustrated, up to N modular sub-cores. Graphics processor core 219 can also include, for each set of N sub-cores, shared function logic 235, shared and/or cache memory 236, geometry/fixed function pipelines 237, and additional fixed function logic 238 to accelerate various graphics and computing processing operations. Shared function logic 235 can include logic units (e.g., samplers, math and/or inter-thread communication logic) associated with shared function logic 420 of fig. 4 that can be shared by every N subcores within graphics processor core 219. The shared and/or cache memory 236 can be a last level cache for a set of N sub-cores 221A-221F within the graphics processor core 219, and can also act as a shared memory accessible by multiple sub-cores. A geometry/fixed function pipeline 237 can be included in place of the geometry/fixed function pipeline 231 within the fixed function block 230 and can include the same or similar logic units.

In one embodiment, graphics processor core 219 includes additional fixed function logic 238, which can include various fixed function acceleration logic for use by graphics processor core 219. In one embodiment, the additional fixed function logic 238 includes an additional geometry pipeline for use in location-only shading. In position-only shading, there are two geometric pipelines: a full geometry pipeline within the geometry/fixed function pipelines 238, 231; and a culling pipeline (cu pipe), which is another geometric pipeline that may be included within the other fixed function logic 238. In one embodiment, the culling pipeline is a pruned version of the full geometry pipeline. The full pipeline and the culling pipeline are capable of executing different instances of the same application, each instance having a separate context. Location-only shading can hide long culling runs of discarded triangles so that shading can be done earlier in some instances. For example, and in one embodiment, the culling pipeline logic within the additional fixed function logic 238 is able to execute position shaders in parallel with the host application and generally generates critical results faster than a full pipeline, because the culling pipeline only acquires and colors position attributes of vertices without performing rasterization (rasterization) and rendering of pixels to a frame buffer. The culling pipeline can use the generated key results to calculate visibility information for all triangles regardless of whether those triangles were culled. The full pipeline (which may be referred to as a replay pipeline in this example) can consume visibility information to skip culled triangles to color only the visible triangles that are eventually passed to the rasterization stage.

In one embodiment, the additional fixed function logic 238 can also include machine learning acceleration logic, such as fixed function matrix multiplication logic, for implementation that includes optimization for machine learning training or reasoning.

Within each graphics sub-core 221A-221F is included a set of execution resources that may be used to perform graphics, media, and computational operations in response to requests by a graphics pipeline, media pipeline, or shader program. Graphics sub-cores 221A-221F include a plurality of EU arrays 222A-222F, 224A-224F, thread dispatch and inter-thread communication (TD/IC) logic 223A-223F, 3D (e.g., texture) samplers 225A-225F, media samplers 206A-206F, shader processors 227A-227F, and Shared Local Memories (SLMs) 228A-228F. The EU arrays 222A-222F, 224A-224F each include a plurality of execution units, which are general purpose graphics processing units capable of performing floating point and integer/fixed point logical operations for servicing graphics, media, or computational operations, including graphics, media, or compute shader programs. The TD/IC logic 223A-223F performs local thread dispatch and thread control operations for execution units within the sub-cores and facilitates communication between threads executing on the execution units of the sub-cores. 3-D samplers 225A-225F can read texture or other 3-D graphics related data into memory. The 3D sampler can read texture data in different ways based on the configured sample states and the texture format associated with a given texture. Media samplers 206A-206F can perform similar read operations based on the type and format associated with the media data. In one embodiment, each graphics sub-core 221A-221F can alternately include a unified 3D and media sampler. Threads executing on execution units within each of the sub-cores 221A-221F can utilize the shared local memory 228A-228F within each sub-core to enable threads executing within a thread group to execute using a common pool of on-chip memory.

FIG. 2C illustrates a Graphics Processing Unit (GPU) 239 that includes a dedicated set of graphics processing resources arranged into multi-core groups 240A-240N. While details are provided for only a single multi-core group 240A, it will be appreciated that the other multi-core groups 240B-240N may be equipped with the same or similar set of graphics processing resources.

As illustrated, the multi-core group 240A may include a set of graphics cores 243, a set of tensor cores 244, and a set of ray tracing cores 245. Scheduler/dispatcher 241 schedules and dispatches graphics threads for execution on the various cores 243, 244, 245. The set of register files 242 stores operand values used by the cores 243, 244, 245 in executing the graphics threads. These registers may include, for example, integer registers for storing integer values, floating point registers for storing floating point values, vector registers for storing packed data elements (integer and/or floating point data elements), and tile registers (tile registers) for storing tensor/matrix values. In one embodiment, the tile registers are implemented as a combined set of vector registers.

One or more combined level 1 (L1) cache and shared memory units 247 locally store graphics data, such as texture data, vertex data, pixel data, ray data, bounding volume data, and the like, within each multi-core group 240A. One or more texture units 247 can also be used to perform texture operations, such as texture mapping and sampling. A level 2 (L2) cache 253 shared by all or a subset of the multi-core groups 240A-240N stores graphics data and/or instructions for multiple concurrent graphics threads. As illustrated, the L2 cache 253 may be shared across multiple multi-core groups 240A-240N. The one or more memory controllers 248 couple the GPU 239 to memory 249, which memory 249 may be system memory (e.g., DRAM) and/or dedicated graphics memory (e.g., GDDR6 memory).

Input/output (I/O) circuitry 250 couples the GPU 239 to one or more I/O devices 252, such as Digital Signal Processors (DSPs), network controllers, or user input devices. On-chip interconnects may be used to couple the I/O devices 252 to the GPU 239 and memory 249. One or more I/O memory management units (IOMMU) 251 of the I/O circuitry 250 directly couple the I/O devices 252 to the system memory 249. In one embodiment, IOMMU 251 manages multiple sets of page tables to map virtual addresses to physical addresses in system memory 249. In this embodiment, the I/O device 252, CPU(s) 246, and GPU(s) 239 may share the same virtual address space.

In one implementation, IOMMU 251 supports virtualization. In this case, it may manage a first set of page tables to map guest/graphics virtual addresses to guest/graphics physical addresses and a second set of page tables to map guest/graphics physical addresses to system/host physical addresses (e.g., within system memory 249). The base of each of the first and second sets of page tables may be stored in a control register and swapped out on a context switch (e.g., so that access to the relevant set of page tables is provided for the new context). Although not illustrated in FIG. 2C, each of the cores 243, 244, 245 and/or multi-core groups 240A-240N may include a Translation Lookaside Buffer (TLB) to cache guest virtual-to-guest physical translations, guest physical-to-host physical translations, and guest virtual-to-host physical translations.

In one embodiment, the CPU 246, GPU 239, and I/O devices 252 are integrated on a single semiconductor chip and/or chip package. The illustrated memory 249 may be integrated on the same chip or may be coupled to the memory controller 248 via an off-chip interface. In one implementation, memory 249 comprises GDDR6 memory, which GDDR6 memory shares the same virtual address space as other physical system-level memory, although the underlying principles of the invention are not limited to this particular implementation.

In one embodiment, the tensor core 244 includes a plurality of execution units specifically designed to perform matrix operations, which are the basic computation operations used to perform deep learning operations. For example, simultaneous matrix multiplication operations may be used for neural network training and reasoning. The tensor core 244 may perform matrix processing using various operand precisions, including single precision floating point (e.g., 32 bits), half precision floating point (e.g., 16 bits), integer word (16 bits), byte (8 bits), and nibble (4 bits). In one embodiment, a neural network implementation extracts features of each rendered scene, potentially combining details from multiple frames to construct a high quality final image.

In a deep learning implementation, parallel matrix multiplication work may be scheduled for execution on the tensor core 244. Training of neural networks requires, in particular, a large number of matrix dot-product operations. To handle the inner product formulation of the nx N matrix multiplication, the tensor core 244 may include at least N dot product processing elements. Before the start of matrix multiplication, a complete matrix is loaded into the tile register, and at least one column of the second matrix is loaded in each of the N cycles. At each cycle, there are N dot products processed.

Depending on the particular implementation, the matrix elements may be stored with different precisions, including 16-bit words, 8-bit bytes (e.g., INT 8), and 4-bit nibbles (e.g., INT 4). Different precision modes can be specified for the tensor core 244 to ensure that the most efficient precision is used for different workloads (e.g., such as an inference workload that can tolerate quantization to bytes and nibbles).

In one embodiment, the ray tracing core 245 speeds up ray tracing operations for both real-time ray tracing and non-real-time ray tracing implementations. In particular, ray tracing core 245 includes ray traversal/intersection circuitry to perform ray traversal using a bounding volume level (BVH) and identify intersections between rays and primitives enclosed within the BVH volume. Ray tracing core 245 may also include circuitry for performing depth testing and culling (e.g., using a Z-buffer or similar arrangement). In one implementation, the ray tracing kernel 245 performs traversal and intersection operations in conjunction with the image denoising techniques described herein, at least a portion of which may be performed on the tensor kernel 244. For example, in one embodiment, the tensor kernel 244 implements a deep learning neural network to perform denoising of frames generated by the ray tracing kernel 245. However, CPU(s) 246, graphics kernel 243 and/or ray tracing kernel 245 may also implement all or a portion of a denoising and/or deep learning algorithm.

Additionally, as described above, a distributed approach to denoising may be employed, where the GPU 239 is in a computing device coupled to other computing devices through a network or high speed interconnect. In this embodiment, interconnected computing devices share neural network learning/training data to improve the speed with which the overall system learns to perform denoising for different types of image frames and/or different graphics applications.

In one embodiment, ray tracing core 245 handles all BVH traversals and ray-primitive intersections, protecting the graphics core 243 from being overloaded with thousands of instructions per ray. In one embodiment, each ray tracing core 245 includes a first set of dedicated circuitry for performing bounding box tests (e.g., for traversal operations) and a second set of dedicated circuitry for performing ray-triangle intersection tests (e.g., intersecting traversed rays). Thus, in one embodiment, the multi-core group 240A can simply launch the ray probe, and the ray tracing core 245 independently performs ray traversals and intersections and returns hit data (e.g., hit, no hit, multiple hits, etc.) to the thread context. While the ray tracing core 245 performs traversal and intersection operations, the other cores 243, 244 are freed to perform other graphics or computational work.

In one embodiment, each ray tracing core 245 includes a traversal unit to perform BVH test operations and an intersection unit to perform ray-primitive intersection tests. The crossbar unit generates a "hit", "no-hit", or "multiple-hit" response, and the crossbar unit provides the response to the appropriate thread. During traversal and intersection operations, execution resources of other cores (e.g., graphics core 243 and tensor core 244) are freed to perform other forms of graphics work.

In one particular embodiment described below, a hybrid rasterization/ray tracing approach is used in which work is distributed between graphics core 243 and ray tracing core 245.

In one embodiment, the ray tracing core 245 (and/or other cores 243, 244) includes hardware support for a ray tracing instruction set such as Microsoft's DirectX ray tracing (DXR), which includes the DispatchRays command along with ray generation, closest hits, arbitrary hits, and miss shaders, which enable each object to be assigned a unique set of textures and shaders. Another ray tracing platform that may be supported by ray tracing core 245, graph core 243, and tensor core 244 is Vulkan 1.1.85. Note, however, that the underlying principles of the invention are not limited to any particular ray tracing ISA.

In general, the various cores 245, 244, 243 may support a ray trace instruction set that includes instructions/functions for ray generation, closest hits, arbitrary hits, ray-primitive intersections, per-primitive and hierarchical bounding box constructions, misses, accesses, and exceptions (exceptions). More specifically, one embodiment includes ray tracing instructions to perform the following functions:

ray generation-ray generation instructions may be executed for each pixel, sample, or other user-defined job assignment.

Closest hit-the closest hit instruction may be executed to locate the closest intersection of rays with primitives within the scene.

Any hit-any hit instruction identifies multiple intersections between primitives and rays within the scene, potentially identifying a new closest intersection.

The cross-cross instruction performs a ray-primitive cross test and outputs a result.

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

Miss-indicating a ray misses all geometric shapes within a scene or a specified region of a scene.

Visit-child (child volume) that indicates the ray will traverse.

Exceptions-include various types of exception handlers (e.g., invoked for various error conditions).

Fig. 2D is a block diagram of a General Purpose Graphics Processing Unit (GPGPU) 270 that can be configured as a graphics processor and/or compute accelerator according to embodiments described herein. The GPGPU 270 can be interconnected with a host processor (e.g., one or more CPUs 246) and memories 271, 272 via one or more system and/or memory buses. In one embodiment, memory 271 is a system memory that may be shared with one or more CPUs 246, while memory 272 is a device memory dedicated to GPGPU 270. In one embodiment, components within device memory 272 and GPGPU 270 may be mapped into memory addresses accessible to one or more CPUs 246. Access to the memories 271 and 272 may be facilitated via the memory controller 268. In one embodiment, memory controller 268 includes an internal Direct Memory Access (DMA) controller 269 or can include logic to perform operations that would otherwise be performed by a DMA controller.

The GPGPU 270 includes a plurality of caches including an L2 cache 253, an L1 cache 254, an instruction cache 255, and a shared memory 256, at least a portion of which shared memory 256 may also be partitioned into caches. GPGPU 270 also includes a plurality of compute units 260A-260N. Each compute unit 260A-260N includes a set of vector registers 261, scalar registers 262, vector logic 263, and scalar logic 264. The compute units 260A-260N can also include a local shared memory 265 and a program counter 266. The compute units 260A-260N can be coupled with a constant cache 267, which constant cache 267 can be used to store constant data, which is data that will not change during the execution of kernel or shader programs executing on the GPGPU 270. In one embodiment, the constant cache 267 is a scalar data cache, and cached (cached) data can be fetched directly into the scalar registers 262.

During operation, one or more CPUs 246 can write commands to registers or memory in GPGPU 270 that have been mapped into an accessible address space. The command processor 257 is able to read commands from registers or memory and determine how those commands will be processed within the GPGPU 270. Thread dispatcher 258 can then be used to dispatch threads to compute units 260A-260N to execute those commands. Each compute unit 260A-260N is capable of executing threads independently of the other compute units. In addition, each of the computation units 260A-260N can be independently configured for conditional computations and can conditionally output the results of the computations to memory. The command processor 257 is able to interrupt one or more CPUs 246 when the submitted command is complete.

3A-3C illustrate block diagrams of additional graphics processor and compute accelerator architectures provided by embodiments described herein. The elements of fig. 3A-3C having the same reference numbers (or names) as the elements of any other figure herein can operate or function in any manner similar to that described elsewhere herein, but are not limited to such.

Fig. 3A is a block diagram of a graphics processor 300, which graphics processor 300 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. In some embodiments, the graphics processor communicates via a memory mapped I/O interface to registers on the graphics processor and with commands placed in processor memory. In some embodiments, graphics processor 300 includes a memory interface 314 to access memory. Memory interface 314 can be an interface to local memory, one or more internal caches, one or more shared external caches, and/or to system memory.

In some embodiments, graphics processor 300 also includes display controller 302 to drive display output data to display device 318. The display controller 302 includes hardware for one or more overlay planes for displaying and combining multiple layers of video or user interface elements. The display device 318 can be an internal or external display device. In one embodiment, display device 318 is a head mounted display device, such as a Virtual Reality (VR) display device or an Augmented Reality (AR) display device. In some embodiments, graphics processor 300 includes a video codec engine 306 to encode, decode, or transcode media into, from, or between one or more media encoding formats, including, but not limited to, Motion Picture Experts Group (MPEG) formats such as MPEG-2, Advanced Video Coding (AVC) formats such as h.264/MPEG-4 AVC, h.265/HEVC, open media alliance (AOMedia) VP8, VP9, and Society of Motion Picture and Television Engineers (SMPTE) 421M/VC-1 and Joint Photographic Experts Group (JPEG) formats such as JPEG, and motion JPEG (mjpeg) formats.

In some embodiments, graphics processor 300 includes a block image transfer (BLIT) engine 304 to perform two-dimensional (2D) rasterizer operations, including, for example, bit boundary block transfers. However, in one embodiment, 2D graphics operations are performed using one or more components of a Graphics Processing Engine (GPE) 310. In some embodiments, GPE 310 is a compute engine for performing graphics operations including three-dimensional (3D) graphics operations and media operations.

In some embodiments, GPE 310 includes a 3D pipeline 312 for performing 3D operations, such as rendering three-dimensional images and scenes using processing functions that act on 3D primitive shapes (e.g., rectangles, triangles, etc.). The 3D pipeline 312 includes programmable and fixed function elements that perform various tasks within the elements and/or spawn (spawn) threads of execution to the 3D/media subsystem 315. While the 3D pipeline 312 can be used to perform media operations, embodiments of the GPE 310 also include a media pipeline 316 that is particularly used to perform media operations, such as video post-processing and image enhancement.

In some embodiments, media pipeline 316 includes fixed-function or programmable logic units to perform one or more dedicated media operations, such as video decoding acceleration, video de-interleaving, and video encoding acceleration, in place of or on behalf of video codec engine 306. In some embodiments, media pipeline 316 additionally includes a thread spawn unit to spawn threads for execution on 3D/media subsystem 315. The spawned threads perform computations for media operations on one or more graphics execution units included in 3D/media subsystem 315.

In some embodiments, 3D/media subsystem 315 includes logic for executing threads spawned by 3D pipeline 312 and media pipeline 316. In one embodiment, the pipeline sends thread execution requests to the 3D/media subsystem 315, the 3D/media subsystem 315 including thread dispatch logic for arbitrating and dispatching various requests to available thread execution resources. The execution resources include an array of graphics execution units to process 3D and media threads. In some embodiments, 3D/media subsystem 315 includes one or more internal caches for thread instructions and data. In some embodiments, the subsystem further includes a shared memory including registers and addressable memory to share data between the threads and to store output data.

FIG. 3B illustrates a graphics processor 320 having a tiled (tiled) architecture, according to embodiments described herein. In one embodiment, the graphics processor 320 includes a graphics processing engine cluster 322, the graphics processing engine cluster 322 having multiple instances of the graphics processing engine 310 of FIG. 3A within the graphics engine tiles 310A-310D. Each graphics engine tile 310A-310D can be interconnected 267-via a set of tile interconnects 323A-323F. Each graphics engine tile 310A-310D can also be connected to a memory module or memory device 326A-326D via a memory interconnect 325A-325D. Memory devices 326A-326D can use any graphics memory technology. For example, memory devices 326A-326D may be Graphics Double Data Rate (GDDR) memories. Memory devices 326A-326D are, in one embodiment, High Bandwidth Memory (HBM) modules that can be on-die with their respective graphics engine tiles 310A-310D. In one embodiment, memory devices 326A-326D are stacked memory devices that can be stacked on top of their respective graphics engine tiles 310A-310D. In one embodiment, as described in further detail in FIGS. 11B-11D, each graphics engine tile 310A-310D and associated memory 326A-326D reside on a separate chiplet (chiplet) that is bonded to a base die or base substrate.

Graphics processing engine cluster 322 can be connected with on-chip or on-package fabric interconnect 324. The fabric interconnect 324 may enable communication between the graphics engine tiles 310A-310D and components such as the video codec 306 and one or more replication engines 304. The replication engine 304 can be used to move data out of, in, and between: memory devices 326A-326D and memory external to graphics processor 320 (e.g., system memory). The fabric interconnect 324 can also be used to interconnect the graphics engine tiles 310A-310D. The graphics processor 320 may optionally include a display controller 302 to enable connection to an external display device 318. The graphics processor may also be configured as a graphics or compute accelerator. In an accelerator configuration, the display controller 302 and the display device 318 may be omitted.

The graphics processor 320 is capable of connecting to a host system via a host interface 328. The host interface 328 may enable communication between the graphics processor 320, system memory, and/or other system components. The host interface 328 can be, for example, a PCI express bus or another type of host system interface.

FIG. 3C illustrates a computation accelerator 330 according to embodiments described herein. The compute accelerator 330 can include architectural similarities to the graphics processor 320 of fig. 3B and is optimized for compute acceleration. Compute engine cluster 332 can include a set of compute engine tiles 340A-340D that include execution logic optimized for parallel or vector-based general purpose compute operations. In some embodiments, the compute engine tiles 340A-340D do not include fixed function graphics processing logic, although in one embodiment, one or more of the compute engine tiles 340A-340D can include logic to perform media acceleration. The compute engine tiles 340A-340D can be connected to the memories 326A-326D via the memory interconnects 325A-325D. The memories 326A-326D and the memory interconnects 325A-325D may be similar technologies as in the graphics processor 320, or can be different. The graphics compute engine tiles 340A-340D may also be interconnected via a set of tile interconnects 323A-323F, and may be connected with the fabric interconnect 324 and/or interconnected by the fabric interconnect 324. In one embodiment, the compute accelerator 330 includes a large L3 cache 336 that can be configured as a device-wide cache. The compute accelerator 330 can also be connected to a host processor and memory via a host interface 328 in a similar manner as the graphics processor 320 of FIG. 3B.

Graphics processing engine

FIG. 4 is a block diagram of a graphics processing engine 410 of a graphics processor, according to some embodiments. In one embodiment, the Graphics Processing Engine (GPE) 410 is a version of the GPE 310 shown in FIG. 3A, and may also represent the graphics engine tiles 310A-310D of FIG. 3B. The elements of fig. 4 having the same reference numbers (or names) as the elements of any other figure herein can operate or function in any manner similar to that described elsewhere herein, but are not limited to such. For example, the 3D pipeline 312 and the media pipeline 316 of fig. 3A are illustrated. The media pipeline 316 is optional in some embodiments of the GPE 410 and may not be explicitly included within the GPE 410. For example, and in at least one embodiment, a separate media and/or image processor is coupled to GPE 410.

In some embodiments, GPE 410 is coupled with or includes a command streamer 403 that provides a command stream to 3D pipeline 312 and/or media pipeline 316. In some embodiments, command streamer 403 is coupled with a memory, which can be a system memory, or one or more of an internal cache and a shared cache. In some embodiments, command streamer 403 receives commands from memory and sends the commands to 3D pipeline 312 and/or media pipeline 316. The command is an indication retrieved from a ring buffer that stores commands for the 3D pipeline 312 and the media pipeline 316. In one embodiment, the ring buffer can additionally include a batch command buffer that stores a batch of the plurality of commands. The commands for 3D pipeline 312 can also include references to data stored in memory, such as, but not limited to, vertex and geometry data for 3D pipeline 312 and/or image data and memory objects for media pipeline 316. The 3D pipeline 312 and the media pipeline 316 process commands and data by performing operations via logic within the respective pipelines or by dispatching one or more execution threads to the graphics core array 414. In one embodiment, graphics core array 414 includes one or more blocks of graphics cores (e.g., graphics core(s) 415A, graphics core(s) 415B), each block including one or more graphics cores. Each graphics core includes: a set of graphics execution resources comprising general purpose and graphics specific execution logic to perform graphics and computational operations; and fixed function texture processing and/or machine learning and artificial intelligence acceleration logic.

In various embodiments, 3D pipeline 312 can include fixed functionality and programmable logic to process one or more shader programs (such as vertex shaders, geometry shaders, pixel shaders, fragment shaders, compute shaders, or other shader programs) by processing instructions and dispatching execution threads to graphics core array 414. Graphics core array 414 provides a uniform block of execution resources for use in processing these shader programs. Multipurpose execution logic (e.g., execution units) within graphics core(s) 415A-415B of graphics core array 414 includes support for various 3D API shader languages and is capable of executing multiple simultaneous execution threads associated with multiple shaders.

In some embodiments, graphics core array 414 includes execution logic to perform media functions, such as video and/or image processing. In one embodiment, the execution unit includes general purpose logic that is programmable to perform parallel general purpose computing operations in addition to graphics processing operations. The general purpose logic is capable of performing processing operations in parallel or in conjunction with the general purpose logic within the processor core(s) 107 of fig. 1 or cores 202A-202N as in fig. 2A.

Output data generated by threads executing on graphics core array 414 can output the data to memory in Unified Return Buffer (URB) 418. The URB 418 is capable of storing data for multiple threads. In some embodiments, the URB 418 may be used to send data between different threads executing on the graphics core array 414. In some embodiments, the URB 418 may additionally be used for synchronization between threads on the graphics core array and fixed function logic within the shared function logic 420.

In some embodiments, the graphics core array 414 is scalable such that the array includes a variable number of graphics cores each having a variable number of execution units based on the target power and performance level of the GPE 410. In one embodiment, the execution resources are dynamically scalable such that the execution resources may be enabled or disabled as needed.

Graphics core array 414 is coupled to shared function logic 420, which shared function logic 420 includes a plurality of resources shared between graphics cores in the graphics core array. The shared function within shared function logic 420 is a hardware logic unit that provides dedicated supplemental functionality to graphics core array 414. In various embodiments, shared function logic 420 includes, but is not limited to, sampler 421, math 422, and inter-thread communication (ITC) 423 logic. Additionally, some embodiments implement one or more caches 425 within shared function logic 420.

Shared functionality is implemented, at least in cases where the demand for a given dedicated function is insufficient to be contained within graphics core array 414. Rather, the individual instantiations of the specialized function are implemented as separate entities in the shared function logic 420 and are shared among the execution resources within the graphics core array 414. The exact set of functions shared within graphics core array 414 and included within graphics core array 414 varies across embodiments. In some embodiments, a particular shared function within shared function logic 420 that is widely used by graphics core array 414 may be included within shared function logic 416 within graphics core array 414. In various embodiments, shared function logic 416 within graphics core array 414 can include some or all of the logic within shared function logic 420. In one embodiment, all logic elements within shared function logic 420 may be duplicated within shared function logic 416 of graphics core array 414. In one embodiment, shared function logic 420 is eliminated in favor of shared function logic 416 within graphics core array 414.

Execution unit

5A-5B illustrate thread execution logic 500 including an array of processing elements employed in a graphics processor core according to embodiments described herein. The elements of fig. 5A-5B having the same reference numbers (or names) as the elements of any other figure herein can operate or function in any manner similar to that described elsewhere herein, but are not limited to such. 5A-5B illustrate an overview of thread execution logic 500, which thread execution logic 500 may represent hardware logic illustrated with each of the sub-cores 221A-221F of FIG. 2B. FIG. 5A represents an execution unit within a general purpose graphics processor, while FIG. 5B represents an execution unit that may be used within a compute accelerator.

As illustrated in fig. 5A, in some embodiments, thread execution logic 500 includes shader processor 502, thread dispatcher 504, instruction cache 506, scalable execution unit array including multiple execution units 508A-508N, sampler 510, shared local memory 511, data cache 512, and data port 514. In one embodiment, the scalable array of execution units is capable of being dynamically scaled by enabling or disabling one or more execution units (e.g., any of execution units 508A, 508B, 508C, 508D through 508N-1 and 508N) based on the computational requirements of the workload. In one embodiment, the included components are interconnected via an interconnection fabric linked to each of the components. In some embodiments, the thread execution logic 500 includes one or more connections to memory (such as system memory or cache memory) through one or more of the instruction cache 506, data port 514, sampler 510, and execution units 508A-508N. In some embodiments, each execution unit (e.g., 508A) is an independently programmable general purpose computing unit capable of executing multiple simultaneous hardware threads while processing multiple data elements in parallel for each thread. In various embodiments, the array of execution units 508A-508N is scalable to include any number of individual execution units.

In some embodiments, the EUs 508A-508N are primarily used to execute shader programs. Shader processor 502 is capable of processing various shader programs and dispatching threads of execution associated with the shader programs via thread dispatcher 504. In one embodiment, the thread dispatcher includes logic to arbitrate thread initiation requests from the graphics and media pipelines and to instantiate the requested thread on one or more of the execution units 508A-508N. For example, the geometry pipeline can dispatch vertices, tessellations (tessellation), or geometry shaders to thread execution logic for processing. In some embodiments, the thread dispatcher 504 is also capable of processing runtime thread yield requests from executing shader programs.

In some embodiments, execution units 508A-508N support an instruction set that includes native support for many standard 3D graphics shader instructions, such that shader programs from graphics libraries (e.g., Direct 3D and OpenGL) are executed with minimal translation. Execution units support vertex and geometry processing (e.g., vertex programs, geometry programs, vertex shaders), pixel processing (e.g., pixel shaders, fragment shaders), and general purpose processing (e.g., compute and media shaders). Each of the execution units 508A-508N is capable of multi-issue (multi-issue) Single Instruction Multiple Data (SIMD) execution, and multi-threading enables an efficient execution environment in the face of higher latency memory accesses. Each hardware thread within each execution unit has a dedicated high bandwidth register file and associated independent thread state. Execution is a multiple issue per clock to a pipeline capable of integer, single and double precision floating point operations, SIMD branch capability, logical operations, override operations, and other miscellaneous operations. While waiting for data from one of the memory or shared functions, dependency logic within the execution units 508A-508N sleeps the waiting thread until the requested data has been returned. While the waiting thread is sleeping, the hardware resources may be dedicated to processing other threads. For example, during a delay associated with vertex shader operations, the execution unit can execute operations for a pixel shader, a fragment shader, or another type of shader program (including a different vertex shader). Various embodiments can be adapted to use execution by using Single Instruction Multiple Threads (SIMT) as an alternative to using SIMD or in addition to using SIMD. References to SIMD cores or operations can also be applicable to SIMT or to SIMD combined with SIMT.

Each of the execution units 508A-508N operates on an array of data elements. The number of data elements is the "execution size" or number of lanes for the instruction. An execution channel is a logical unit for the execution of data element access, masking, and flow control within an instruction. The number of lanes may be independent of the number of physical Arithmetic Logic Units (ALUs) or Floating Point Units (FPUs) for a particular graphics processor. In some embodiments, execution units 508A-508N support both integer and floating point data types.

The execution unit instruction set includes SIMD instructions. Various data elements can be stored as packed data types in registers, and execution units will process the various elements based on their data sizes. For example, in operating on a 256-bit wide vector, 256 bits of the vector are stored in a register, and the execution unit operates on the vector as four separate 54-bit packed data elements (four word (QW) size data elements), eight separate 32-bit packed data elements (double word (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, one or more execution units can be combined into a fused execution unit 509A-509N having thread control logic (507A-507N) that is common to the fused EU. Multiple EUs can be fused into an EU group. Each EU in the fused EU group can be configured to execute a separate SIMD hardware thread. The number of EUs in the fused EU group can vary according to the embodiment. In addition, various SIMD widths can be performed per EU, including but not limited to SIMD8, SIMD16, and SIMD 32. Each fused graphics execution unit 509A-509N includes at least two execution units. For example, the fused execution unit 509A includes a first EU 508A, a second EU 508B and thread control logic 507A, the thread control logic 507A being common to the first EU 508A and the second EU 508B. The thread control logic 507A controls the threads executing on the fused graphics execution unit 509A, allowing each EU within the fused execution units 509A-509N to execute using a common instruction pointer register.

One or more internal instruction caches (e.g., 506) are included in the thread execution logic 500 to cache thread instructions for the execution units. In some embodiments, one or more data caches (e.g., 512) are included to cache thread data during thread execution. Threads executing on the execution logic 500 are also capable of storing explicitly managed data in the shared local memory 511. In some embodiments, sampler 510 is included to provide texture samples for 3D operations and media samples for media operations. In some embodiments, sampler 510 includes dedicated texture or media sampling functionality to process texture or media data during a sampling process prior to providing the sampled data to an execution unit.

During execution, the graphics and media pipeline sends thread initiation requests to the thread execution logic 500 via thread spawn and dispatch logic. Once a group of geometric objects has been processed and rasterized into pixel data, pixel processor logic (e.g., pixel shader logic, fragment shader logic, etc.) within shader processor 502 is invoked to further compute output information and cause the results to be written to an output surface (e.g., a color buffer, a depth buffer, a stencil buffer, etc.). In some embodiments, a pixel shader or fragment shader computes values for various vertex attributes to be interpolated across rasterized objects. In some embodiments, pixel processor logic within shader processor 502 then executes an Application Programming Interface (API) supplied pixel or fragment shader program. To execute shader programs, shader processor 502 dispatches threads to execution units (e.g., 508A) via thread dispatcher 504. In some embodiments, shader processor 502 uses texture sampling logic in sampler 510 to access texture data in a texture map stored in memory. Arithmetic operations on the texture data and the input geometry data compute pixel color data for each geometric segment, or discard one or more pixels from further processing.

In some embodiments, the data port 514 provides a memory access mechanism for the thread execution logic 500 to output processed data to memory for further processing on a graphics processor output pipeline. In some embodiments, data port 514 includes or is coupled to one or more cache memories (e.g., data cache 512) to cache data for memory access via the data port.

In one embodiment, the execution logic 500 can further include a ray tracker 505, the ray tracker 505 can provide ray tracing acceleration functionality. The ray tracker 505 can support a ray tracing instruction set that includes instructions/functionality for ray generation. The ray trace instruction set can be similar to or different from the ray trace instruction set supported by the ray trace core 245 of FIG. 2C.

Fig. 5B illustrates exemplary internal details of the execution unit 508 according to an embodiment. The graphics execution unit 508 can include an instruction fetch unit 537, a general register file array (GRF) 524, an architectural register file Array (ARF) 526, a thread arbiter 522, a send unit 530, a branch unit 532, a set of SIMD Floating Point Units (FPUs) 534, and in one embodiment, a set of dedicated integer SIMD ALUs 535. The GRF 524 and ARF 526 comprise a set of general purpose register files and architectural register files associated with each simultaneous hardware thread that may be active in the graphics execution unit 508. In one embodiment, per-thread architectural state is maintained in the ARF 526, while data used during thread execution is stored in the GRF 524. The execution state of each thread (including the instruction pointer for each thread) can be maintained in thread specific registers in the ARF 526.

In one embodiment, the graphics execution unit 508 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 execution unit and the target number of simultaneous threads, where execution unit resources are partitioned across logic used to execute multiple simultaneous threads. The number of logical threads that can be executed by the graphics execution unit 508 is not limited to the number of hardware threads, and multiple logical threads can be assigned to each hardware thread.

In one embodiment, the graphics execution unit 508 is capable of collectively issuing multiple instructions, which may each be different instructions. The thread arbiter 522 of the graphics execution unit thread 508 is capable of dispatching instructions to one of the issue unit 530, branch unit 532, or SIMD FPU(s) 534 for execution. Each execution thread has access to 128 general purpose registers within the GRF 524, where each register is capable of storing 32 bytes, which are accessible as a SIMD 8 element vector of 32-bit data elements. In one embodiment, each execution unit thread may access 4 kilobytes within GRF 524, although embodiments are not so limited and in other embodiments more or less register resources may be provided. In one embodiment, the graphics execution unit 508 is partitioned into seven hardware threads capable of independently performing computational operations, although the number of threads per execution unit may vary depending on the embodiment. For example, up to 16 hardware threads are supported in one embodiment. In an embodiment where seven threads have access to 4 kilobytes, the GRF 524 is capable of storing a total of 28 kilobytes. In the case where 16 threads have access to 4 kilobytes, the GRF 524 is able to store a total of 64 kilobytes. The flexible addressing mode can allow registers to be addressed together to efficiently build wider registers or to represent a strided rectangular block data structure.

In one embodiment, memory operations, sampler operations, and other longer latency system communications are dispatched via a "send" instruction executed by the messaging transmit unit 530. In one embodiment, branch instructions are dispatched to a dedicated branch unit 532 to facilitate SIMD divergence and eventual convergence.

In one embodiment, graphics execution unit 508 includes one or more SIMD floating-point units ((one or more) FPUs) 534 to perform floating-point operations. In one embodiment, FPU(s) 534 also support integer computations. In one embodiment, FPU(s) 534 are capable of performing up to a number M of 32-bit floating point (or integer) operations on SIMD's, or up to 2M of 16-bit integer or 16-bit floating point operations on SIMD's. In one embodiment, at least one of the FPU(s) provides extended mathematical capabilities to support high throughput beyond mathematical functions and double precision 54-bit floating point. In some embodiments, there is also a set of 8-bit integer SIMD ALUs 535, and the set of 8-bit integer SIMD ALUs 535 may be specifically optimized to perform operations associated with machine learning computations.

In one embodiment, an array of multiple instances of graphics execution unit 508 can be instantiated in a graphics sub-core grouping (e.g., a subslice). For scalability, the product architect can select the exact number of execution units per sub-core grouping. In one embodiment, execution unit 508 is capable of executing instructions across multiple execution lanes. In further embodiments, each thread executing on the graphics execution unit 508 is executing on a different channel.

Fig. 6 illustrates a further execution unit 600 according to an embodiment. Execution unit 600 may be a compute optimized execution unit for use, for example, in compute engine tiles 340A-340D as in fig. 3C, but is not so limited. Variations of the execution unit 600 may also be used in the graphics engine tiles 310A-310D as in fig. 3B. In one embodiment, the execution unit 600 includes a thread control unit 601, a thread state unit 602, an instruction fetch/pre-fetch unit 603, and an instruction decode unit 604. Execution unit 600 additionally includes a register file 606, the register file 606 storing registers that can be assigned to hardware threads within the execution unit. The execution unit 600 additionally comprises a dispatch unit 607 and a branch unit 608. In one embodiment, the dispatch unit 607 and branch unit 608 can operate similarly to the dispatch unit 530 and branch unit 532 of the graphics execution unit 508 of FIG. 5B.

The execution unit 600 further comprises a calculation unit 610, said calculation unit 610 comprising a plurality of different types of functional units. In one embodiment, compute unit 610 includes an ALU unit 611, the ALU unit 611 including an array of arithmetic logic units. ALU unit 611 can be configured to perform 64-bit, 32-bit, and 16-bit integer and floating point operations. Integer and floating point operations may be performed simultaneously. The computing unit 610 can also include a systolic array 612 and a math unit 613. Systolic array 612 comprises a network of width W and depth D of data processing elements that can be used to perform vector or other data parallel operations in a systolic manner. In one embodiment, systolic array 612 can be configured to perform matrix operations, such as matrix dot product operations. In one embodiment, systolic array 612 supports 16-bit floating point operations and 8-bit and 4-bit integer operations. In one embodiment, the systolic array 612 can be configured to accelerate machine learning operations. In such embodiments, systolic array 612 can be configured with support for a bfloat 16-bit floating point format. In one embodiment, math unit 613 can be included to perform a particular subset of math operations in an efficient and less powerful manner than ALU unit 611. Math unit 613 can include variations of math logic that can be found in shared function logic of graphics processing engines provided by other embodiments (e.g., math logic 422 of shared function logic 420 of fig. 4). In one embodiment, the math unit 613 can be configured to perform 32-bit and 64-bit floating point operations.

Thread control unit 601 includes logic to control execution of threads within execution units. Thread control unit 601 can include thread arbitration logic to start, stop, and preempt execution of threads within execution unit 600. The thread state unit 602 can be used to store thread states for threads assigned to execute on the execution unit 600. Storing thread states within execution unit 600 enables fast preemption of threads when those threads become blocked or idle. The instruction fetch/pre-fetch unit 603 is capable of fetching instructions from an instruction cache of higher level execution logic (e.g., as instruction cache 506 in fig. 5A). The instruction fetch/prefetch unit 603 is also capable of issuing a prefetch request for instructions to be loaded into the instruction cache based on an analysis of the currently executing thread. Instruction decode unit 604 can be used to decode instructions to be executed by a compute unit. In one embodiment, the instruction decode unit 604 can be used as a secondary decoder to decode complex instructions into constituent micro-operations.

Execution unit 600 additionally includes a register file 606 that can be used by hardware threads executing on execution unit 600. The registers in register file 606 can be partitioned across logic used to execute multiple simultaneous threads within compute unit 610 of execution unit 600. The number of logical threads that can be executed by the graphics execution unit 600 is not limited to the number of hardware threads, and multiple logical threads can be assigned to each hardware thread. The size of the register file 606 can vary across embodiments based on the number of hardware threads supported. In one embodiment, register renaming may be used to dynamically allocate registers to hardware threads.

Fig. 7 is a block diagram illustrating a graphics processor instruction format 700, according to some embodiments. In one or more embodiments, a graphics processor execution unit supports an instruction set with instructions in multiple formats. The solid box illustrates components typically included in an execution unit instruction, while the dashed line includes components that are optional or included only in a subset of instructions. In some embodiments, the instruction format 700 described and illustrated is a macro-instruction because they are instructions supplied to the execution units, as opposed to micro-operations that result from instruction decoding once the instruction is processed.

In some embodiments, the graphics processor execution unit natively supports instructions in the 128-bit instruction format 710. The 64-bit packed instruction format 730 may be used for some instructions based on the selected instruction, instruction options, and number of operands. The native 128-bit instruction format 710 provides access to all instruction options, while in the 64-bit format 730 some options and operations are restricted. The available native instructions in the 64-bit format 730 vary from embodiment to embodiment. In some embodiments, instructions are partially compressed using a set of index values in the index field 713. The execution unit hardware references a set of compression tables based on the index values and uses the compression table outputs to reconstruct native instructions in the 128-bit instruction format 710. Other sizes and formats of instructions can be used.

For each format, instruction opcode 712 defines the operation to be performed by the execution unit. An execution unit executes each instruction in parallel across multiple data elements of each operand. For example, in response to an add instruction, the execution unit performs a simultaneous add operation across each color channel representing a texture element or a picture element. By default, the execution unit executes each instruction across all data lanes of operands. In some embodiments, instruction control field 714 enables control of certain execution options such as channel selection (e.g., prediction) and data channel order (e.g., swizzle). For instructions that employ the 128-bit instruction format 710, the execution size field 716 limits the number of data lanes to be executed in parallel. In some embodiments, the execution size field 716 is not available for use in the 64-bit compact instruction format 730.

Some execution unit instructions have up to three operands, including two source operands, src 0720, src 1722, and one destination 718. In some embodiments, the execution unit supports dual destination instructions, where one of the destinations is implicit. The data manipulation instruction can have a third source operand (e.g., SRC 2724), where the instruction opcode 712 determines the number of source operands. The last source operand of the instruction can be an immediate (e.g., hard-coded) value passed with the instruction.

In some embodiments, 128-bit instruction format 710 includes an access/address mode field 726, the access/address mode field 726 specifying, for example, whether direct register addressing mode or indirect register addressing mode is used. When using the direct register addressing mode, the register address of one or more operands is provided directly by bits in the instruction.

In some embodiments, 128-bit instruction format 710 includes an access/address mode field 726, the access/address mode field 726 specifying an address mode and/or an access mode of the instruction. In one embodiment, an access pattern is used to define the data access alignment of an instruction. Some embodiments support access patterns that include 16 byte aligned access patterns and 1 byte aligned access patterns, where the byte alignment of the access patterns determines the access alignment of instruction operands. For example, when in the first mode, the instruction may use byte aligned addressing for the source operand and the destination operand, and when in the second mode, the instruction may use 16 byte aligned addressing for all of the source operand and the destination operand.

In one embodiment, the address mode portion of the access/address mode field 726 determines whether the instruction will use direct addressing or indirect addressing. When using the direct register addressing mode, bits in the instruction directly provide the register address of one or more operands. When using the indirect register addressing mode, register addresses for one or more operands may be calculated based on an address register value and an address immediate field in the instruction.

In some embodiments, instructions are grouped based on opcode 712 bit fields to simplify opcode decoding 740. For an 8-bit opcode, bits 4, 5, and 6 allow the execution unit to determine the type of opcode. The exact opcode groupings shown are examples only. In some embodiments, the move and logical opcode group 742 includes data move and logical instructions (e.g., move (mov), compare (cmp)). In some embodiments, move and logical group 742 share five Most Significant Bits (MSBs), with a move (mov) instruction taking the form 0000 xxxxxxb and a logical instruction taking the form 0001 xxxxb. The group of flow control instructions 744 (e.g., fetch, jump (jmp)) includes instructions in the form of 0010 xxxxxxb (e.g., 0x 20). The miscellaneous instruction group 746 includes a mix of instructions, including synchronous instructions (e.g., wait, send) in the form of 0011 xxxxxxb (e.g., 0x 30). The parallel mathematical instruction group 748 includes component-wise arithmetic instructions (e.g., add, multiply (mul)) in the form of 0100 xxxxxxb (e.g., 0x 40). The parallel math group 748 performs arithmetic operations in parallel across the data channels. The vector math group 750 includes arithmetic instructions (e.g., dp 4) in the form 0101xxxxb (e.g., 0x 50). Vector math groups perform arithmetic such as dot product calculations on vector operands. The illustrated opcode decoding 740 can be used in one embodiment to determine which portion of the execution unit will be used to execute the decoded instruction. For example, some instructions may be designated as systolic instructions to be executed by the systolic array. Other instructions, such as ray tracing instructions (not shown), can be routed to ray tracing cores or ray tracing logic within a slice or partition of execution logic.

Graphics pipeline

Fig. 8 is a block diagram of another embodiment of a graphics processor 800. The elements of fig. 8 having the same reference numbers (or names) as the elements of any other figure herein can operate or function in any manner similar to that described elsewhere herein, but are not limited to such.

In some embodiments, graphics processor 800 includes geometry pipeline 820, media pipeline 830, display engine 840, thread execution logic 850, and render output pipeline 870. In some embodiments, graphics processor 800 is a graphics processor within a multi-core processing system that includes one or more general purpose processing cores. The graphics processor is controlled by register writes to one or more control registers (not shown) or via commands issued to the graphics processor 800 over the ring interconnect 802. In some embodiments, ring interconnect 802 couples graphics processor 800 to other processing components, such as other graphics processors or general purpose processors. Commands from ring interconnect 802 are interpreted by command streamer 803, which command streamer 803 supplies instructions to the various components of geometry pipeline 820 or media pipeline 830.

In some embodiments, command stream transmitter 803 directs the operation of vertex fetcher 805, which vertex fetcher 805 reads vertex data from memory and executes vertex processing commands provided by command stream transmitter 803. In some embodiments, vertex fetcher 805 provides vertex data to vertex shader 807, which vertex shader 807 performs coordinate space transformations and lighting operations on each vertex. In some embodiments, vertex fetcher 805 and vertex shader 807 execute vertex processing instructions by dispatching execution threads to execution units 852A-852B via thread dispatcher 831.

In some embodiments, execution units 852A-852B are an array of vector processors having sets of instructions for performing graphics and media operations. In some embodiments, execution units 852A-852B have an attached L1 cache 851, the L1 cache 851 being specific to each array, or shared between arrays. The cache can be configured as a data cache, an instruction cache, or a single cache partitioned to contain data and instructions in different partitions.

In some embodiments, geometry pipeline 820 includes a tessellation component to perform hardware accelerated tessellation of 3D objects. In some embodiments, a programmable hull shader (programmable hull shader) 811 configures the tessellation operations. The programmable domain shader 817 provides back-end evaluation of the tessellation output. The tessellator 813 operates under the direction of the hull shader 811 and contains dedicated logic to generate a detailed set of geometric objects based on a coarse geometric model provided as input to the geometry pipeline 820. In some embodiments, tessellation components (e.g., hull shader 811, tessellator 813, and domain shader 817) can be bypassed if tessellation is not used.

In some embodiments, a full geometry object can be processed by the geometry shader 819 via one or more threads assigned to the execution units 852A-852B, or can travel directly to a clipper 829. In some embodiments, the geometry shader operates on the entire geometry object, rather than on vertices or patches of vertices (patches) as in previous stages of the graphics pipeline. If tessellation is disabled, geometry shader 819 receives input from vertex shader 807. In some embodiments, the geometry shader 819 may be programmed by a geometry shader program to perform geometry tessellation when the tessellation unit is disabled.

Before rasterization, the clipper 829 processes the vertex data. The clipper 829 may be a programmable clipper or a fixed function clipper with clipping and geometry shader functions. In some embodiments, a rasterizer and depth test component 873 in the render output pipeline 870 dispatches pixel shaders to convert the geometric objects into a per-pixel representation. In some embodiments, pixel shader logic is included in thread execution logic 850. In some embodiments, the application can bypass the rasterizer and depth test component 873 and access the un-rasterized vertex data via the stream output unit 823.

Graphics processor 800 has an interconnect bus, interconnect fabric, or some other interconnect mechanism that allows data and messages to be passed between the main components of the processor. In some embodiments, the execution units 852A-852B and associated logic units (e.g., L1 cache 851, sampler 854, texture cache 858, etc.) are interconnected via data ports 856 to perform memory accesses and communicate with the rendering output pipeline components of the processor. In some embodiments, sampler 854, caches 851, 858 and execution units 852A-852B each have separate memory access paths. In one embodiment, the texture cache 858 can also be configured as a sampler cache.

In some embodiments, the render output pipeline 870 contains a rasterizer and depth test component 873 that converts vertex-based objects into associated pixel-based representations. In some embodiments, the rasterizer logic includes a windower (windower)/masker unit to perform fixed function triangle and line rasterization. An associated render cache 878 and depth cache 879 may also be available in some embodiments. The pixel operations component 877 performs pixel-based operations on the data, although in some examples, pixel operations associated with 2D operations (e.g., bitblock image transfer with blending) are performed by the 2D engine 841 or replaced by the display controller 843 when displaying with an overlay display plane. In some embodiments, shared L3 cache 875 is available to all graphics components, allowing data to be shared without using main system memory.

In some embodiments, graphics processor media pipeline 830 includes a media engine 837 and a video front end 834. In some embodiments, the video front end 834 receives pipeline commands from the command streamer 803. In some embodiments, media pipeline 830 includes a separate command streamer. In some embodiments, the video front end 834 processes media commands before sending the commands to the media engine 837. In some embodiments, media engine 837 includes thread spawning functionality to spawn a thread for dispatch to thread execution logic 850 via thread dispatcher 831.

In some embodiments, graphics processor 800 includes a display engine 840. In some embodiments, display engine 840 is external to processor 800 and is coupled with the graphics processor via ring interconnect 802 or some other interconnect bus or fabric. In some embodiments, display engine 840 includes a 2D engine 841 and a display controller 843. In some embodiments, the display engine 840 contains dedicated logic that can operate independently of the 3D pipeline. In some embodiments, the display controller 843 is coupled with a display device (not shown), which may be a system-integrated display device (as in a laptop computer) or may be an external display device attached via a display device connector.

In some embodiments, geometry pipeline 820 and media pipeline 830 may be configured to perform operations based on multiple graphics and media programming interfaces and are not specific to any one Application Programming Interface (API). In some embodiments, driver software for the graphics processor translates API calls specific to a particular graphics or media library into commands that can be processed by the graphics processor. In some embodiments, support is provided for an open graphics library (OpenGL), open computing language (OpenCL), and/or Vulkan graphics and computing APIs, all from the Khronos Group. In some embodiments, support may also be provided for the Direct3D library from microsoft corporation. In some embodiments, a combination of these libraries may be supported. Support may also be provided for an open source computer vision library (OpenCV). Future APIs with compatible 3D pipelines will also be supported if a mapping from the pipeline of the future API to the pipeline of the graphics processor can be made.

Graphics pipeline programming

FIG. 9A is a block diagram illustrating a graphics processor command format 900 according to some embodiments. FIG. 9B is a block diagram illustrating a graphics processor command sequence 910, according to an embodiment. The solid box in fig. 9A illustrates components 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. The exemplary graphics processor command format 900 of FIG. 9A includes data fields to identify the client 902, command operation code (opcode) 904, and data 906 of the command. Some commands also include a subopcode 905 and a command size 908.

In some embodiments, the client 902 specifies a client unit of the graphics device that processes command data. In some embodiments, the graphics processor command parser examines the client field of each command to adjust the further processing of the command and routes the command data to the appropriate client unit. In some embodiments, a graphics processor client unit includes a memory interface unit, a rendering unit, a 2D unit, a 3D unit, and a media unit. Each client unit has a corresponding processing pipeline that processes commands. Once the client unit receives the command, the client unit reads the operation code 904 and the sub-operation code 905 (if the sub-operation code 905 is present) to determine the operation to perform. The client unit uses the information in data field 906 to execute the command. For some commands, an explicit command size 908 is contemplated to specify the size of the command. In some embodiments, the command parser automatically determines the size of at least some of the commands based on the command opcode. In some embodiments, the commands are aligned via multiples of a doubleword. Other command formats can be used.

The flowchart in FIG. 9B illustrates an exemplary graphics processor command sequence 910. In some embodiments, software or firmware of a data processing system featuring an embodiment of a graphics processor uses the illustrated version of the command sequence to set, execute, and terminate a set of graphics operations. Sample command sequences are shown and described for purposes of example only, as embodiments are not limited to these particular commands or this sequence of commands. Additionally, the commands may be issued as a batch of commands in a sequence of commands such that the graphics processor will process the sequence of commands at least partially concurrently.

In some embodiments, graphics processor command sequence 910 may begin with a pipeline flush command 912 to cause any active graphics pipeline to complete the current pending commands of the pipeline. In some embodiments, 3D pipeline 922 and media pipeline 924 do not operate concurrently. A pipeline flush is performed to cause the active graphics pipeline to complete any pending commands. In response to a pipeline flush, a command parser for a graphics processor will halt command processing until the active drawing engine completes pending operations and the associated read cache is invalidated. Alternatively, any data in the render cache marked as dirty can be flushed to memory. In some embodiments, pipeline flush command 912 can be used for pipeline synchronization or used before placing the graphics processor into a low power state.

In some embodiments, the pipeline select command 913 is used when the command sequence requires the graphics processor to explicitly switch between pipelines. In some embodiments, the pipeline select command 913 is required only once within the execution context before issuing the pipeline command unless the context is to issue commands for both pipelines. In some embodiments, a pipeline flush command 912 is required immediately prior to a pipeline switch via pipeline select command 913.

In some embodiments, pipeline control commands 914 configure the graphics pipeline for operation and are used to program 3D pipeline 922 and media pipeline 924. In some embodiments, the pipeline control commands 914 configure the pipeline state for the active pipeline. In one embodiment, the pipeline control command 914 is used for pipeline synchronization and to flush data from one or more caches within the active pipeline before processing a batch of commands.

In some embodiments, the return buffer status command 916 is used to configure a set of return buffers for the respective pipeline to write data. Some pipelining operations require allocation, selection, or configuration of one or more return buffers into which these operations write intermediate data during processing. In some embodiments, the graphics processor also uses one or more return buffers to store output data and perform cross-thread communications. In some embodiments, return buffer status 916 includes selecting the size and number of return buffers to be used for a set of pipelining operations.

The remaining commands in the command sequence differ based on the active pipeline used for the operation. Based on the pipeline determination 920, the command sequence is customized to either the 3D pipeline 922, which starts in a 3D pipeline state 930, or the media pipeline 924, which starts in a media pipeline state 940.

The commands used to configure the 3D pipeline state 930 include 3D state set commands for vertex buffer state, vertex element state, constant color state, depth buffer state, and other state variables to be configured before processing the 3D primitive command. The values of these commands are determined based at least in part on the particular 3D API in use. In some embodiments, the 3D pipeline state 930 commands can also selectively disable or bypass certain pipeline elements if those elements are not to be used.

In some embodiments, the 3D primitive 932 command is used to submit a 3D primitive to be processed by the 3D pipeline. Commands and associated parameters passed to the graphics processor via the 3D primitive 932 commands are forwarded to vertex fetch functions in the graphics pipeline. The vertex fetch function uses the 3D primitive 932 command data to generate the vertex data structure. The vertex data structure is stored in one or more return buffers. In some embodiments, 3D primitive 932 commands are used to perform vertex operations on 3D primitives via a vertex shader. To process the vertex shader, 3D pipeline 922 dispatches shader execution threads to the graphics processor execution unit.

In some embodiments, the 3D pipeline 922 is triggered via an execute 934 command or event. In some embodiments, the register write triggers the command execution. In some embodiments, execution is triggered via a "go" or "kick" command in the command sequence. In one embodiment, a pipeline synchronization command to flush a sequence of commands through a graphics pipeline is used to trigger command execution. The 3D pipeline will perform geometric processing for the 3D primitives. Once the operation is complete, the resulting geometric object is rasterized and the pixel engine colors the resulting pixels. For those operations, additional commands to control pixel shading and pixel back-end operations may also be included.

In some embodiments, graphics processor command sequence 910 follows the path of media pipeline 924 when performing media operations. In general, the particular use and manner of programming for media pipeline 924 depends on the media or computing operation to be performed. Certain media decoding operations may be offloaded to the media pipeline during media decoding. In some embodiments, the media pipeline can also be bypassed and media decoding can be performed in whole or in part using resources provided by one or more general purpose processing cores. In one embodiment, the media pipeline further includes elements for General Purpose Graphics Processor Unit (GPGPU) operations, wherein the graphics processor is used to perform SIMD vector operations using a compute shader program that is not explicitly related to the rendering of graphics primitives.

In some embodiments, media pipeline 924 is configured in a similar manner as 3D pipeline 922. A set of commands to configure the media pipeline state 940 is dispatched or placed into a command queue prior to the media object command 942. In some embodiments, the commands for the media pipeline state 940 include data to configure the media pipeline elements that will be used to process the media object. This includes data, such as encoding and decoding formats, used to configure the video decoding and video encoding logic within the media pipeline. In some embodiments, the commands for the media pipeline state 940 also support the use of one or more pointers to "indirect" state elements containing a collection of state settings.

In some embodiments, media object command 942 supplies a pointer to a media object for processing by the media pipeline. The media object includes a memory buffer containing video data to be processed. In some embodiments, all of the media pipeline state must be valid before issuing the media object command 942. Once the pipeline state is configured and the media object command 942 is queued, the media pipeline 924 is triggered via an execute command 944 or equivalent execute event (e.g., a register write). The output from media pipeline 924 may then be post-processed by operations provided by 3D pipeline 922 or media pipeline 924. In some embodiments, GPGPU operations are configured and performed in a similar manner as media operations.

Graphics software architecture

FIG. 10 illustrates an exemplary graphics software architecture for data processing system 1000 in accordance with some embodiments. In some embodiments, the software architecture includes a 3D graphics application 1010, an operating system 1020, and at least one processor 1030. In some embodiments, processor 1030 includes a graphics processor 1032 and one or more general purpose processor cores 1034. Graphics application 1010 and operating system 1020 each execute in system memory 1050 of the data processing system.

In some embodiments, 3D graphics application 1010 contains one or more shader programs, including shader instructions 1012. The shader language instructions may be in a high level shader language, such as High Level Shader Language (HLSL) or OpenGL shader language (GLSL) of Direct3D, and so forth. The application also includes executable instructions 1014 in a machine language suitable for execution by the general purpose processor core 1034. The application also includes a graphical object 1016 defined by the vertex data.

In some embodiments, the operating system 1020 is Microsoft Windows ® operating system from Microsoft corporation, a proprietary UNIX-like operating system, or an open source UNIX-like operating system using a variant of the Linux kernel. The operating system 1020 can support a graphics API 1022, such as the Direct3D API, the OpenGL API, or the Vulkan API. When the Direct3D API is in use, the operating system 1020 uses a front-end shader compiler 1024 to compile any shader instructions 1012 that employ HLSL into a lower-level shader language. The compilation may be just-in-time (JIT) compilation or the application may be able to perform shader precompilation. In some embodiments, the high-level shaders are compiled into low-level shaders during compilation of the 3D graphics application 1010. In some embodiments, the shader instructions 1012 are provided in an intermediate form, such as a version of the Standard Portable Intermediate Representation (SPIR) used by the Vulkan API.

In some embodiments, user mode graphics driver 1026 contains a back-end shader compiler 1027 to convert shader instructions 1012 into a hardware-specific representation. When the OpenGL API is in use, shader instructions 1012 in the GLSL high-level language are passed to user-mode graphics driver 1026 for compilation. In some embodiments, the user mode graphics driver 1026 uses operating system kernel mode functions 1028 to communicate with the kernel mode graphics driver 1029. In some embodiments, the kernel mode graphics driver 1029 communicates with the graphics processor 1032 to dispatch commands and instructions.

IP check cash

One or more aspects of at least one embodiment may be implemented by representative code stored on a machine-readable medium which 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 fabricate logic to perform the techniques described herein. Such a representation, referred to as an "IP core," is a reusable unit of logic for an integrated circuit that may be stored on a tangible machine-readable medium as a hardware model that describes the structure of the integrated circuit. The hardware model may be supplied to various customers or manufacturing facilities that load the hardware model on fabrication machines that manufacture integrated circuits. An integrated circuit may be fabricated such that the circuit performs the operations described in association with any of the embodiments described herein.

Fig. 11A is a block diagram illustrating an IP core development system 1100 that may be used to fabricate integrated circuits to perform operations, according to an embodiment. The IP core development system 1100 may be used to generate a modular, reusable design that can be incorporated into a larger design or used to construct an entire integrated circuit (e.g., an SOC integrated circuit). Design facility 1130 is capable of generating software simulations 1110 of IP core designs in a high-level programming language (e.g., C/C + +). Software simulation 1110 can be used to design, test, and verify the behavior of an IP core using simulation model 1112. Simulation model 1112 may include functional, behavioral, and/or timing simulations. A Register Transfer Level (RTL) design 1115 can then be created or synthesized from simulation model 1112. RTL design 1115 is an abstraction of the behavior of an integrated circuit that models the flow of digital signals between hardware registers, including associated logic that executes using the modeled digital signals. In addition to RTL design 1115, lower level designs at the logic level or transistor level may be created, designed, or synthesized. Thus, the specific details of the initial design and simulation may differ.

The RTL design 1115 or equivalent may be further synthesized by the design facility into a hardware model 1120, which hardware model 1120 may employ a Hardware Description Language (HDL) or some other representation of physical design data. The HDL may be further simulated or tested to verify the IP core design. Non-volatile memory 1140 (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 third party fabrication facility 1165. Alternatively, the IP core design may be communicated over a wired connection 1150 or a wireless connection 1160 (e.g., via the Internet). Fabrication facility 1165 may then fabricate an integrated circuit based at least in part on the IP core design. The integrated circuit fabricated can be configured to perform operations according to at least one embodiment described herein.

Fig. 11B illustrates a cross-sectional side view of an integrated circuit package assembly 1170 according to some embodiments described herein. The integrated circuit package assembly 1170 illustrates an implementation of one or more processor or accelerator devices as described herein. The packaging assembly 1170 includes a plurality of hardware logic units 1172, 1174 connected to a substrate 1180. The logic 1172, 1174 may be implemented at least partially in configurable logic or fixed functionality logic hardware, and can include one or more portions of any of the processor core(s), graphics processor(s), or other accelerator device described herein. Each cell of logic 1172, 1174 can be implemented within a semiconductor die and coupled with a substrate 1180 via an interconnect 1173. Interconnect structure 1173 may be configured to route electrical signals between logic 1172, 1174 and substrate 1180, and can include interconnects such as, but not limited to, bumps or posts. In some embodiments, the interconnect fabric 1173 may 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 1172, 1174. In some embodiments, the substrate 1180 is an epoxy-based laminate substrate. In other embodiments, the substrate 1180 may comprise other suitable types of substrates. The package assembly 1170 can be connected to other electrical devices via package interconnect 1183. Package interconnect 1183 may be coupled to a surface of substrate 1180 to route electrical signals to other electrical devices, such as a motherboard, other chipset, or a multi-chip module.

In some embodiments, the logic units 1172, 1174 are electrically coupled with a bridge 1182, the bridge 1182 configured to route electrical signals between the logics 1172, 1174. Bridge 1182 may be a dense interconnect structure that provides routing for electrical signals. The bridge 1182 may 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 1172, 1174.

Although two logic units 1172, 1174 and a bridge 1182 are illustrated, embodiments described herein may include more or fewer logic units on one or more dies. Since bridge 1182 may be excluded when logic is included on a single die, one or more dies may be connected through zero or more bridges. Alternatively, multiple dies or logic units can be connected by one or more bridges. Additionally, multiple logic cells, dies, and bridges can be connected together in other possible configurations, including three-dimensional configurations.

Fig. 11C illustrates a package assembly 1190 comprising a plurality of hardware logic chiplet units connected to a substrate 1180 (e.g., base die). A graphics processing unit, parallel processor, and/or compute accelerator as described herein can be constructed from a multiplicity of silicon chiplets that are individually fabricated. 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. A diverse set of chiplets with different IP core logic can be assembled into a single device. In addition, the 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. Different process technologies can be used to manufacture and build the IP core during manufacturing, which avoids the complexity of converging multiple IPs (especially on large socs with several feature (colors) IPs) to 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. In addition, disaggregated IP is more susceptible to being independently power gated, and components that are not in use on a given workload can be powered down, thereby reducing overall power consumption.

The hardware logic chiplets can include dedicated hardware logic chiplets 1172, logic or I/O chiplets 1174, and/or memory chiplets 1175. Hardware logic chiplet 1172 and logic or I/O chiplets 1174 can be implemented at least in part in configurable logic or fixed functionality logic hardware and can include one or more portions of processor core(s), graphics processor(s), parallel processors, or any of the other accelerator devices described herein. The memory chiplets 1175 can be DRAM (e.g., GDDR, HBM) memory or cache (SRAM) memory.

Each chiplet can be fabricated as a separate semiconductor die and coupled to the substrate 1180 via interconnect structures 1173. Interconnect structure 1173 can be configured to route electrical signals between the various chiplets and logic within substrate 1180. Interconnect structure 1173 can include interconnects such as, but not limited to, bumps or pillars. In some embodiments, interconnect fabric 1173 may 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 logic, I/O, and memory chiplets.

In some embodiments, the substrate 1180 is an epoxy-based laminate substrate. In other embodiments, the substrate 1180 may comprise other suitable types of substrates. The package assembly 1190 can be connected to other electrical devices via a package interconnect 1183. Package interconnect 1183 may be coupled to a surface of substrate 1180 to route electrical signals to other electrical devices, such as a motherboard, other chipset, or a multi-chip module.

In some embodiments, logic or I/O chiplet 1174 and memory chiplet 1175 can be electrically coupled via a bridge 1187, the bridge 1187 configured to route electrical signals between logic or I/O chiplet 1174 and memory chiplet 1175. Bridge 1187 may be a dense interconnect structure that provides routing for electrical signals. The bridge 1187 may 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 chiplets 1174 and the memory chiplets 1175. Bridge 1187 may also be referred to as a silicon bridge or an interconnect bridge. For example, in some embodiments, bridge 1187 is an embedded multi-die interconnect bridge (EMIB). In some embodiments, bridge 1187 may simply be a direct connection from one chiplet to another.

The substrate 1180 can include hardware components for the I/O1191, cache memory 1192, and other hardware logic 1193. Fabric 1185 can be embedded in substrate 1180 to enable communication between various logic chiplets and logic 1191, 1193 within substrate 1180. In one embodiment, I/O1191, fabric 1185, cache, bridges, and other hardware logic 1193 can be integrated into a base die that is stacked on top of substrate 1180.

In various embodiments, package assembly 1190 can include a fewer or greater number of components and chiplets interconnected by fabric 1185 or one or more bridges 1187. The chiplets within the package assembly 1190 can be arranged in a 3D or 2.5D arrangement. In general, bridge structure 1187 may be used to facilitate point-to-point interconnections between, for example, logic or I/O chiplets and memory chiplets. Fabric 1185 can be used to interconnect various logic and/or I/O chiplets (e.g., chiplets 1172, 1174, 1191, 1193) with other logic and/or I/O chiplets. In one embodiment, the in-substrate cache 1192 can act as a global cache for the package assembly 1190, part of a distributed global cache, or as a private cache for the fabric 1185.

FIG. 11D illustrates a packaged assembly 1194 including interchangeable chiplets 1195 according to an embodiment. The interchangeable chiplets 1195 can be assembled into standardized slots on one or more base chiplets 1196, 1198. The base chiplets 1196, 1198 can be coupled via bridge interconnects 1197, which bridge interconnects 1197 can be similar to other bridge interconnects described herein and can be, for example, EMIBs. The memory chiplets can also be connected to logic or I/O chiplets via bridge interconnects. The I/O and logic chiplets are capable of communicating via the interconnect fabric. The base chiplets can each support one or more slots in a standardized format for one of logic or I/O or memory/cache.

In one embodiment, the SRAM and power delivery circuitry can be fabricated into one or more of the base chiplets 1196, 1198, which base chiplets 1196, 1198 can be fabricated using different process technologies relative to the interchangeable chiplets 1195 stacked on top of the base chiplets. For example, the base chiplets 1196, 1198 can be made using larger process technologies, while the interchangeable chiplets can be made using smaller process technologies. One or more of the interchangeable chiplets 1195 can be memory (e.g., DRAM) chiplets. Different memory densities can be selected for the packaged assembly 1194 based on the power and/or performance that is the target of the product in which the packaged assembly 1194 is used. In addition, logic chiplets with different numbers of types of functional units can be selected at assembly based on the power and/or performance targeted for the product. In addition, chiplets containing different types of IP logic cores can be inserted into interchangeable chiplet slots, enabling hybrid processor designs that can mix and match different technology IP blocks.

Exemplary System-on-chip Integrated Circuit

Fig. 12-14 illustrate an example integrated circuit and associated graphics processor that may be fabricated using one or more IP cores, according to various embodiments described herein. Other logic and circuitry may be included in addition to that illustrated, including additional graphics processors/cores, peripheral interface controllers, or general purpose processor cores.

Fig. 12 is a block diagram illustrating an exemplary system-on-chip integrated circuit 1200 that may be fabricated using one or more IP cores, according to an embodiment. The exemplary integrated circuit 1200 includes one or more application processors 1205 (e.g., CPUs), at least one graphics processor 1210, and may additionally include an image processor 1215 and/or a video processor 1220, any of which may be modular IP cores from the same or multiple different design facilities. Integrated circuit 1200 includes peripheral or bus logic including USB controller 1225, UART controller 1230, SPI/SDIO controller 1235, and I2S/I2C controller 1240. Additionally, the integrated circuit can include a display device 1245 coupled to one or more of a High Definition Multimedia Interface (HDMI) controller 1250 and a Mobile Industrial Processor Interface (MIPI) display interface 1255. Storage may be provided by a flash memory subsystem 1260 including flash memory and a flash memory controller. A memory interface may be provided via the memory controller 1265 to access SDRAM or SRAM memory devices. Some integrated circuits additionally include an embedded security engine 1270.

Fig. 13-14 are block diagrams illustrating an exemplary graphics processor for use within a SoC, according to embodiments described herein. FIG. 13 illustrates an example graphics processor 1310 of a system-on-chip integrated circuit that may be fabricated using one or more IP cores, according to an embodiment. FIG. 14 illustrates a further exemplary graphics processor 1340 of a system-on-chip integrated circuit that may be fabricated using one or more IP cores, according to an embodiment. Graphics processor 1310 of FIG. 13 is an example of a low power graphics processor core. Graphics processor 1340 of fig. 14 is an example of a higher performance graphics processor core. Each of the graphics processors 1310, 1340 can be a variation of the graphics processor 1210 of fig. 12.

As shown in FIG. 13, graphics processor 1310 includes a vertex processor 1305 and one or more fragment processors 1315A-1315N (e.g., 1315A, 1315B, 1315C, 1315D through 1315N-1 and 1315N). Graphics processor 1310 is capable of executing different shader programs via separate logic, such that vertex processor 1305 is optimized to perform operations for vertex shader programs, while one or more fragment processors 1315A-1315N perform fragment (e.g., pixel) shading operations for fragment or pixel shader programs. Vertex processor 1305 executes the vertex processing stages of the 3D graphics pipeline and generates primitive and vertex data. The fragment processor(s) 1315A-1315N use the primitive and vertex data generated by the vertex processor 1305 to produce a frame buffer for display on a display device. In one embodiment, fragment processor(s) 1315A-1315N are optimized to execute fragment shader programs as provided for in the OpenGL API, which can be used to perform similar operations as for the pixel shader programs as provided in the Direct 3D API.

Graphics processor 1310 additionally includes one or more Memory Management Units (MMUs) 1320A-1320B, cache(s) 1325A-1325B, and circuit interconnect(s) 1330A-1330B. The one or more MMUs 1320A-1320B provide virtual address to physical address mapping for the graphics processor 1310, including for the vertex processor 1305 and/or the fragment processor(s) 1315A-1315N, which may reference vertex or image/texture data stored in memory in addition to vertex or image/texture data stored in one or more caches 1325A-1325B. In one embodiment, one or more MMUs 1320A-1320B may be synchronized with other MMUs within the system, including one or more MMUs associated with one or more application processors 1205, image processors 1215, and/or video processors 1220 of FIG. 12, enabling each processor 1205-1220 to participate in a shared or unified virtual memory system. According to an embodiment, one or more circuit interconnects 1330A-1330B enable graphics processor 1310 to interface with other IP cores within the SoC via the SoC's internal bus or via a direct connection.

As shown in FIG. 14, graphics processor 1340 includes one or more MMUs 1320A-1320B, cache(s) 1325A-1325B and circuit interconnect(s) 1330A-1330B of graphics processor 1310 of FIG. 13. Graphics processor 1340 includes one or more shader cores 1355A-1355N (e.g., 1355A, 1355B, 1355C, 1355D, 1355E, 1355F through 1355N-1 and 1355N) that provide a unified shader core architecture in which a single core or a single type of core is capable of executing all types of programmable shader code, including shader program code to implement vertex shaders, fragment shaders, and/or compute shaders. The exact number of shader cores present can vary between embodiments and implementations. In addition, the graphic processor 1340 includes: an inter-core task manager 1345, the inter-core task manager 1345 acting as a thread dispatcher to dispatch execution threads to one or more shader cores 1355A-1355N; and a tiling unit 1358 to accelerate tiling operations (tiling operations) for tile-based rendering in which rendering operations for a scene are subdivided in image space, e.g., to take advantage of local spatial coherence within the scene or to optimize internal cache usage.

Ray tracing with machine learning

As mentioned above, ray tracing is a graphics processing technique in which light transmission is emulated through physics-based rendering. One of the key operations in ray tracing is to process visibility queries that require traversal and cross-testing of nodes in a bounding volume level (BVH).

Ray tracing and path tracing based techniques compute images by tracing rays and paths through each pixel and computing high-level effects such as shading, gloss, indirect lighting, etc. using random sampling. Using only a few samples is fast but produces noisy images, while using many samples produces high quality images but is cost prohibitive.

Machine learning includes any circuit, program code, or combination thereof that can incrementally improve the performance of a given task or incrementally render more accurate predictions or decisions. Some machine learning engines may perform these tasks or present these predictions/decisions without being explicitly programmed to perform the tasks or present the predictions/decisions. Various machine learning techniques exist, including (but not limited to) supervised and semi-supervised learning, unsupervised learning, and reinforcement learning.

In recent years, a breakthrough solution for ray tracing/path tracing for real-time use has emerged in the form of "denoising" (the process of using image processing techniques to produce high quality filtered/denoised images from noisy, low sample count inputs). The most effective denoising technique relies on a machine learning technique in which a machine learning engine learns what a noisy image would likely look like if it had been computed with more samples. In one particular implementation, machine learning is performed by a Convolutional Neural Network (CNN); however, the underlying principles of the invention are not limited to CNN implementations. In such an implementation, training data is generated with a low sample count input and ground-truths. The CNN is trained to predict a converging pixel from a neighborhood of noisy pixel inputs around the pixel in question.

Although imperfect, this AI-based denoising technique has proven surprisingly effective. However, it is noted that good training data is required, since otherwise the network may predict erroneous results. For example, if an animated film studio trains a de-noising CNN over a past film with scenes on land and then attempts to de-noise frames from a new film set on water using the trained CNN, the de-noising operation will perform poorly.

To address this issue, learning data may be dynamically collected at rendering time and a machine learning engine (such as CNN) may be continuously trained based on the data for which it is currently running, thereby continuously improving the machine learning engine for upcoming tasks. Thus, the training phase may still be performed before runtime, but the machine learning weights are continually adjusted as needed during runtime. Thus, the high cost of computing the reference data required for training is avoided by limiting the generation of learning data to sub-regions of the image per frame or every N frames. In particular, a noisy input of the frame is generated for denoising the full frame with the current network. Additionally, reference pixels for small regions are generated and used for continuous training, as described below.

Although CNN implementations are described herein, any form of machine learning engine may be used, including, but not limited to, systems that perform supervised learning (e.g., building a mathematical model of a set of data containing inputs and desired outputs), unsupervised learning (e.g., that evaluates input data for certain types of structures), and/or a combination of supervised and unsupervised learning.

Existing denoising implementations operate in a training phase and a runtime phase. During the training phase, a network topology is defined that receives a region of N × N pixels with various per-pixel data channels such as pixel color, depth, normal deviation, primitive ID, and reflectance, and generates a final pixel color. A set of "representative" training data is generated using a low sample count input equivalent to one frame, and reference is made to the "expected" pixel color calculated with a very high sample count. The network is trained towards these inputs, generating a set of "ideal" weights for the network. In these implementations, the reference data is used to train the weights of the network so that the output of the network most closely matches the desired result.

At run time, a given pre-computed ideal network weight is loaded and the network is initialized. For each frame, a low sample count image of the denoised input (i.e., as used for training) is generated. For each pixel, a given neighborhood of the input of the pixel is run through the network to predict the "denoised" pixel color, generating a denoised frame.

FIG. 15 illustrates an initial training implementation. The machine learning engine 1500 (e.g., CNN) receives a region of N × N pixels as high sample count image data 1502 with various per-pixel data channels such as pixel color, depth, normal bias, primitive ID, and reflectance, and generates a final pixel color. Representative training data is generated using a low sample count input 1501 equivalent to one frame. The network is trained towards these inputs, generating a set of "ideal" weights 1505, which the machine learning engine 1500 then uses to denoise low sample count images at run time.

To improve the above technique, the denoising phase for generating new training data per frame or subset of frames (e.g., every N frames, where N =2, 3, 4, 10, 25, etc.) is enhanced. In particular, as illustrated in fig. 16, one or more regions in each frame, referred to herein as "new reference regions" 1602 are selected, which new reference regions 1602 are rendered with a high sample count into a separate high sample count buffer 1604. Low sample count buffer 1603 stores low sample count input frame 1601 (including low sample region 1604 corresponding to new reference region 1602).

The location of the new reference region 1602 may be randomly selected. Alternatively, the location of the new reference region 1602 may be adjusted in a pre-specified manner for each new frame (e.g., using predefined movement of regions between frames, being limited to a specified region in the center of a frame, etc.).

Regardless of how the new reference region is selected, it is used by the machine learning engine 1600 to continually refine and update the trained weights 1605 for denoising. In particular, the reference pixel color from each new reference region 1602 and the noisy reference pixel input from the corresponding low sample count region 1607 are rendered. A supplemental training is then performed on the machine learning engine 1600 using the high sample count reference region 1602 and the corresponding low sample count region 1607. In contrast to the initial training, the training is continuously performed on each new reference region 1602 during runtime — thereby ensuring that the machine learning engine 1600 is accurately trained. For example, per-pixel data channels (e.g., pixel color, depth, normal bias, etc.) may be evaluated, which the machine learning engine 1600 uses to adjust the trained weights 1605. As in the training case (fig. 21), the machine learning engine 1600 is trained towards a set of ideal weights 1605 for removing noise from the low sample count input frame 1601 to generate a denoised frame 1620. However, the trained weights 1605 are continuously updated based on new image characteristics of the new type of low sample count input frame 1601.

The retraining operations performed by machine learning engine 1600 may be performed concurrently in a background process on a Graphics Processor Unit (GPU) or host processor. A rendering loop, which may be implemented as a driver component and/or GPU hardware component, may continue to generate new training data (e.g., in the form of new reference region 1602), which it places in a queue. A background training process executing on the GPU or host processor may continue to read new training data from the queue, retrain the machine learning engine 1600, and update it with new weights 1605 at appropriate intervals.

Fig. 17 illustrates an example of one such implementation in which background training process 1700 is implemented by host CPU 1710. In particular, the background training process 1700 continuously updates the trained weights 1605 using the high sample count new reference region 1602 and the corresponding low sample region 1604, thereby updating the machine learning engine 1600.

As illustrated in FIG. 18A by way of a non-limiting example for a multiplayer online game, different hosts 1820-1822 individually generate reference regions that the background training processes 1700A-C communicate to a server 1800 (such as a game server, for example). The server 1800 then performs training on the machine learning engine 1810 using the new reference regions received from each of the hosts 1821-1822, updating the weights 1805 as previously described. It communicates these weights 1805 to the host 1820 that stores the weights 1605A-C, thereby updating each individual machine learning engine (not shown). Because a large number of reference areas can be provided to the server 1800 in a short period of time, it can efficiently and accurately update the weights for any given application (e.g., online game) executed by the user.

As illustrated in fig. 18B, the different host may generate and share new trained weights (e.g., based on the training/reference area 1602 as previously described) with the server 1800 (e.g., such as a game server), or alternatively, use a peer-to-peer sharing protocol. A machine learning management component 1810 on the server generates a set of combining weights 1805 using the new weights received from each of the hosts. Combining weights 1805 may be, for example, averages generated from new weights and continuously updated as described herein. Once generated, a copy of the combining weights 1605A-C can be transferred and stored on each of the hosts 1820-1821, which hosts 1820-1821 can then perform a denoising operation using the combining weights as described herein.

The semi-closed loop update mechanism may also be used by hardware manufacturers. For example, the reference network may be included as part of a driver distributed by a hardware manufacturer. As the driver generates new training data using the techniques described herein and submits these back to the hardware manufacturer on, the hardware manufacturer uses this information to continue to improve its machine learning implementation for the next driver update.

In an example implementation (e.g., in batch movie rendering on a rendering farm), the renderer transmits the newly generated training area to a dedicated server or database (in the studio's rendering farm) that aggregates the data from multiple rendering nodes over time. Separate processes on separate machines continue to improve the studio's specialized denoising network, and new rendering jobs always use the latest trained network.

The machine learning method is illustrated in fig. 19. The method may be implemented on the architectures described herein, but is not limited to any particular system or graphics processing architecture.

At 1901, low sample count image data and high sample count image data are generated for a plurality of image frames as part of an initial training phase. At 1902, a machine learning denoising engine is trained using the high/low sample count image data. For example, a set of convolutional neural network weights associated with pixel features may be updated according to the training. However, any machine learning architecture may be used.

At 1903, at run time, a low sample count image frame is generated along with at least one reference region having a high sample count. At 1904, the high sample count reference region is used by the machine learning engine and/or separate training logic (e.g., background training module 1700) to continuously improve training of the machine learning engine. For example, a high sample count reference region may be used in conjunction with a corresponding portion of a low sample count image to continually teach the machine learning engine 1904 how to most efficiently perform denoising. In a CNN implementation, for example, this may involve updating the weights associated with the CNN.

Many of the variations described above may be implemented, such as the manner in which the feedback loop is configured to the machine learning engine, the entity that generates the training data, the manner in which the training data is fed back to the training engine, and how the improved network is provided to the rendering engine. Additionally, although the examples described above perform continuous training using a single reference region, any number of reference regions may be used. Also, as previously mentioned, the reference regions may have different sizes, may be used over different numbers of image frames, and may be positioned in different locations within the image frames using different techniques (e.g., randomly, according to a predetermined pattern, etc.).

Additionally, although a Convolutional Neural Network (CNN) is described as one example of machine learning engine 1600, the underlying principles of the invention may be implemented using any form of machine learning engine capable of continuously improving its results using new training data. By way of example and not limitation, other machine implementations include data processing packet methods (GMDH), long-term short-term memory, deep reserve pool computing (deep reserve pool computing), deep belief networks, tensor deep stacking networks, and deep predictive coding networks, to name a few.

Apparatus and method for efficient distributed denoising

As described above, denoising has become a key feature of real-time ray tracing with smooth, noiseless images. Rendering can be done across a distributed system on multiple devices, but the denoising frameworks existing so far all operate on a single instance on a single machine. If rendered across multiple devices, they may not have all of the rendered pixels accessible for computing the denoised portion of the image.

Distributed denoising algorithms are presented that work with Artificial Intelligence (AI) and non-AI based denoising techniques. The regions of the image have been distributed across nodes from a distributed rendering operation, or partitioned and distributed from a single frame buffer. Ghost regions of the neighboring regions needed to compute sufficient denoising are collected from neighboring nodes as needed, and the resulting patches are combined into a final image.

Distributed processing

FIG. 20 illustrates a plurality of nodes 2021 and 2023 that perform rendering. Although only three nodes are illustrated for simplicity, the underlying principles of the invention are not limited to any particular number of nodes. Indeed, a single node may be used to implement some embodiments of the invention.

Nodes 2021 and 2023 each render a portion of the image, thereby generating region 2011 and 2013 in this example. Although rectangular areas 2011-2013 are shown in fig. 20, areas having any shape may be used and any device may handle any number of areas. The region of the node required to perform a sufficiently smooth denoising operation is referred to as a ghost region 2011 and 2013. That is, the ghost region 2001-2003 represents all the data required to perform denoising at a specified quality level. Reducing the quality level reduces the size of the ghost region and thus the amount of data required, and increasing the quality level increases the ghost region and the corresponding data required.

If a node, such as node 2021, does have a local copy of a portion of the ghost region 2001 required to de-noise its region 2011 at a specified quality level, the node will retrieve the required data from one or more "neighboring" nodes, such as node 2022, which possess a portion of the ghost region 2001 as illustrated. Similarly, if node 2022 does have a local copy of a portion of the ghost region 2002 required to de-noise its region 2012 at a specified quality level, node 2022 will retrieve the required ghost region data 2032 from node 2021. The retrieval may be performed over a bus, an interconnect, a high-speed memory fabric, a network (e.g., high-speed ethernet), or may even be an on-chip interconnect in a multi-core chip capable of distributing rendering work among multiple nodes (e.g., for rendering large images at extreme resolutions or time variations). Each node 2021-2023 may comprise a separate execution unit or a designated set of execution units within the graphics processor.

The particular amount of data to be transmitted depends on the denoising technique being used. Moreover, the data from ghost regions may include any data needed to improve denoising of each respective region. For example, ghost region data may include image color/wavelength, intensity/alpha data, and/or normals. However, the underlying principles of the invention are not limited to any particular set of ghost region data.

Additional details

For slower networks or interconnects, the compression of the data may be utilized using existing general lossless or lossy compression. Examples include, but are not limited to, zlib, gzip, and Lempel-Ziv-Markov chain algorithms (LZMA). By noting that the delta in ray hit information between frames may be very sparse, further content specific compression may be used and only the samples that contributed to the delta need be sent when the node already has the delta collected from previous frames. These can be selectively pushed to the nodes that collect those samplesiOr nodeiSamples from other nodes may be requested. Lossless compression is used for certain types of data and program code, while lossy data is used for other types of data.

FIG. 21 illustrates additional details of the interaction between nodes 2021-2022. Each node 2021-2022 includes a ray trace rendering circuit 2081-2082 for rendering a corresponding image area 2011-2012 and a ghost area 2001-2002. The denoiser 2100 + 2111 performs denoising operations on the region 2011 + 2012, and each node 2021 + 2022 is responsible for rendering and denoising. The denoisers 2021 and 2022 may comprise, for example, circuitry, software, or any combination thereof to generate the denoised regions 2121 and 2122, respectively. As mentioned, the denoiser 2021-2022 may need to rely on data within ghost regions owned by different nodes when generating the denoised regions (e.g., the denoiser 2100 may need data from the ghost region 2002 owned by the node 2022).

Thus, denoiser 2100-. The area data manager 2101 & 2102 may manage data transmission from the ghosting area 2001 & 2002 as described herein. The compressor/decompressor units 2131-2132 may perform compression and decompression, respectively, of the ghost region data exchanged between the nodes 2021-2022.

For example, the region data manager 2101 of the node 2021 may, upon request from the node 2022, send data from the shadow region 2001 to the compressor/decompressor 2131, which compressor/decompressor 2131 compresses the data to generate compressed data 2106 that it transmits to the node 2022, thereby reducing bandwidth on the interconnect, network, bus, or other data communication link. The compressor/decompressor 2132 of node 2022 then decompresses the compressed data 2106 and the denoiser 2111 uses the decompressed ghost data to generate a denoised region 2012 of higher quality than would be possible with only data from the region 2012. The region data manager 2102 may store the decompressed data from the ghost region 2001 in a cache, memory, register file, or other storage device to make it available to the denoiser 2111 in generating the denoised region 2122. A similar set of operations may be performed to provide data from the ghost region 2002 to the denoiser 2100 on node 2021, which denoiser 2100 uses this data in conjunction with data from the region 2011 to generate a higher quality denoised region 2121.

Grabbing data or rendering

If the connection between devices such as node 2021-2022 is slow (i.e., below a threshold latency and/or threshold bandwidth), it may be faster to render the ghost region locally rather than request results from other devices. This may be determined at run-time by tracking network transaction speed and linearly extrapolating the number of renderings for ghost region size. In such cases where rendering the entire ghost region is faster, multiple devices may eventually render the same portion of the image. The resolution of the rendered portion of the ghost region may be adjusted based on the variance of the base region and the determined degree of blur.

Load balancing

Static and/or dynamic load balancing schemes may be used to distribute the processing load among the various nodes 2021-2023. For dynamic load balancing, the variance determined by the denoising filter may require more time in denoising, but drives the amount of samples used to render a particular region of the scene, where low variance and blurred regions of the image require fewer samples. The particular region assigned to a particular node may be dynamically adjusted based on data from previous frames or dynamically passed across devices while they are rendering, so that all devices will have the same amount of work.

FIG. 22 illustrates how the monitor 2201- > 2202 running on each respective node 2021- > 2022 collects performance metric data including, but not limited to, the time consumed in transmitting data over the network interface 2211- > 2212, the time consumed in de-noising regions (with and without ghost region data), and the time consumed in rendering each region/ghost region. The monitor 2201-. Manager node 2201 then distributes the new workload for the new zone to nodes 2021 and 2022 based on the detected load. For example, manager node 2201 may communicate more work to those nodes that are not heavily loaded and/or redistribute work from those nodes that are overloaded. Additionally, the load balancer node 2201 may transmit reconfiguration commands to adjust the particular manner in which rendering and/or denoising is performed by each of the nodes (some examples of which are described above).

Determining ghost regions

The size and shape of the ghost region 2001-2002 may be determined based on a denoising algorithm implemented by the denoiser 2100-2111. Their respective sizes may then be dynamically modified based on the detected variance of the sample being denoised. The learning algorithm for AI denoising can itself be used to determine the appropriate region size, or in other cases such as bi-directional blurring, the predetermined filter width will determine the size of the ghost region 2001-2002. In an exemplary implementation using a learning algorithm, the machine learning engine can execute on the manager node 2201 and/or portions of the machine learning can execute on each of the various nodes 2021-2023 (see, e.g., fig. 18A-B and associated text above).

Collecting a final image

The final image may be generated by collecting the rendered and denoised regions from each of the nodes 2021-2023 without ghost regions or normals. In FIG. 22, for example, the denoised regions 2121-2122 are communicated to the region processor 2280 of the manager node 2201, which combines the regions to generate a final denoised image 2290, which final denoised image 2290 is then displayed on the display 2290. The region processor 2280 may combine the regions using various 2D synthesis techniques. Although illustrated as separate components, the region processor 2280 and denoised image 2290 may be integral with the display 2290. The various nodes 2021-2022 may use direct-send techniques to transmit the denoised regions 2121-2122 and potentially use various lossy or lossless compression of the region data.

AI denoising remains a costly operation and moves into the cloud as the game progresses. In this way, it may become desirable to distribute the processing of denoising across multiple nodes 2021-2022 to achieve a real-time frame rate for traditional games or Virtual Reality (VR) that require higher frame rates. Movie studios often also render in large rendering fields that can be utilized for faster denoising.

An exemplary method for performing distributed rendering and denoising is illustrated in fig. 23. The method may be implemented within the context of the system architecture described above, but is not limited to any particular system architecture.

At 2301, a graphics job is dispatched to a plurality of nodes that perform ray tracing operations to render regions of an image frame. Each node may already have the data in memory required to perform the operation. For example, two or more of the nodes may share a common memory or a local memory of the node may have stored data from a previous ray tracing operation. Alternatively, or additionally, some data may be transmitted to each node.

At 2302, the "ghost regions" required to specify a level of denoising (i.e., at an acceptable performance level) are determined. The ghost region includes any data required to perform a specified level of denoising, including data owned by one or more other nodes.

At 2303, data relating to the ghost region (or portions thereof) is exchanged between nodes. At 2304, each node performs denoising (e.g., using the exchanged data) on its respective region and combines the results at 2305 to generate a final denoised image frame.

A manager node or master node such as that shown in fig. 22 may dispatch work to the nodes and then combine the work performed by the nodes to generate a final image frame. A peer-based architecture may be used where nodes are peers that exchange data to render and de-noise the final image frame.

The nodes described herein (e.g., nodes 2021-2023) may be graphics processing computing systems interconnected via a high-speed network. Alternatively, the nodes may be individual processing elements coupled to a high-speed memory fabric. All nodes may share a common virtual memory space and/or a common physical memory. Alternatively, the node may be a combination of a CPU and a GPU. For example, manager node 2201 described above may be a CPU and/or software executing on a CPU and nodes 2021-2022 may be GPUs and/or software executing on GPUs. Various different types of nodes may be used while still complying with the underlying principles of the invention.

Example neural network implementation

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

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 representing the problem being modeled by the network, and adjusting weights until the network model executes with minimal error on all instances of the training data set. For example, during a supervised learning training process for a neural network, the output produced by the network in response to an input representing an instance in a 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 the weights associated with the connections are adjusted to minimize the error as it propagates back through the layers of the network. The network is considered "trained" when the error of each of the outputs generated from the instances of the training data set is minimized.

The accuracy of machine learning algorithms can be significantly affected by the quality of the data set used to train the algorithm. The training process may be computationally intensive and may require a significant amount of time on a conventional general purpose processor. Thus, parallel processing hardware is used to train many types of machine learning algorithms. This is particularly useful for training of optimized neural networks, as the computations performed in adjusting the coefficients in the neural network are naturally applicable to parallel implementations. In particular, many machine learning algorithms and software applications have been adapted to utilize parallel processing hardware within general purpose graphics processing devices.

Fig. 24 is a general diagram of a machine learning software stack 2400. The machine learning application 2402 may be configured to train a neural network using a training data set or implement machine intelligence using a trained deep neural network. The machine learning application 2402 may include specialized software that can be used to train the neural network prior to deployment and/or the training and reasoning functionality of the neural network. The machine learning application 2402 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.

Hardware acceleration for the machine learning application 2402 may be implemented via the machine learning framework 2404. The machine learning framework 2404 may be implemented on hardware described herein, such as the processing system 100 including the processors and components described herein. Elements described with respect to fig. 24 having the same or similar names as the elements of any other figure herein describe the same elements as in the other figures, may operate or function in a similar manner as the same elements in the other figures, may include the same components, and may be linked to other entities (such as, but not limited to, those described elsewhere herein). The machine learning framework 2404 may provide a library of machine learning primitives. Machine learning primitives are the basic operations typically performed by machine learning algorithms. Without the machine learning framework 2404, developers of machine learning algorithms would be required to create and optimize the primary computational logic associated with the machine learning algorithms, and then re-optimize the computational logic as new parallel processors are developed. Instead, the machine learning application may be configured to perform the necessary computations using primitives provided by machine learning framework 2404. Exemplary primitives include tensor convolution, activation functions, and pooling, which are computational operations performed in training a Convolutional Neural Network (CNN). The machine learning framework 2404 may also provide primitives used to implement basic linear algebra subroutines (such as matrix and vector operations) performed by many machine learning algorithms.

The machine learning framework 2404 can process input data received from the machine learning application 2402 and generate suitable inputs to the computing framework 2406. The compute framework 2406 may abstract underlying instructions provided to the GPGPU driver 2408 to enable the machine learning framework 2404 to leverage hardware acceleration via the GPGPU hardware 2410 without requiring the machine learning framework 2404 to have in depth knowledge of the architecture of the GPGPU hardware 2410. Additionally, the compute framework 2406 may implement hardware acceleration for the machine learning framework 2404 across various types and generations of GPGPU hardware 2410.

GPGPU machine learning acceleration

Fig. 25 illustrates a multi-GPU computing system 2500, which can be a variation of processing system 100. Accordingly, a corresponding combination with the multi-GPU computing system 2500 is also disclosed herein in connection with a disclosure of any feature of the processing system 100, but is not limited thereto. Elements of fig. 25 having the same or similar names as elements of any other figure herein describe the same elements as in the other figures, may operate or function in a similar manner as the same elements in the other figures, may include the same components, and may be linked to other entities (as described elsewhere herein, but not limited thereto). The multi-GPU computing system 2500 can include a processor 2502 coupled to a plurality of GPGPUs 2506A-D via a host interface switch 2504. The host interface switch 2504 may be, for example, a PCI express switch that couples the processor 2502 to a PCI express bus through which the processor 2502 may communicate with the set of GPGPUs 2506A-D. Each GPGPU of the multiple GPGPGPGPGPUs 2506A-D may be an example of a GPGPU described above. GPGPGPUs 2506A-D may be interconnected via a set of high speed point-to-point GPU-to-GPU links 2516. The high-speed GPU-to-GPU link may be connected to each of the GPGPUs 2506A-D via a dedicated GPU link. P2P GPU link 2516 enables direct communication between each of the GPGPGPUs 2506A-D without requiring communication through a host interface bus to which processor 2502 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 the multi-GPU computing system 2500, e.g., via one or more network devices. Instead of connecting GPGPUs 2506A-D to processor 2502 via host interface switch 2504, processor 2502 may include direct support for P2P GPU link 2516 and thus connect directly to GPGPGPUs 2506A-D.

Machine learning neural network implementation

The computing architecture described herein may be configured to perform parallel processing of a type particularly suited to training and deploying neural networks for machine learning. A neural network can be generalized as a network having a function of a graph relationship. As is well known in the art, there are various types of neural network implementations used in machine learning. One exemplary type of neural network is a feed-forward network, as previously described.

A second exemplary type of neural network is a Convolutional Neural Network (CNN). CNN is a specialized feed-forward neural network for processing data with a known mesh-like topology, such as image data. CNNs are therefore 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 CNN input layer are organized as a set of "filters" (feature detectors inspired by the receptive field found in the retina), and the output of each set of filters is propagated to nodes in successive layers of the network. The calculation of the CNN includes applying a convolution mathematical operation to each filter to produce the output of that filter. Convolution is a special type of mathematical operation performed by two functions to produce a third function, which is a modified version of one of the two original functions. In convolutional network terminology, the first function to convolution may be referred to as the input, while the second function may be referred to as the convolution kernel. The output may be referred to as a profile. For example, the input to the convolutional layer may be a multi-dimensional array of data defining various color components of the input image. The convolution kernel may be a multi-dimensional array of parameters, where these parameters are adapted through a training process for the neural network.

A Recurrent Neural Network (RNN) is a series of feed-forward neural networks that include feedback connections between layers. RNNs enable modeling of sequential data by sharing parametric data across different parts of a neural network. The architecture of the RNN includes loops. The loop represents the effect of the current value of the variable on its own value at a future time, since 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 linguistic processing due to the variable nature in which linguistic data can be composed.

The diagrams described below present exemplary feed-forward, CNN, and RNN networks, and describe general procedures for training and deploying each of those types of networks, respectively. It will be appreciated that these descriptions are exemplary and non-limiting and that the illustrated concepts are generally applicable to deep neural networks and machine learning techniques.

The exemplary neural networks described above may be used to perform deep learning. Deep learning is machine learning using a deep neural network. A deep neural network used in deep learning is an artificial neural network composed of a plurality of hidden layers, as opposed to a shallow neural network including only a single hidden layer. Deeper neural networks are typically more computationally intensive to train. However, the additional hidden layer of the network enables multi-step pattern recognition, which results in reduced output errors relative to shallow machine learning techniques.

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

Once the neural network is constructed, a learning model may be applied to the network to train the network to perform a particular task. The learning model describes how to adjust the weights within the model to reduce the output error of the network. Back propagation of errors is a common method for training neural networks. The input vectors are presented to the network for processing. The output of the network is compared to the expected output using a loss function and an error value is calculated for each of the neurons in the output layer. These error values are then propagated back until each neuron has an associated error value that approximately represents its contribution to the original output. The network may then learn from those errors using an algorithm, such as a random gradient descent algorithm, to update the weights of the neural network.

26-27 illustrate exemplary convolutional neural networks. Fig. 26 illustrates various layers within a CNN. As shown in fig. 26, an exemplary CNN for modeling image processing may receive an input 2602 that describes red, green, and blue (RGB) components of an input image. Input 2602 may be processed by multiple convolutional layers (e.g., convolutional layers 2604, 2606). Output from multiple convolutional layers may optionally be processed by a set of fully-connected layers 2608. Neurons in a fully connected layer have a full connection to all activations in a previous layer, as described previously for the feed forward network. Output from the fully connected layer 2608 may be used to generate output results from the network. Activation within fully connected layer 2608 may be computed using matrix multiplication rather than convolution. Not all CNN implementations utilize a fully connected layer. For example, in some implementations, convolutional layer 2606 may generate an output for the CNN.

Convolutional layers are sparsely connected, unlike traditional neural network configurations found in fully connected layers 2608. The traditional neural network layer is fully connected such that each output unit interacts with each input unit. However, because the output of the convolution of the domain (rather than the respective state values of each of the nodes in the domain) is input to the nodes of the subsequent layers, the convolution layers are sparsely connected, as illustrated. 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 convolution layer is one aspect that enables the CNN to be scaled to handle large images.

Fig. 27 illustrates exemplary computation stages within the convolutional layer of CNN. CNN's input 2712 to convolutional layer may be processed in three stages of convolutional layer 2714. These three stages may include a convolution stage 2716, a detector stage 2718, and a pooling stage 2720. Convolutional layer 2714 can then output data to the continuous convolutional layer. The last convolutional layer of the network may generate output profile data or provide input to the fully-connected layer, e.g., to generate a classification value for the input to the CNN.

Several convolutions are performed in parallel with generating a set of linear activations in convolution stage 2716. Convolution stage 2716 may include an affine transform, which is any transform that may be specified as a linear transform plus a translation (translation). Affine transformations include rotation, translation, scaling, and combinations of these transformations. The convolution stage computes the output of a function (e.g., a neuron) that is connected to a particular region in the input, which may be determined to be a local region associated with the neuron. The neuron calculates a dot product between the weight of the neuron and the region in the local input to which the neuron is connected. The output from convolution stage 2716 defines a set of linear activations that are processed by successive stages of convolution layer 2714.

The linear activation may be processed by the detector stage 2718. In the detector stage 2718, each linear activation is processed by a nonlinear activation function. The nonlinear activation function increases the nonlinear properties of the overall network without affecting the corresponding domain of the convolutional layer. Several types of non-linear activation may be used. One particular type is a modified linear unit (ReLU) that uses an activation function defined as f (x) = max (0, x) such that the activation threshold is at zero.

The pooling stage 2720 uses a pooling function that replaces the output of the convolutional layer 2706 with the summary statistics of nearby outputs. The pooling function may be used to introduce translation invariance into the neural network such that small translations to the input do not change the pooled output. Invariance to local translation may be useful in scenarios where the presence of features in the input data is more important than the precise location of the features. Various types of pooling functions may be used during the pooling stage 2720, including maximum pooling, average pooling, and l2 norm pooling. Additionally, some CNN implementations do not include a pooling stage. Instead, such an implementation is replaced with an additional convolution stage with an increased step size (stride) relative to the previous convolution stage.

The output from convolutional layer 2714 may then be processed by the next layer 2722. The next layer 2722 may be one of the additional convolutional layers or the fully connected layer 2708. For example, the first convolutional layer 2704 of fig. 27 may output to the second convolutional layer 2706, and the second convolutional layer may output to the first layer of the fully connected layer 2808.

Fig. 28 illustrates an exemplary recurrent neural network 2800. In a Recurrent Neural Network (RNN), a previous state of the network affects the output of the current state of the network. RNNs may be constructed in various ways using various functions. The use of RNNs generally revolves around the use of mathematical models to predict the future based on a priori sequences of inputs. For example, RNNs can be used to perform statistical language modeling to predict an upcoming word given a previous sequence of words. The illustrated RNN 2800 may be described as having an input layer 2802 to receive input vectors, an implication layer 2804 to implement a looping function, a feedback mechanism 2805 to implement "memory" of previous states, and an output layer 2806 to output results. The RNN 2800 operates on a time step basis. The state of the RNN at a given time step is affected based on previous time steps via the feedback mechanism 2805. For a given time step, the state of the implied layer 2804 is defined by the inputs of the previous state and the current time step. The initial input at the first time step (x 1) may be processed by the hidden layer 2804. The second input (x 2) may be processed by the implication layer 2804 using state information determined during processing of the initial input (x 1). A given state may be calculated as s _ t = f (Ux _ t + Ws _ (t-1)), where U and W are parameter matrices. The function f is typically non-linear, such as a hyperbolic tangent function (Tanh) or a variant of the modification function f (x) = max (0, x). However, the particular mathematical function used in the hidden layer 2804 may vary depending on the particular implementation details of the RNN 2800.

Variations of these networks may be implemented in addition to the basic CNN and RNN networks described. One example RNN variant is a Long Short Term Memory (LSTM) RNN. LSTM RNNs are able to learn long-term dependencies, which may be necessary to process longer language sequences. A variant of CNN is a convolutional deep belief network, which has a similar structure to CNN and is trained in a similar way to deep belief networks. A Deep Belief Network (DBN) is a generative neural network that consists of layers of stochastic (random) variables. The DBN may 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 an optimal initial set of weights for the neural network.

Fig. 29 illustrates training and deployment of a deep neural network. Once a given network has been constructed for a task, the neural network is trained using the training data set 2902. Various training frameworks 2904 have been developed to implement hardware acceleration of the training process. For example, the machine learning framework described above may be configured as a training framework. A training framework 2904 may be mounted into the untrained neural network 2906 and enable the untrained neural network to be trained using parallel processing resources described herein to generate a trained neural network 2908.

To begin the training process, the initial weights may be selected randomly or by pre-training using a deep belief network. The training cycle is then executed in a supervised or unsupervised manner.

Supervised learning is a learning method in which training is performed as an intervening operation, such as when the training data set 2902 includes inputs paired with desired outputs for the inputs, or where the training data set includes inputs with known outputs and the outputs of the neural network are manually ranked. The network processes the inputs and compares the resulting outputs to a set of expected or expected outputs. The error is then propagated back through the system. The training framework 2904 may make adjustments to adjust the weights that control the untrained neural network 2906. The training framework 2904 may provide tools to monitor how well the untrained neural network 2906 converges to a model suitable for generating a correct response based on known input data. The training process is repeated as the weights of the network are adjusted to improve the output of the neural network generation. The training process may continue until the neural network reaches a statistically desired accuracy associated with the trained neural network 2908. The trained neural network 2908 may then be deployed to implement any number of machine learning operations.

Unsupervised learning is a learning method in which the network attempts to train itself using unlabeled data. Thus, for unsupervised learning, the training data set 2902 will include input data without any associated output data. The untrained neural network 2906 may learn the groupings within the unlabeled inputs and may determine how the individual inputs correlate to the overall data set. Unsupervised training may be used to generate a self-organizing map, which is a class of trained neural networks 2907 that are capable of performing operations useful in reducing dimensionality of data. Unsupervised training may 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 data.

Variations of supervised and unsupervised training may also be employed. Semi-supervised learning is a technique in which the training data set 2902 includes a mixture of identically distributed labeled and unlabeled data. Incremental learning is a variant of supervised learning in which input data is continuously used to further train the model. Incremental learning enables the trained neural network 2908 to adapt to new data 2912 without forgetting the knowledge instilled within the network during initial training.

The training process, particularly for deep neural networks, whether supervised or unsupervised, may be too computationally intensive for a single compute node. Instead of using a single compute node, a distributed network of compute nodes may be used to accelerate the training process.

Fig. 30A is an exemplary block diagram illustrating distributed learning. Distributed learning is a training model that uses a plurality of distributed computing nodes (such as the nodes described above) to perform supervised or unsupervised training of a neural network. The distributed computing nodes may each include one or more host processors and one or more of the general purpose processing nodes, such as highly parallel general purpose graphics processing units. As illustrated, distributed learning may be performed by model parallel 3002, data parallel 3004, or a combination of model and data parallel.

In model parallel 3002, different compute nodes in a distributed system may perform training computations on different parts of a single network. For example, each layer of the neural network may be trained by a different processing node of the distributed system. Benefits of model parallelism include the ability to scale to particularly large models. Splitting computations associated with different layers of a neural network enables training of very large neural networks where the weights of all layers would not fit into the memory of a single compute node. In some instances, model parallelism can be particularly useful in performing unsupervised training of large neural networks.

In data parallelism 3004, 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 approaches all require techniques to combine results and synchronize model parameters between each node. Exemplary methods of combining data include parameter averaging and update-based data parallelism. Parameter averaging each node is trained on a subset of the training data and global parameters (e.g., weights, biases) are set 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 updates to the model are passed instead of passing parameters from the nodes to the parameter server. Additionally, update-based data parallelism can be performed in a decentralized manner, where updates are compressed and passed between nodes.

The combined model and data parallelism 3006 can be implemented, for example, in a distributed system where each compute node includes multiple GPUs. Each node may have a complete instance of the model, with a separate GPU within each node being used to train different portions of the model.

Distributed training has increased overhead relative to training on a single machine. However, the parallel processors and gpgpgpu described herein may each implement various techniques to reduce overhead for distributed training, including techniques to achieve high bandwidth GPU-to-GPU data transfer and accelerate remote data synchronization.

Exemplary machine learning application

Machine learning may be applied to solve a 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. Applications of computer vision range from rendering human visual capabilities (such as recognizing faces) to creating new classes of visual capabilities. For example, a computer vision application may be configured to recognize sound waves from vibrations induced in objects visible in a video. Parallel processor accelerated machine learning enables computer vision applications to be trained using significantly larger training data sets than previously possible and inference systems to be deployed using low power parallel processors.

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 may be used to train a driving model based on a data set that defines an appropriate response to a particular training input. The parallel processors described herein may enable fast training of increasingly complex neural networks for autonomous driving solutions and enable deployment of low-power inference processors suitable for integration into mobile platforms within autonomous vehicles.

Parallel processor accelerated deep neural networks have implemented machine learning approaches to Automatic Speech Recognition (ASR). ASR involves creating a function that computes the most likely sequence of languages 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.

Parallel processor accelerated machine learning may also be used to accelerate natural language processing. The automatic learning process may 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.

Parallel processing platforms for machine learning can be divided into training platforms and deployment platforms. The training platform is generally highly parallel and includes optimizations to accelerate multi-GPU single-node training and multi-node multi-GPU training. Exemplary parallel processors suitable for training include highly parallel general purpose graphics processing units and/or multiple GPU computing systems described herein. In contrast, deployed machine learning platforms generally include lower power parallel processors suitable for use in products such as cameras, autonomous robots, and autonomous vehicles.

FIG. 30B illustrates an exemplary inference System on a chip (SOC) 3100 suitable for performing inference using a trained model. Elements of fig. 30B that have the same or similar names as elements of any other figure herein describe the same elements as in the other figures, may operate or function in a similar manner as the same elements in the other figures, may include the same components, and may be linked to other entities (such as, but not limited to, those described elsewhere herein). SOC 3100 may integrate processing components, including media processor 3102, vision processor 3104, GPGPU 3106, and multicore processor 3108. The SOC 3100 may additionally include an on-chip memory 3105, which may implement a shared on-chip data pool accessible by each of the processing components. The processing components may be optimized for low power operation to enable deployment to various machine learning platforms, including autonomous vehicles and autonomous robots. For example, one implementation of SOC 3100 may be used as part of a master control system of an autonomous vehicle. Where SOC 3100 is configured for use in an autonomous vehicle, the SOC is designed and configured to comply with functional safety standards related to deployment authorities.

During operation, the media processor 3102 and the vision processor 3104 may work in concert to accelerate computer vision operations. The media processor 3102 may implement low-latency decoding of multiple high-resolution (e.g., 4K, 8K) video streams. The decoded video stream may be written to a buffer in the on-chip memory 3105. The vision processor 3104 may then parse the decoded video and perform preliminary processing operations on the frames of the decoded video in preparation for processing the frames using the trained image recognition model. For example, the vision processor 3104 may accelerate the convolution operations of CNN for performing image recognition on high-resolution video data, while the back-end model calculations are performed by the GPGPU 3106.

The multi-core processor 3108 may include control logic to facilitate the sequencing and synchronization of data transfers and shared memory operations performed by the media processor 3102 and the visual processor 3104. The multi-core processor 3108 may also function as an application processor to execute software applications that may utilize the speculative computing capabilities of the GPGPU 3106. For example, at least a portion of the navigation and driving logic may be implemented in software executing on the multicore processor 3108. Such software may issue the computational workload directly to the GPGPU 3106, or the computational workload may be issued to the multi-core processor 3108, which multi-core processor 3108 may then offload at least a portion of those operations to the GPGPU 3106.

The GPGPU 3106 may include a low power configuration of processing clusters, such as processing clusters DPLAB06A-DPLAB06H within a highly parallel general purpose graphics processing unit DPLAB 00. The V-collar cluster within GPGPU 3106 may support instructions that are specifically optimized to perform inference calculations on a trained neural network. For example, the GPGPU 3106 may support instructions to perform low precision computations such as 8-bit and 4-bit integer vector operations.

Ray tracing architecture

In one implementation, a graphics processor includes circuitry and/or program code to perform real-time ray tracing. A dedicated set of ray tracing cores may be included in a graphics processor to perform the various ray tracing operations described herein, including ray traversal and/or ray intersection operations. In addition to ray tracing cores, sets of graphics processing cores to perform programmable shading operations and sets of tensor cores to perform matrix operations on tensor data may be included.

FIG. 31 illustrates an exemplary portion of one such Graphics Processing Unit (GPU) 3105 comprising a dedicated set of graphics processing resources arranged in multi-core groups 3100A-N. Graphics Processing Unit (GPU) 3105 may be a variation of the following disclosed herein: graphics processor 300, GPGPU 1340, and/or any other graphics processor. Accordingly, disclosure of any feature of the graphics processor also discloses a corresponding combination with the CPU 3105, but is not limited thereto. Further, elements of fig. 31 having the same or similar names as elements of any other figure herein describe the same elements as in the other figures, may operate or function in a similar manner as the same elements in the other figures, may include the same components, and may be linked to other entities (as described elsewhere herein, but not limited thereto). While details are provided of only a single multi-core group 3100A, it will be appreciated that other multi-core groups 3100B-N may be equipped with the same or similar set of graphics processing resources.

As illustrated, the multi-core group 3100A may include a set of graphics cores 3130, a set of tensor cores 3140, and a set of ray trace cores 3150. Scheduler/dispatcher 3110 schedules and dispatches graphics threads for execution on various cores 3130, 3140, 3150. The set of register files 3120 store operand values used by the cores 3130, 3140, 3150 in executing the graphics threads. These registers may include, for example, integer registers for storing integer values, floating point registers for storing floating point values, vector registers for storing packed data elements (integer and/or floating point data elements), and tile registers (tile registers) for storing tensor/matrix values. The tile registers may be implemented as a combined set of vector registers.

One or more level 1 (L1) caches and texture units 3160 store graphics data, such as texture data, vertex data, pixel data, ray data, bounding volume data, and the like, locally within each multi-core group 3100A. A level 2 (L2) cache 3180 shared by all or a subset of the multi-core groups 3100A-N stores graphics data and/or instructions for multiple concurrent graphics threads. As illustrated, the L2 cache 3180 may be shared across multiple multi-core groups 3100A-N. One or more memory controllers 3170 couple the GPU 3105 to memory 3198, which memory 3198 may be system memory (e.g., DRAM) and/or dedicated graphics memory (e.g., GDDR6 memory).

Input/output (IO) circuitry 3195 couples GPU 3105 to one or more IO devices 3195, such as a Digital Signal Processor (DSP), network controller, or user input device. On-chip interconnects may be used to couple I/O devices 3190 to GPU 3105 and memory 3198. One or more IO memory management units (IOMMUs) 3170 of IO circuitry 3195 couple IO device 3190 directly to system memory 3198. IOMMU 3170 may manage multiple sets of page tables to map virtual addresses to physical addresses in system memory 3198. In addition, IO device 3190, CPU(s) 3199, and GPU(s) 3105 may share the same virtual address space.

IOMMU 3170 may also support virtualization. In this case, it may manage a first set of page tables to map guest/graphics virtual addresses to guest/graphics physical addresses and a second set of page tables to map guest/graphics physical addresses to system/host physical addresses (e.g., within system memory 3198). The base of each of the first and second sets of page tables may be stored in a control register and swapped out on a context switch (e.g., so that access to the relevant set of page tables is provided for the new context). Although not illustrated in FIG. 31, each of the cores 3130, 3140, 3150 and/or the multi-core groups 3100A-N may include a Translation Lookaside Buffer (TLB) to cache guest virtual-to-guest physical translations, guest physical-to-host physical translations, and guest virtual-to-host physical translations.

CPU 3199, GPU 3105, and IO device 3190 may be integrated on a single semiconductor chip and/or chip package. The illustrated memory 3198 may be integrated on the same chip or may be coupled to the memory controller 3170 via an off-chip interface. In one implementation, memory 3198 includes GDDR6 memory, which GDDR6 memory shares the same virtual address space as other physical system-level memory, although the underlying principles of the invention are not limited to this particular implementation.

The tensor core 3140 includes a plurality of execution units specifically designed to perform matrix operations, which are basic computation operations used to perform deep learning operations. For example, simultaneous matrix multiplication operations may be used for neural network training and reasoning. The tensor core 3140 may perform matrix processing using various operand precisions, including single precision floating point (e.g., 32 bits), half precision floating point (e.g., 16 bits), integer word (16 bits), byte (8 bits), and nibble (4 bits). The neural network implementation may also extract features of each rendered scene, potentially combining details from multiple frames to construct a high quality final image.

In a deep learning implementation, parallel matrix multiplication work may be scheduled for execution on the tensor core 3140. Training of neural networks requires, in particular, a large number of matrix dot-product operations. To handle the inner product formulation of the nx N matrix multiplication, the tensor kernel 3140 may include at least N dot product processing elements. Before the start of matrix multiplication, a complete matrix is loaded into the tile register, and at least one column of the second matrix is loaded in each of the N cycles. At each cycle, there are N dot products processed.

Depending on the particular implementation, the matrix elements may be stored with different precisions, including 16-bit words, 8-bit bytes (e.g., INT 8), and 4-bit nibbles (e.g., INT 4). Different precision modes can be specified for the tensor core 3140 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).

Ray tracing core 3150 may be used to accelerate ray tracing operations for both real-time ray tracing and non-real-time ray tracing implementations. In particular, the ray tracing core 3150 may include ray traversal/intersection circuitry to perform ray traversal using a bounding volume level (BVH) and identify intersections between rays and primitives enclosed within the BVH volume. Ray tracing core 3150 may also include circuitry for performing depth testing and culling (e.g., using a Z-buffer or similar arrangement). In one implementation, ray tracing kernel 3150 performs traversal and intersection operations in conjunction with the image denoising techniques described herein, at least a portion of which may be performed on tensor kernel 3140. For example, the tensor kernel 3140 may implement a deep learning neural network to perform denoising of frames generated by the ray tracing kernel 3150. However, the CPU(s) 3199, graphics kernel 3130, and/or ray tracing kernel 3150 may also implement all or a portion of a denoising and/or deep learning algorithm.

Additionally, as described above, a distributed approach to denoising may be employed, where the GPU 3105 is in a computing device coupled to other computing devices through a network or high speed interconnect. Additionally, interconnected computing devices may share neural network learning/training data to improve the speed with which the overall system learning performs denoising for different types of image frames and/or different graphics applications.

Ray tracing core 3150 may handle all BVH traversals and ray-primitive intersections, protecting graphics core 3130 from being overloaded with thousands of instructions per ray. Each ray tracing core 3150 may include a first set of dedicated circuitry for performing bounding box tests (e.g., for traversal operations) and a second set of dedicated circuitry for performing ray-triangle intersection tests (e.g., intersecting traversed rays). Thus, the multi-core group 3100A can simply launch the ray probe, and the ray tracing core 3150 independently performs ray traversal and intersection and returns hit data (e.g., hit, no hit, multiple hits, etc.) to the thread context. While the ray tracing core 3150 performs traversal and intersection operations, the other cores 3130, 3140 are released to perform other graphics or computational work.

Each ray tracing core 3150 may include a traversal unit to perform BVH test operations and an intersection unit to perform ray-primitive intersection tests. The crossbar unit may then generate a "hit", "no hit", or "multiple hit" response, and the crossbar unit provides the response to the appropriate thread. During traversal and intersection operations, execution resources of other cores (e.g., graphics core 3130 and tensor core 3140) may be freed to perform other forms of graphics work.

A hybrid rasterization/ray tracing approach may also be used in which work is distributed between graphics kernel 3130 and ray tracing kernel 3150.

The ray tracing core 3150 (and/or other cores 3130, 3140) may include hardware support for ray tracing instruction sets such as Microsoft's DirectX ray tracing (DXR), which includes the DispatchRays commands as well as ray generation, closest hits, arbitrary hits, and miss shaders, which enable each object to be assigned a unique set of textures and shaders. Another ray tracing platform that may be supported by ray tracing kernel 3150, graphics kernel 3130 and tensor kernel 3140 is Vulkan 1.1.85. Note, however, that the underlying principles of the invention are not limited to any particular ray tracing ISA.

In general, the various cores 3150, 3140, 3130 may support a ray trace instruction set that includes instructions/functions for ray generation, closest hits, arbitrary hits, ray-primitive intersections, per-primitive and hierarchical bounding box constructions, misses, accesses, and exceptions. More specifically, ray tracing instructions may be included to perform the following functions:

light generationThe ray generation instructions may be executed for each pixel, sample or other user-defined job assignment.

Closest hitThe closest hit instruction may be executed to locate the closest intersection point of the ray with a primitive within the scene.

Hit arbitrarilyAny hit instruction identifies a number of intersections between primitives and rays within the scene, potentially identifying a new closest intersection point.

Crossing-the intersection instruction performs a ray-primitive intersection test and outputs a result.

Per-primitive bounding box structureThis instruction builds a bounding box around a given primitive or group of primitives (e.g., when building a new BVH or other acceleration data structure).

MissIndicating that the ray missed all geometric shapes within the scene or a specified region of the scene.

AccessSub-volumes indicating the ray will traverse.

Abnormality (S)Including various types of exception handlers (e.g., invoked for various error conditions).

The media processing circuitry 3197 includes fixed-function circuitry to encode, decode, or transcode media into and between one or more media encoding formats, including, but not limited to, Motion Picture Experts Group (MPEG) formats (e.g., MPEG-2), Advanced Video Coding (AVC) formats (e.g., h.264/MPEG-4 AVC), and the Society of Motion Picture and Television Engineers (SMPTE)421M/VC-1 and Joint Photographic Experts Group (JPEG) formats (e.g., JPEG and motion JPEG (mjpeg)) formats. To process these formats, one embodiment of the media processing circuitry includes logic to perform scaling/quantization, intra prediction, inter prediction, motion compensation, and motion estimation.

The display processor 3193 may include a 2D display engine and a display controller. The display processor 3193 may also include dedicated logic capable of operating independently of the other circuitry of the GPU 3105. The display controller includes an interface for coupling to a display device (not shown), which may be a system-integrated display device (e.g., in a laptop computer), or an external display device connected via a display device connector.

Hierarchical beam tracking

Bounding volume levels are often used to improve the efficiency of performing operations on graphics primitives and other graphics objects. BVHs are hierarchical tree structures that are constructed based on a set of geometric objects. At the top of the tree structure is a root node that encapsulates all geometric objects in a given scene. The individual geometric objects are wrapped in bounding volumes that form leaf nodes of the tree. These nodes are then grouped into small sets and enclosed within a larger enclosure. These are in turn also grouped and enveloped within other larger bounding volumes in a recursive manner, resulting in a tree structure with a single bounding volume at the top of the tree represented by the root node. The bounding volume hierarchy is used to efficiently support various operations on a collection of geometric objects, such as collision detection, primitive culling, and ray traversal/intersection operations used in ray tracing.

In the ray tracing architecture, ray-primitive intersections are determined by traversing the ray through the BVH. For example, if the ray does not pass through the root node of the BVH, the ray does not intersect any of the primitives enveloped by the BVH and no further processing of the ray is required with respect to the set of primitives. If the ray passes through the first child node of the BVH instead of the second child node, then there is no need to test the ray against any primitives enveloped by the second child node. In this manner, BVH provides an efficient mechanism for testing ray-primitive intersections.

Groups of contiguous light rays, called "light beams," may be tested against BVHs, rather than individual light rays. Fig. 32 illustrates an exemplary light beam 3201 that is contoured by four different rays. Any light ray that intersects the patch 3200 defined by the four light rays is considered to be within the same beam. Although the light beam 3201 in fig. 32 is defined by a rectangular arrangement of light rays, the light beam may be defined in a variety of other ways while still conforming to the underlying principles of the invention (e.g., circular, elliptical, etc.).

Fig. 33 illustrates how the ray trace engine 3310 of the GPU 3320 implements the beam tracing techniques described herein. In particular, the ray generation circuit 3304 generates a plurality of rays on which traversal and intersection operations are to be performed. However, instead of performing traversal and intersection operations on individual rays, traversal and intersection are performed using the beam level 3307 generated by beam level construction circuit 3305. The beam level is similar to the bounding volume level (BVH). For example, fig. 34 provides an example of a main beam 3400, which main beam 3400 may be subdivided into a plurality of different components. In particular, the main beam 3400 may be divided into quadrants 3401-3404 and each quadrant itself may be divided into sub-quadrants, such as sub-quadrants A-D within quadrant 3404. The main beam can be subdivided in a number of ways. For example, the main beam may be split into two halves (rather than quadrants) and each half may be split into two halves, and so on. Regardless of how the subdivision is performed, the hierarchy is generated in a manner similar to BVH, e.g., with the root node representing the primary beam 3400, the child nodes of the first level each represented by quadrants 3401-3404, the child nodes of the second level for each of the sub-quadrants A-D, and so on.

Once beam level 3307 is constructed, traversal/intersection circuitry 3306 may perform traversal/intersection operations using beam level 3307 and BVH 3308. In particular, it may test the beam for BVHs and reject portions of the beam that do not intersect any portion of the BVH. Using the data shown in fig. 34, for example, if the sub-beams associated with sub-regions 3402 and 3403 do not intersect a BVH or a particular branch of BVH, they may be culled with respect to BVH or branch. The remaining portions 3401, 3404 may be tested for BVH by performing a depth first search or other search algorithm.

A method for ray tracing is illustrated in fig. 35. The method may be implemented within the context of the graphics processing architecture described above, but is not limited to any particular architecture.

At 3500, a primary beam is constructed comprising a plurality of rays, and at 3501, the beam is subdivided and a hierarchy data structure is generated to create a beam hierarchy. Operations 3500-3501 may be performed as a single integrated operation of constructing a beam hierarchy from a plurality of rays. At 3502, the beam level is used with the BVH to cull nodes/primitives and/or rays from the BVH (from the beam level). At 3503, ray-primitive intersections are determined for the remaining rays and primitives.

Lossy and lossless packet compression in distributed ray tracing systems

Ray tracing operations may be distributed across multiple compute nodes coupled together by a network. FIG. 36 illustrates, for example, a ray trace cluster 3600 that includes a plurality of ray trace nodes 3610-3613 that perform ray trace operations in parallel, potentially combining results on one of the nodes. In the illustrated architecture, the ray tracing node 3610-3613 is communicatively coupled to a client-side ray tracing application 3630 via a gateway.

One of the difficulties with distributed architectures is the large amount of packetized (packetised) data that must be transferred between each of the ray trace nodes 3610-3613. Both lossless compression techniques and lossy compression techniques are used to reduce the data transferred between ray tracing nodes 3610 and 3613.

To achieve lossless compression, data or commands are sent that allow the receiving node to reconstruct the results, rather than sending packets that are padded with the results of some type of operation. For example, randomly sampled local light and Ambient Occlusion (AO) operations do not necessarily require direction. Thus, the transmitting node may simply send a random seed, which is then used by the receiving node to perform random sampling. For example, if the scenes are distributed across nodes 3610-3612 to sample light 1 at points p1-p3, then only the light ID and origin need be sent to nodes 3610-3612. Each of the nodes may then independently randomly sample the light. The random seed may be generated by the receiving node. Similarly, for primary ray (primary ray) hit points, Ambient Occlusion (AO) and soft shadow sampling can be computed on nodes 3610 & 3612 without waiting for the origin of consecutive frames. Additionally, if it is known that a set of rays will go to the same point source, an instruction identifying the light source may be sent to the receiving node, which will apply the light source to the set of rays. As another example, if there are N ambient occlusion rays passing through a single point, a command may be sent to generate N samples from that point.

Various additional techniques may be applied to lossy compression. For example, a quantization factor may be employed to quantize all coordinate values associated with BVH, primitives, and rays. In addition, 32-bit floating point values for data (such as BVH nodes and primitives) may be converted to 8-bit integer values. In the exemplary implementation, the boundaries of ray packets are stored with full precision, but the individual ray points P1-P3 are transmitted as index offsets to the boundaries. Similarly, a plurality of local coordinate systems may be generated that use 8-bit integer values as local coordinates. The location of the origin of each of these local coordinate systems may be encoded using a full precision (e.g., 32-bit floating point) value, effectively connecting the global and local coordinate systems.

The following is an example of lossless compression. An example of a ray data format used internally in the ray tracing program is as follows:

instead of sending the generated raw data for each node, the data may be compressed by grouping values and creating implicit rays using applicable metadata, if possible.

Bundling and grouping light data

A flag may be used for a mask or public data with a modifier.

For example:

origin is all shared

All ray data is packed except that only a single origin is stored across all rays. Flags is set for RAYPACKET _ COMMON _ ORIGIN. When a RayPacket is unpacked when it is received, the origin is populated with a single origin value.

Sharing origin only among some rays

All ray data is packed except for rays that share the origin. For each set of unique shared origins, operators are packed that identify the operation (shared origin), store the origin, and mask which rays share information. Such an operation may be performed on any shared value among the nodes, such as a material ID, a primitive ID, an origin, a direction, a normal, and so forth.

Sending implicit rays

Typically, the ray data is derived at the receiving end using minimal meta-information for generating the ray data. A very common example is to generate a plurality of secondary rays (secondary rays) to randomly sample a region. Instead of the sender generating the secondary ray, sending it and the receiver operating on it, the sender may send a command that requires the ray to be generated with any relevant information, and the ray is generated on the receiving end. In the case where a ray needs to be first generated by the transmitter to determine which receiver to send it to, the ray is generated and a random seed may be sent to regenerate exactly the same ray.

For example, to sample hit points, where 64 shadow rays sample the area light source, all 64 rays intersect the area from the same calculation N4. A RayPacket is created with a common origin and normal. More data can be sent if one wants the receiver to color the resulting pixel contribution, but for this example let us assume if we want to return only whether the ray hits another node data. A RayOperation is created for the generate shadow ray operation and is assigned a value of the lightID to be sampled and a random number seed. When N4 receives a ray packet, it generates fully-filled ray data by filling all rays with shared origin data and setting directions based on the lightID sampled randomly with a random number of sub-random numbers to generate the same ray generated by the original sender. When returning results, only the binary results for each ray need be returned, which can be delivered through a mask on the ray.

Sending the original 64 rays in this example would have used 104 bytes by 64 rays =6656 bytes. This is also doubled to 13912 bytes if the return rays are also sent in their original form. Lossless compression is used, where only the common ray origin and normal, ray generation operation with seed and ID are sent, only 29 bytes are sent, where 8 bytes are returned for the cross-mask. This results in a need to transmit 360:1 data compression rates over the network. This does not include the overhead to process the message itself, which would need to be identified in some way, but depending on the implementation. Other operations may be performed for recalculating the ray origin and direction from the pixelID of the original ray, recalculating the pixelID based on ranges in the ray grouping, and other possible implementations for recalculation of values. Similar operations may be used for any single ray or group of rays sent, including shadows, reflections, refractions, ambient obscuration, intersections, volume intersections, coloring, bounce reflections, etc. in path tracing.

FIG. 37 illustrates additional details of two ray trace nodes 3710 and 3711 performing compression and decompression of ray trace packets. In particular, when the first ray trace engine 3730 is ready to transmit data to the second ray trace engine 3731, the ray compression circuit 3720 performs lossy and/or lossless compression of the ray trace data as described herein (e.g., converts a 32-bit value to an 8-bit value, replaces the original data with instructions to reconstruct the data, etc.). The compressed optical packet 3701 is transmitted from the network interface 3725 to the network interface 3726 over a local area network (e.g., 10Gb/s, 100Gb/s Ethernet). The light decompression circuit then decompresses the light packets at the appropriate time. For example, it may execute commands to reconstruct ray traced data (e.g., perform random sampling using a random seed for illumination operations). The ray tracing engine 3731 then uses the received data to perform ray tracing operations.

In the reverse direction, the ray compression circuit 3741 compresses the ray data, the network interface 3726 transmits the compressed ray data over the network (e.g., using the techniques described herein), the ray decompression circuit 3740 decompresses the ray data as necessary and the ray trace engine 3730 uses the data in ray trace operations. Although illustrated as separate units in FIG. 37, the ray decompression circuits 3740-3741 may be integrated into the ray trace engines 3730-3731, respectively. For example, to the extent that the compressed ray data includes commands to reconstruct the ray data, the commands may be executed by each respective ray tracing engine 3730 and 3731.

As illustrated in fig. 38, the light compression circuit 3720 may include: lossy compression circuitry 3801 for performing lossy compression techniques described herein (e.g., converting 32-bit floating point coordinates to 8-bit integer coordinates) and lossless compression circuitry 3803 for performing lossless compression techniques (e.g., transferring commands and data to allow ray decompression circuitry 3821 to reconstruct the data). The light decompression circuit 3721 includes a lossy decompression circuit 3802 and a lossless decompression circuit 3804 for performing lossless decompression.

Another exemplary method is illustrated in fig. 39. The method may be implemented on the ray tracing architecture described herein or other architectures, but is not limited to any particular architecture.

At 3900, ray data is received, which is to be transferred from the first ray tracing node to the second ray tracing node. At 3901, lossy compression circuitry performs lossy compression on the first ray trace data, and at 3902, lossless compression circuitry performs lossless compression on the second ray trace data. At 3903, the compressed ray traced data is transmitted to a second ray traced node. At 3904, the lossy/lossless decompression circuit performs lossy/lossless decompression of the ray traced data, and at 3905, the second ray tracing node performs ray tracing operations using the decompressed data.

Graphics processor with hardware accelerated hybrid ray tracing

Next, a hybrid rendering pipeline is presented that performs rasterization on graphics core 3130 and ray tracing operations on ray tracing core 3150, graphics core 3130, and/or CPU 3199 cores. For example, rasterization and depth testing may be performed on graphics kernel 3130 instead of the primary ray projection stage. Ray tracing kernel 3150 may then generate secondary rays for ray reflection, refraction, and shading. In addition, certain areas of the scene will be selected where ray tracing core 3150 will perform ray tracing operations (e.g., based on material property thresholds, such as high reflectivity levels), while other areas of the scene will be rendered with rasterization on graphics core 3130. This hybrid implementation can be used for real-time ray tracing applications where latency is a critical issue.

The ray traversal architecture described below may perform programmable shading and control of ray traversal, for example, using existing Single Instruction Multiple Data (SIMD) and/or Single Instruction Multiple Thread (SIMT) graphics processors, while accelerating critical functions, such as BVH traversal and/or interleaving, using dedicated hardware. SIMD occupancy for non-coherent paths may be improved by regrouping the spawned shaders at certain points during traversal and before shading. This is accomplished using dedicated hardware that dynamically classifies the shaders on-chip. Recursion is managed by splitting functions into continuations that are executed on return (continuations) and regrouping the continuations prior to execution in order to increase SIMD occupancy.

Programmable control of ray traversal/intersection is achieved by breaking the traversal functionality into an inner traversal that can be implemented as fixed function hardware and an outer traversal that is executed on the GPU processor and that is programmably controlled by a user-defined traversal shader. The cost of transferring traversal context between hardware and software is reduced by conservatively truncating the intra-traversal state during the transition between intra-and extra-traversal.

Programmable control of ray tracing may be expressed by the different shader types listed in table a below. There may be multiple shaders for each type. For example, each material may have a different hit shader.

Shader type	Functionality
		Master and slave	Starting the primary ray
Hit in	Bidirectional Reflectance Distribution Function (BRDF) sampling, initiating secondary rays
		Hit arbitrarily	Calculating the transmittance of alpha texture geometry
Miss	Calculating radiation from a light source
		Crossing	Cross custom shape
Go through	Instance selection and transformation
		Can call	General function

TABLE A

Recursive ray tracing may be initiated by an API function that instructs the graphics processor to start a set of main shaders or intersection circuits that may produce ray-scene intersections for the original ray. This in turn spawns other shaders, such as a walk, hit shader, or miss shader. The shader that spawns the child shader may also receive a return value from the child shader. A callable shader is a general-purpose function that can be directly spawned by another shader and can also return values to the calling shader.

Fig. 40 illustrates a graphics processing architecture that includes shader execution circuitry 4000 and fixed function circuitry 4010. The general-purpose execution hardware subsystem includes a plurality of Single Instruction Multiple Data (SIMD) and/or Single Instruction Multiple Thread (SIMT) cores/Execution Units (EU) 4001 (i.e., each core may include multiple execution units), one or more samplers 4002, and a level 1 (L1) cache 4003 or other form of local memory. The fixed function hardware subsystem 4010 includes a message unit 4004, a scheduler 4007, a ray-BVH traversal/crossbar circuit 4005, a classification circuit 4008, and a local L1 cache 4006.

In operation, the master dispatcher 4009 dispatches a set of elementary rays to a scheduler 4007, which scheduler 4007 schedules work to shaders executing on the SIMD/SIMT cores/EUs 4001. The SIMD core/EU 4001 may be the ray trace core 3150 and/or graphics core 3130 described above. Execution of the master shader yields additional work to be performed (e.g., by one or more child shaders and/or fixed function hardware). Message unit 4004 distributes work spawned by SIMD core/EU 4001 to scheduler 4007, sort circuit 4008, or ray-BVH crossbar 4005, which accesses the free stack pool as needed. If additional work is sent to scheduler 4007, it is scheduled for processing on the SIMD/SIMT core/EU 4001. Prior to scheduling, the classification circuit 4008 may classify the rays into groups or bins (bins) as described herein (e.g., grouping rays having similar characteristics). The ray-BVH crossing circuit 4005 performs ray crossing tests using the BVH bank. For example, the ray-BVH crossing circuit 4005 may compare the ray coordinates to each level of BVH to identify the ray crossed volume.

Shaders may be referenced using shader records, user-assigned structures (which include pointers to entry functions), vendor-specific metadata, and global arguments (global arguments) to shaders executed by the SIMD core/EU 4001. Each execution instance of a shader is associated with a call stack, which may be used to store arguments passed between parent and child shaders. The call stack may also store references to continuation functions that are executed on call returns.

Fig. 41 illustrates an example set of assigned stacks 4101, including a main shader stack, a hit shader stack, a walk shader stack, a continuation function stack, and a ray-BVH intersection stack (which as described may be performed by fixed function hardware 4010). New shader calls may implement a new stack from the free stack pool 4102. The stack (e.g., the stack included in the assigned stack set) may be cached in the local L1 cache 4003, 4006 to reduce latency of accesses.

There may be a limited number of call stacks each having a fixed maximum size "Sstart" allocated in a contiguous region of memory. Thus, the base address of the stack may be calculated directly from the Stack Index (SID) as base address = SID × Sstack. The stack ID may be allocated and deallocated by scheduler 4007 when scheduling work to SIMD core/EU 4001.

The master dispatcher 4009 can include a graphics processor command processor that dispatches a master shader in response to a dispatch command from a host (e.g., a CPU). Scheduler 4007 may receive these dispatch requests and launch the primary shader on the SIMD processor thread if scheduler 4007 is able to allocate a stack ID for each SIMD channel. The stack ID may be assigned from the free stack pool 4102, which free stack pool 4102 is initialized at the beginning of the dispatch command.

An executing shader may spawn a child shader by sending a spawn message to the message pass unit 4004. The command includes a stack ID associated with the shader and also includes pointers to child shader records for each active SIMD channel. The parent shader may issue the message only once for the active channel. After sending yield messages for all relevant channels, the parent shader may terminate.

Shaders executing on SIMD core/EU 4001 may also spawn fixed-function tasks, such as ray-BVH crosses, using a spawn message with shader log pointers reserved for fixed-function hardware. As mentioned, the message passing unit 4004 sends the resulting ray-BVH cross work to the fixed function ray-BVH cross circuit 4005 and the callable shaders directly to the classification circuit 4008. The sorting circuitry may derive SIMD batches with similar characteristics by shader recording pointer shader groupings. Thus, stack IDs from different parent shaders may be grouped by the classification circuit 4008 in the same batch. The classification circuit 4008 sends the grouped batches to a scheduler 4007, which scheduler 4007 accesses shader records from a graphics memory 2511 or a Last Level Cache (LLC) 4020 and starts shaders on processor threads.

The continuation may be considered a callable shader and may also be referenced by a shader record. When the child shader is spawned and returns a value to the parent shader, a pointer to the continuation shader may be pushed onto the call stack 4101. When the child shader returns, the continuation shader record may then be popped off the call stack 4101 and the continuation shader may be spawned. Alternatively, the produced continuation may pass through a classification unit similar to the callable shader and be launched on the processor thread.

As illustrated in fig. 42, the classification circuit 4008 creates SIMD batches for shading by grouping spawned tasks through the shader record pointers 4201A, 4201B, 4201 n. The stack IDs or context IDs in the sorted batches may be grouped according to different assignments and different input SIMD lanes. The grouping circuit 4210 may perform the classification using a Content Addressable Memory (CAM) structure 4201, the content addressable memory structure 4201 including a plurality of entries, each entry identified with a tag 4201. As mentioned, the tags 4201 may be corresponding shader record pointers 4201A, 4201B, 4201 n. The CAM structure 4201 may store a limited number of tags (e.g., 32, 64, 128, etc.) that are each associated with an incomplete SIMD lot corresponding to a shader record pointer.

For incoming yield commands, each SIMD channel has a corresponding stack ID (shown as 16 context IDs 0-15 in each CAM entry) and shader record pointers 4201A-B, …, n (acting as tag values). The grouping circuit 4210 may compare the shader record pointer for each channel to the tag 4201 in the CAM structure 4201 to find a matching batch. If a matching batch is found, the stack ID/context ID may be added to the batch. Otherwise a new entry with a new shader record pointer tag may be created, possibly evicting older entries with incomplete batches.

The execution shader may deallocate the call stack by sending a deallocation message to the message unit when the call stack is empty. The deallocation message is relayed to the scheduler, which returns the stack ID/context ID to the free pool for the active SIMD lane.

Hybrid approaches for ray traversal operations using a combination of fixed-function ray traversal and software ray traversal are presented. Thus, it provides flexibility for software traversal while maintaining efficiency for fixed function traversal. FIG. 43 illustrates an acceleration structure that may be used for hybrid traversal, which is a two-level tree with a single top-level BVH 4300 and several bottom-level BVHs 4301 and 4302. Graphic elements are shown on the right to indicate an inner traversal path 4303, an outer traversal path 4304, a traversal node 4305, a leaf node 4306 having a triangle, and a leaf node 4307 having custom primitives.

A leaf node 4306 with triangles in the top level BVH 4300 may reference triangles, cross shader records for custom primitives, or traverse shader records. The leaf node 4306 with triangles for the bottom level BVH 4301-4302 may reference only triangles and cross shader records for custom primitives. The type of reference is encoded within the leaf node 4306. The inner traversal 4303 refers to the traversal within each BVH 4300-4302. The inner traversal operation includes the computation of ray-BVH intersections and the traversal across BVH structures 4300-4302 is referred to as the outer traversal. The inner traversal operation can be implemented efficiently in fixed-function hardware, while the outer traversal operation can be performed with acceptable performance with a programmable shader. Thus, the inner walk operation may be performed using fixed function circuitry 4010 and the outer walk operation may be performed using shader execution circuitry 4000, the shader circuitry 4000 comprising SIMD/SIMT cores/EUs 4001 for executing programmable shaders.

Note that for simplicity, SIMD/SIMT core/EU 4001 is sometimes referred to herein simply as a "core", "SIMD core", "EU", or "SIMD processor". Similarly, ray-BVH traversal/intersection circuit 4005 is sometimes referred to simply as a "traversal unit," traversal/intersection unit, "or" traversal/intersection circuit. When alternative terminology is used, the particular designation used to designate the corresponding circuit/logic does not alter the underlying function performed by the circuit/logic, as described herein.

Furthermore, although illustrated as a single component in fig. 40 for purposes of explanation, traverse/crossbar circuit 4005 may comprise different traverse cells and separate crossbar cells, each of which may be implemented in circuitry and/or logic as described herein.

When a ray intersects a traversal node during an inner traversal, a traversal shader can be spawned. The classification circuit 4008 may group these shaders by shader log pointers 4201A-B, …, n to create a SIMD batch, which is initiated by the scheduler 4007 for SIMD execution on the graphics SIMD core/EU 4001. The traversal shader can modify the traversal in several ways, enabling a wide range of applications. For example, the traversal shader may select BVHs or transform rays at a coarser level of detail (LOD) to implement rigid body transformation. The traversal shader may then spawn an intra-traversal for the selected BVH.

The inner traversal computes the ray-BVH intersection by traversing the BVH and computing the ray box and ray triangle intersection. The intra traversal is spawned in the same manner as the shader by sending a message to the message passing circuitry 4004, which message passing circuitry 4004 relays the corresponding spawn message to the ray-BVH intersection circuit 4005, which computes the ray-BVH intersection.

The stack traversed within may be stored locally in fixed function circuitry 4010 (e.g., within L1 cache 4006). When a ray intersects a leaf node corresponding to either a traversal shader or a intersection shader, the inner traversal may be terminated and the inner stack truncated. The truncated stack, along with pointers to rays and BVHs, may be written to memory at the location specified by the calling shader and then the corresponding traversal shader or intersection shader may be spawned. If a ray intersects any triangles during an inner traversal, the corresponding hit information may be provided as input arguments to these shaders, as shown in the code below. These spawned shaders may be grouped by the classification circuit 4008 to create a SIMD batch for execution.

Truncating the inner traversal stack reduces the cost of overflowing the inner traversal stack to memory. Can be applied toRestart Trail for Stackless BVH TraversalThe method described in High Performance Graphics (2010) page 107-111 truncates the stack into a small number of entries at the top of the stack, a 42-bit restart trail and a 6-bit depth value. The restart trace indicates branches that have been taken inside the BVH and the depth value indicates the traversal depth corresponding to the last stack entry. This is sufficient information to traverse within a later time recovery.

The inner traversal is completed when the inner stack is empty and there are no more BVH nodes to test. In this case, an external stack handler is spawned that pops the top of the external stack and resumes traversal if the external stack is not empty.

The outer traversal may execute a main traversal state machine and may be implemented in program code executed by the shader execution circuitry 4000. It may produce an intra-traversal query under the following conditions: (1) when a new ray is produced by the hit shader or the main shader; (2) when the traversal shader selects a BVH for traversal; and (3) when the external stack handler resumes the internal traversal of the BVH.

As illustrated in FIG. 44, prior to in-yield traversal, space is allocated on call stack 4405 for fixed function circuitry 4010 to store a truncated internal stack 4410. The offsets 4403-4404 to the top of the call stack and the internal stack are maintained in the walk state 4400, which walk state 4400 is also stored in memory 2511. Traversal state 4400 also includes rays in world space 4401 and object space 4402, as well as hit information for the nearest intersected primitives.

The traversal shader, the intersection shader, and the external stack handler are all spawned by the ray-BVH intersection circuit 4005. The walk shader allocates on the call stack 4405 before initiating a new intra walk for the second level BVH. The external stack handler is the shader responsible for updating the hit information and recovering any pending in-flight traversal tasks. The external stack handler is also responsible for yielding either a hit or miss shader when the traversal is complete. Traversal is completed when there are no pending traversal queries to yield. When traversal is complete and an intersection is found, yield hits the shader; otherwise a miss shader is produced.

Although the hybrid traversal scheme described above uses two levels of BVH levels, any number of BVH levels with corresponding changes in the outer traversal implementation may be implemented.

Additionally, although fixed function circuitry 4010 is described above for performing ray-BVH intersection, other system components may also be implemented in the fixed function circuitry. For example, the external stack handler described above may be an internal (not user visible) shader, which may potentially be implemented in the fixed-function BVH traversal/intersection circuit 4005. This implementation may be used to reduce the round-trip between the fixed-function crossbar hardware 4005 and the processor and the number of shader stages allocated.

The examples described herein enable programmable shading and ray traversal control using user-defined functions that can be performed with greater SIMD efficiency on existing and future GPU processors. Programmable control of ray traversal implements several important features, such as procedural instantiation (procedural instantiation), random level of detail selection, custom primitive intersection, and lazy BVH update (lazy BVH update).

A programmable Multiple Instruction Multiple Data (MIMD) ray tracing architecture that supports speculative execution of hit shaders and cross shaders is also provided. In particular, the architecture focuses on reducing the scheduling and communication overhead between the programmable SIMD/SIMT core/execution unit 4001 described above with respect to fig. 40 and the fixed-function MIMD walk/crossbar unit 4005 in the hybrid ray-tracing architecture. Multiple speculative execution schemes of hit shaders and cross shaders are described below, which can be dispatched in a single batch from the traversal hardware, avoiding several traversal and shading roundtrips. Dedicated circuitry may be used to implement these techniques.

Embodiments of the present invention are particularly beneficial in use cases where it is desirable to execute multiple hit shaders or cross shaders (which would incur significant overhead when implemented without dedicated hardware support) according to a ray traversal query. These include, but are not limited to, the most recent k-hit query (hit shaders are launched for the k closest intersections) and multiple programmable intersection shaders.

The techniques described herein may be implemented as an extension to the architecture illustrated in fig. 40 (and described with respect to fig. 40-44). In particular, the present embodiment of the invention builds on this architecture with enhancements to improve the performance of the use case mentioned above.

The performance limitation of the hybrid ray trace traversal architecture is the overhead of initiating a traversal query from an execution unit and the overhead of invoking a programmable shader from ray trace hardware. This overhead generates "execution roundtrips" between the programmable core 4001 and the traverse/interleave unit 4005 when multiple hit shaders or interleave shaders are invoked during traversal of the same ray. This also places additional pressure on the classification unit 4008 that needs to extract SIMD/SIMT coherence from individual shader calls.

Several aspects of ray tracing require the ability to be programmable, which can be expressed by the different shader types listed in table a above (i.e., main, hit, any hit, miss, cross, traverse, and callable). There may be multiple shaders for each type. For example, each material may have a different hit shader. In Microsoft at present ^®Some of these shader types are defined in the ray tracing API.

As a brief review, recursive ray tracing is initiated by an API function that instructs the GPU to launch a set of master shaders that are capable of producing ray-scene intersections (implemented in hardware and/or software) for the original ray. This in turn may spawn other shaders (such as traversal, hit, or miss shaders). The shader that spawns the child shader can also receive a return value from the shader. A callable shader is a general-purpose function that can be directly spawned by another shader and can also return values to the calling shader.

Ray traversal computes ray-scene intersections by traversing and intersecting nodes in a bounding volume level (BVH). Recent research has shown that techniques better suited to fixed function hardware (such as reduced precision arithmetic, BVH compression, per ray state machines, dedicated intersection pipelines, and custom caches) can be used to improve the efficiency of computing ray-scene intersections by more than an order of magnitude.

The architecture shown in FIG. 40 includes a system in which an array of SIMD/SIMT cores/execution units 4001 interact with a fixed function ray trace/crossbar unit 4005 to perform programmable ray tracing. Programmable shaders are mapped onto SIMD/SIMT threads on execution unit/core 4001, where SIMD/SIMT utilization, execution, and data coherency are critical for optimal performance. Ray queries typically break coherence for a variety of reasons, such as:

Traversal divergence: the duration of the BVH traversal varies highly.

Asynchronous ray processing tends to occur between rays.

Execution divergence: light produced from different passes of the same SIMD/SIMT thread may result in different shader calls.

Data access divergenceProperty of (2): for example, rays hitting different surfaces sample different BVH nodes and primitives, and shaders access different textures. Various other scenarios may cause data access divergences.

The SIMD/SIMT core/execution unit 4001 may be a variation of the cores/execution units described herein, including graphics core(s) 415A-415B, shader cores 1355A-N, graphics core 3130, graphics execution unit 608, execution units 852A-B, or any other core/execution unit described herein. The SIMD/SIMT core/execution unit 4001 may be used in place of the graphics core(s) 415A-415B, shader cores 1355A-N, graphics core 3130, graphics execution unit 608, execution units 852A-B, or any other core/execution unit described herein. Accordingly, disclosure of any feature in combination with graphics core(s) 415A-415B, shader cores 1355A-N, graphics core 3130, graphics execution unit 608, execution units 852A-B, or any other core/execution unit described herein also discloses a corresponding combination with the SIMD/SIMT core/execution unit 4001 of FIG. 40, but is not so limited. .

The fixed function ray tracing/crossing unit 4005 can overcome the first two challenges by processing each ray individually and out of order. However, this destroys the SIMD/SIMT groups. The classification unit 4008 is therefore responsible for forming new coherent SIMD/SIMT groups to be called up again by the shaders assigned to the execution units.

It is easy to see the benefits of such an architecture compared to a pure software-based ray-tracing implementation directly on a SIMD/SIMT processor. However, there is overhead associated with message passing between SIMD/SIMT core/execution unit 4001 (sometimes referred to herein simply as SIMD/SIMT processor or core/EU) and MIMD walk/crossbar unit 4005. Furthermore, classification unit 4008 may not extract ideal SIMD/SIMT utilization from the non-coherent shader calls.

Use cases may be identified in which shader calls may be particularly frequent during traversal. Enhancements to the hybrid MIMD ray trace processor are described which serve to significantly reduce the overhead of communication between the core/EU 4001 and the walk/cross unit 4005. This may be particularly beneficial when finding the k-nearest intersection and implementing a programmable intersection shader. It is noted, however, that the techniques described herein are not limited to any particular processing scenario.

An overview of the high-level cost of ray trace context switching between the core/EU 4001 and the fixed-function traversal/intersection unit 4005 is provided below. Most of the execution overhead is caused by these two context switches whenever a shader invocation is necessary during a single ray traversal.

Each SIMD/SIMT lane that emits a ray generates a yield message to traverse/crossbar unit 4005 associated with the BVH for traversal. Data (ray traversal context) is relayed via yield messages and (cached) memory to traversal/intersection unit 4005. When the traverse/interleave unit 4005 is ready to assign a new hardware thread to a yield message, the traverse/interleave unit 4005 loads the traversal state and performs traversal on the BVH. There is also a setup cost that needs to be performed before the first traversal step on the BVH.

FIG. 45 illustrates the operational flow of a programmable ray tracing pipeline. The shading elements, including traversal 4502 and intersection 4503, may be implemented in fixed function circuitry, while the remaining elements may be implemented with programmable cores/execution units.

The original ray shader 4501 sends work to the traversal circuit at 4502, which traverses the current ray(s) through the BVH (or other acceleration structure). Upon reaching the leaf node, the traversal circuit invokes the intersection circuit at 4503, which, when identifying the ray-triangle intersection, invokes any hit shader at 4504 (which, as indicated, may provide the result back to the traversal circuit).

Alternatively, the traversal may terminate before reaching the leaf node and the closest hit shader called at 4507 (if a hit is recorded) or the miss shader at 4506 (in the case of a miss).

As indicated at 4505, if the traversal circuit reaches a custom primitive leaf node, then the cross shader may be invoked. The custom primitives may be any non-triangular primitive such as a polygon or polyhedron (e.g., tetrahedron, voxel, hexahedron, wedge, pyramid, or other "unstructured" volume). The cross shader 4505 identifies any cross between rays and custom primitives to any hit shader 4504, which implements any hit processing.

When the hardware walk 4502 reaches the programmable phase, the walk/cross unit 4505 can generate shader dispatch messages to the relevant shaders 4505 and 4507, which relevant shaders 4505 and 4507 correspond to a single SIMD lane of the execution unit(s) used to execute the shaders. Because the dispatch occurs in any order of the ray, and the dispatch is divergent among the called programs, the sort unit 4008 can accumulate multiple dispatch calls to extract coherent SIMD batches. The updated traversal state and optional shader arguments may be written into memory 2511 by traversal/interleave unit 4005.

In the k-nearest intersection problem, the nearest hit shader 4507 executes for the first k intersection. In a conventional manner, this would mean ending the ray traversal when the closest intersection is found, invoking the hit shader, and spawning a new ray from the hit shader to find the next closest intersection (with the ray origin offset, so the same intersection would not happen again). It is easy to see that this implementation would require k ray productions for a single ray. Another implementation operates with an arbitrary hit shader 4504, which arbitrary hit shader 4504 is invoked for all intersections and maintains a global list of recent intersections, using an insert sort operation. The main problem with this approach is that there is no upper bound for any hit shader calls.

As mentioned, the cross shader 4505 can be invoked on non-triangular (custom) primitives. Depending on the results of the intersection tests and the traversal state (pending node and primitive intersection), traversal of the same ray may continue after execution of the intersection shader 4505. Thus, finding the closest hit may require several round trips to the execution unit.

Focus can also be placed on reducing SIMD-MIMD context switching for the cross shader 4505 and the hit shaders 4504, 4507 by changing to traverse hardware and shader scheduling models. First, the ray traversal circuit 4005 delays shader calls by accumulating multiple potential calls and dispatching them in larger batches. In addition, certain calls that prove unnecessary may be culled at this stage. Furthermore, shader scheduler 4007 may aggregate multiple shader calls from the same traversal context into a single SIMD batch, which results in a single thread yield message. In one exemplary implementation, traversal hardware 4005 suspends traversing threads and waits for results of multiple shader calls. This mode of operation is referred to herein as "speculative" shader execution because it allows dispatching multiple shaders, some of which may not be called when sequential calls are used.

FIG. 46A illustrates an example where a traversal operation encounters multiple custom primitives 4650 in a subtree, and FIG. 46B illustrates how this can be solved using three cross-dispatch cycles C1-C3. In particular, the scheduler 4007 may require three cycles to submit work to the SIMD processor 4001, and the traversal circuit 4005 requires three cycles to provide results to the classification unit 4008. The traversal state 4601 required by the traversal circuit 4005 may be stored in a memory, such as a local cache (e.g., an L1 cache and/or an L2 cache).

A. Delayed ray tracing shader fetch

The manner in which the management hardware traverses state 4601 may also be modified to allow multiple potential intersection or hit calls to accumulate in the list. At a given time during the traversal, each input in the list may be used to generate a shader call. For example, k-nearest intersections can be accumulated in traversal state 4601 on traversal hardware 4005 and/or in memory, and if the traversal is complete, the hit shader can be invoked for each element. For hit shaders, multiple potential intersections may be accumulated for subtrees in the BVH.

For the most recent-k use case, the benefit of this approach is that instead of k-1 round trips and k-1 new ray yield messages to the SIMD core/EU 4001, all hit shaders are invoked from the same traversal thread during a single traversal operation on traversal circuit 4005. The challenge for potential implementation is to ensure that the order of execution of the hit shaders is not straightforward (standard "round-trip" methods ensure that the hit shader closest to the intersection is executed first, etc.). This can be addressed by synchronization or relaxation of ordering of the hit shaders.

For the cross shader use case, the traversal circuit 4005 does not know in advance whether a given shader will return a positive cross test. However, it is possible to speculatively execute multiple cross shaders and merge it into a global recent hit if at least one returns a positive hit result. Specific implementations need to find an optimal number of delayed cross-tests to reduce the number of dispatched calls, but avoid invoking too many redundant cross-shaders.

B. Aggregating shader fetches from traversal circuit

When multiple shaders are dispatched from the same ray yield on the traversal circuit 4005, a branch may be created in the flow of the ray traversal algorithm. This may be problematic for the cross shader, as the remainder of the BVH traversal depends on the results of all dispatched cross tests. This means that synchronous operation is necessary to wait for the results of shader calls, which can be challenging on asynchronous hardware.

Two points of merging the results of shader calls may be: a SIMD processor 4001 and a traverse circuit 4005. With respect to SIMD processor 4001, multiple shaders can synchronize and aggregate their results using a standard programming model. One relatively simple way to do this is to use a global atom and aggregate the results in a shared data structure in memory where the interleaved results of multiple shaders can be stored. The last shader can then parse the data structure and call back to the traversal circuit 4005 to continue the traversal.

A more efficient method of limiting execution of multiple shader calls to the path of the same SIMD thread on SIMD processor 4001 may also be achieved. Then, the cross-tests are reduced locally using SIMD/SIMT reduction operations (rather than relying on global atoms). This implementation may rely on new circuitry within the classification unit 4008 to leave a small batch of shader calls in the same SIMD batch.

Execution of traversal threads on traversal circuit 4005 may also be suspended. Using a conventional execution model, when a shader is dispatched during traversal, the traversal thread is terminated, and the ray traversal state is saved to memory to allow other ray yield commands to be executed while the execution unit 4001 processes the shader. If the traversal thread is merely paused, the traversal state need not be stored and each shader result can be individually waited for. The implementation may include circuitry to avoid deadlocks and provide adequate hardware utilization.

Fig. 47-48 illustrate examples of delay models that utilize three shaders 4701 to invoke a single shader invocation on the SIMD core/execution unit 4001. All cross tests are evaluated within the same SIMD/SIMT group when saved. As a result, the nearest intersection can also be computed on the programmable core/execution unit 4001.

As mentioned, all or a portion of the shader aggregation and/or delay may be performed by the traverse/crossbar circuit 4005 and/or the core/EU scheduler 4007. Fig. 47 illustrates how the shader delay/aggregator circuitry 4706 within the scheduler 4007 can delay the scheduling of shaders associated with a particular SIMD/SIMT thread/lane until a specified trigger event has occurred. Upon detection of a triggering event, scheduler 4007 dispatches multiple aggregated shaders in a single SIMD/SIMT batch to core/EU 4001.

Fig. 48 illustrates how a shader delay/aggregator circuit 4805 within the traverse/crossbar circuit 4005 can delay the scheduling of shaders associated with a particular SIMD thread/lane until a specified trigger event has occurred. Upon detecting a trigger event, the traverse/crossbar circuit 4005 submits the aggregated shaders to the sorting units 4008 in a single SIMD/SIMT batch.

Note, however, that the shader delay and aggregation techniques may be implemented within various other components (such as classification unit 4008) or may be distributed across multiple components. For example, the traverse/crossbar circuit 4005 may execute a first set of shader aggregate operations and the scheduler 4007 may execute a second set of shader aggregate operations to ensure shaders of SIMD threads are efficiently scheduled on the cores/EUs 4001.

The "trigger event" that causes the aggregate shader to be assigned to a core/EU may be a processing event, such as a minimum latency associated with a particular thread or a particular number of accumulated shaders. Alternatively or additionally, the trigger event may be a temporal event, such as a duration or a particular number of processor cycles since the delay of the first shader. Other variables, such as the current workload on the core/EU 4001 and the traverse/crossbar unit 4005, may also be evaluated by the scheduler 4007 to determine when to dispatch a shader's SIMD/SIMT batch.

Different embodiments of the invention may be implemented using different combinations of the above approaches, depending on the requirements of the particular system architecture and application used.

Ray tracing instruction

The ray tracing instructions described below are included in an Instruction Set Architecture (ISA) supported by CPU 3199 and/or GPU 3105. If executed by the CPU, Single Instruction Multiple Data (SIMD) instructions may utilize vector/packed source and destination registers to perform the described operations, and may be decoded and executed by the CPU core. If executed by GPU 3105, the instructions may be executed by graphics core 3130. For example, any of the Execution Units (EUs) 4001 described above may execute instructions. Alternatively or in addition, the instructions may be executed by execution circuitry on ray trace core 3150 and/or tensor core 3140.

FIG. 49 illustrates an architecture for executing ray traced instructions described below. The illustrated architecture may be integrated within one or more of the cores 3130, 3140, 3150 described above (see, e.g., fig. 31 and associated text), which may be included in different processor architectures.

In operation, instruction fetch unit 4903 fetches a ray trace instruction 4900 from memory 3198, and decoder 4995 decodes the instruction. In one implementation, the decoder 4995 decodes instructions to generate executable operations (e.g., micro-operations or uops in a micro-coded core). Alternatively, some or all of the ray trace instructions 4900 may be executed without decoding, and as such the decoder 4904 is not required.

In either implementation, scheduler/dispatcher 4905 schedules and dispatches instructions (or operations) across a set of Functional Units (FUs) 4910 and 4912. The illustrated implementation includes: a vector FU 4910 for executing Single Instruction Multiple Data (SIMD) instructions that operate on multiple packed data elements stored in the vector registers 4915 simultaneously, and a scalar FU 4911 for operating on scalar values stored in one or more scalar registers 4916. The optional ray tracing FU 4912 may operate on packed data values stored in vector registers 4915 and/or scalar values stored in scalar registers 4916. In implementations without a special FU 4912, the vector FU 4910 and possibly the scalar FU 4911 may perform the ray tracing instruction described below.

The various FUs 4910 and 4912 access ray traced data 4902 (e.g., traversal/intersection data) needed to execute ray traced instructions 4900 from vector registers 4915, scalar registers 4916, and/or local cache subsystems 4908 (e.g., L1 cache). FU 4910 and 4912 may also perform accesses to memory 3198 via load and store operations, and cache subsystem 4908 may operate independently to cache data locally.

While ray tracing instructions may be used to improve the performance of ray traversal/intersection and BVH construction, ray tracing instructions may also be applicable to other areas, such as high-performance computing (HPC) and general-purpose GPU (GPGPU) implementations.

In the following description, the term double-word is sometimes abbreviateddwAnd, unsigned bytes are abbreviatedub. Additionally, the source and destination registers mentioned below (e.g., src0, src1, dest, etc.) may refer to vector registers 4915, or in some cases, a combination of vector registers 4915 and scalar registers 4916. In general, if the source or destination value used by the instruction includes packed data elements (e.g., where the source or destination stores N data elements), the vector register 4915 is used. Other values may use scalar registers 4916 or vector registers 491 5。

Dequantization

One example of a dequantization instruction "dequantizes" a previously quantized value. For example, in a ray tracing implementation, certain BVH subtrees may be quantized to reduce storage and bandwidth requirements. The dequantization instruction may take the form of dequantization dest src0 src1 src2, where source register src0 stores N unsigned bytes, source register src1 stores 1 unsigned byte, source register src2 stores 1 floating-point value, and destination register dest stores N floating-point values. All of these registers may be vector registers 4915. Alternatively, src0 and dest may be vector registers 4915, and src1 and src2 may be scalar registers 4916.

The following code sequence defines one particular implementation of the dequantization instruction:

in this example, ldexp is the double precision floating point value to a specified integer power of two (i.e., ldexp (x, exp) = x 2)^exp) Multiplication. In the above code, if the execution mask value (execMask i) associated with the current SIMD data element is present]) Set to 1, then the SIMD data element at position i in src0 is converted to a floating point value and multiplied by an integer power of the value in src1 (2)^{src1 value}) And this value is added to the corresponding SIMD data element in src 2.

Selective minimum or maximum value

The selective minimum or maximum instruction may perform a minimum or maximum operation (i.e., return a minimum or maximum of a set of values) for each lane as indicated by the bits in the bitmask. The bit mask may utilize vector registers 4915, scalar registers 4916, or a separate set of mask registers (not shown). The following code sequence defines one particular implementation of the min/max instruction: sel _ min _ max dest src0 src1 src2, where src0 stores N doublewords, src1 stores N doublewords, src2 stores one doubleword, and the destination register stores N doublewords.

The following code sequence defines one particular implementation of the selective min/max instruction:

in this example, the value of (1 < < i) & src2 (a 1 left-shifted by i ANDed with src 2) is used to select the minimum value of the ith data element in src0 and src1 or the maximum value of the ith data element in src0 and src 1. This operation is only performed for the ith data element when the execution mask value (execMask [ i ]) associated with the current SIMD data element is set to 1.

Shuffle index instruction

The shuffle index instruction is capable of copying any set of input ways to an output way. For a SIMD width of 32, the instruction can be executed at a lower throughput. The instructions take the form of: shuffle _ index dest src0 src1 < optional flag >, where src0 stores N doublewords, src1 stores N unsigned bytes (i.e., index values), and dest stores N doublewords.

One particular implementation of the shuffle index instruction is defined by the following code sequence:

in the code above, the index in src1 identifies the current way. If the ith value in the execution mask is set to 1, a check is performed to ensure that the source path is in the range of 0 to SIMD width. If so, the flag is set to (src lanemomod) and the data element i of the destination is set equal to the data element i of src 0. If the lane is in range (i.e., valid), the index value from src1 (src lane 0) is used as the index into src0 (dst [ i ] = src0[ src lane ]).

Immediate value shuffle Up/Dn/XOR instructions

The immediate value shuffle instruction may shuffle the input data elements/ways based on an immediate value of the instruction. The immediate value may specify shifting the input lane by 1, 2, 4, 8, or 16 positions based on the value of the immediate value. Alternatively, additional scalar source registers can be designated as fill values. When the source way index is invalid, the padding value (if provided) is stored to the data element location in the destination. If no padding values are provided, the data element positions are all set to zero.

The flag register may be used as a source mask. If the flag bit of the source way is set to 1, the source way may be marked as invalid and the instruction may proceed.

The following are examples of different implementations of an immediate value shuffle instruction:

shuffle _ < up/dn/xor > _<1/2/4/8/16> dest src0 < optional src1> < optional flag >

shuffle _ < up/dn/xor > _<1/2/4/8/16> dest src0 < optional src1> < optional flag >

In this implementation, src0 stores N doublewords, src1 stores one doubleword for the pad value (if present), and dest stores N doublewords that include the result.

The following code sequence defines one particular implementation of the immediate value shuffle instruction:

here, the input data element/way is shifted by 1, 2, 4, 8 or 16 positions based on the value of the immediate value. Register src1 is an additional scalar source register that serves as a fill value for the data element position stored into the destination when the source lane index is invalid. If no padding value is provided and the source lane index is invalid, the data element position in the destination is set to 0. A FLAG register (FLAG) is used as the source mask. As described above, if the flag bit of the source way is set to 1, the source way is marked invalid and the instruction proceeds.

Indirect shuffle Up/Dn/XOR instruction

The indirect shuffle instruction has a source operand (src 1) that controls the mapping from the source lane to the destination lane. The indirect shuffle instruction may take the form of:

shuffle _ < up/dn/xor > dest src0 src1 < optional flag >

Where src0 stores N doublewords, src1 stores 1 doubleword, and dest stores N doublewords.

The following code sequence defines one particular implementation of the immediate value shuffle instruction:

thus, the indirect shuffle instruction operates in a similar manner to the immediate value shuffle instruction described above, but the mapping of the source lane to the destination lane is controlled by the source register src1 rather than the immediate value.

Cross-lane min/max instruction

Cross-lane min/max instructions may be supported for floating point and integer data types. The cross lane minimum instruction may take the form of lane _ min dest src0 and the cross lane maximum instruction may take the form of lane _ max dest src0, where src0 stores N doublewords and dest stores 1 doubleword.

For example, the following code sequence defines one particular implementation of the cross-path minimum:

in this example, the doubleword value in data element position i of the source register is compared to the data element in the destination register, and the minimum of these two values is copied to the destination register. The cross-lane max instruction operates in substantially the same manner, the only difference being that the maximum of the data element and destination value in location i is selected.

Cross-lane min/max index instruction

The cross-lane minimum index instruction may take the form of lane _ min _ index dest src0 and the cross-lane maximum index instruction may take the form of lane _ max _ index dest src0, where src0 stores N doublewords and dest stores 1 doubleword.

For example, the following code sequence defines one particular implementation of the cross-lane minimum index instruction:

in this example, the destination index is incremented from 0 to SIMD width across the destination register. If the execution mask bit is set, the data element at position i in the source register is copied to the temporary storage location (tmp) and the destination index is set to data element position i.

Ordering network (ordering) instructions across paths

The cross-lane ordering network instruction may use a (stable) ordering network of width N to order all N input elements in ascending (sortnet _ min) or in descending (sortnet _ max) order. The min/max versions of the instruction may take the form of sortnet _ min dest src0 and sortnet _ maxddest src0, respectively. In one implementation, src0 and dest store N doublewords. Min/max sorting is performed on N dwords of src0, and the ascending element (for the min) or descending element (for the max) is stored in its respective sorted order in dest. One example of a code sequence that defines an instruction is: dst = apply _ N _ wide _ debugging _ network _ min/max (src 0).

Cross-lane ordering network index instructions

The cross-lane ordering network index instruction may use a (stable) ordering network of width N to order all N input elements, but return indexes that change in sequence in ascending order (sortnet _ min) or in descending order (sortnet _ max). The min/max version of the instruction may take the form of sortnet _ min _ index dest src0 and sortnet _ max _ index dest src0, where src0 and dest each store N doublewords. One example of a code sequence defining an instruction is dst = apply _ N _ wide _ debugging _ network _ min/max _ index (src 0).

A method for executing any of the above instructions is illustrated in fig. 50. The method may be implemented on the particular processor architecture described above, but is not limited to any particular processor or system architecture.

At 5001, instructions of the main graphics thread are executed on the processor core. This may include, for example, any of the cores described above (e.g., graphics core 3130). When it is determined at 5002 that the ray tracing job is reached within the main graphics thread, the ray tracing instruction is offloaded to ray trace execution circuitry, which may take the form of a Functional Unit (FU) such as described above with respect to fig. 49 or may be located in a dedicated ray trace core 3150 as described with respect to fig. 31.

At 5003, the ray trace instruction is decoded and fetched from memory, and at 5005, the instruction is decoded into executable operations (e.g., in embodiments requiring a decoder). At 5004, the ray traced instruction is scheduled and dispatched for execution by the ray tracing circuitry. At 5005, the ray trace instruction is executed by the ray trace circuitry. For example, instructions may be dispatched and executed on an FU described above (e.g., vector FU 4910, ray trace FU4912, etc.) and/or graphics core 3130 or ray trace core 3150.

Upon completion of execution for the ray traced instruction, the result is stored (e.g., stored back to memory 3198) at 5006 and the main graphics thread is notified at 5007. At 5008, the ray trace results are processed within the context of the main thread (e.g., read from memory and integrated into the graphics rendering results).

In embodiments, the term "engine" or "module" or "logic" may refer to, be part of, or include the following: an Application Specific Integrated Circuit (ASIC), an electronic circuit, a processor (shared, dedicated, or group) and/or memory (shared, dedicated, or group) that executes one or more software or firmware programs, a combinational logic circuit, and/or other suitable components that provide the described functionality. In embodiments, the engine, module, or logic may be implemented in firmware, hardware, software, or any combination of firmware, hardware, and software.

Apparatus and method for asynchronous ray tracing

Embodiments of the present invention include a combination of fixed function acceleration circuitry and general purpose processing circuitry to perform ray tracing. For example, certain operations related to ray traversal and intersection testing of the bounding volume level (BVH) may be performed by fixed function acceleration circuitry, while multiple execution circuits execute various forms of ray tracing shaders (e.g., any hit shader, intersection shader, miss shader, etc.). One embodiment includes a dual high bandwidth store containing multiple entries for storing rays and a corresponding dual stack for storing BVH nodes. In this embodiment, the traversal circuit alternates between dual ray banks and stacks to process rays at each clock cycle. Additionally, one embodiment includes priority selection circuitry/logic that distinguishes between internal nodes, non-internal nodes, and primitives and uses this information to intelligently prioritize BVH nodes and the processing of primitives encompassed by BVH nodes.

One particular embodiment uses a short stack to store a limited number of BVH nodes during the traversal operation to reduce the high speed memory required for the traversal. This embodiment includes stack management circuitry/logic to efficiently push and pop entries to/from the short stack to ensure that the required BVH node is available. In addition, traversal operations are tracked by performing updates to the tracking data structure. When the traversal circuit/logic pauses, it can consult the trace data structure to start the traversal operation at the same location within the BVH it left off. And performs the trace data maintained in the data structure trace so that the traversal circuitry/logic can restart.

Fig. 51 illustrates one embodiment, which includes shader execution circuitry 4000 to execute shader program code and process associated ray trace data 4902 (e.g., BVH node data and ray data), ray trace acceleration circuitry 5110 to perform traversal and intersection operations, and memory 3198 to store program code and associated data processed by RT acceleration circuitry 5110 and shader execution circuitry 4000.

In one embodiment, the shader execution circuitry 4000 includes multiple core/execution units 4001 that execute shader program code to perform various forms of data parallel operations. For example, in one embodiment, the core/execution unit 4001 may execute a single instruction across multiple channels, where each instance of the instruction operates on data stored in a different channel. For example, in a SIMT implementation, each instance of an instruction is associated with a different thread. During execution, the L1 cache stores certain ray traced data (e.g., recently or frequently accessed data) for efficient access.

A set of original rays may be assigned to scheduler 4007, which schedules work to shaders executed by the cores/EUs 4001. The core/EU 4001 may be a ray trace core 3150, a graphics core 3130, a CPU core 3199, or other type of circuitry that may execute shader program code. The one or more original ray shaders 5101 process the original rays and yield additional work to be performed by the ray trace acceleration circuit 5110 and/or the core/EU 4001 (e.g., to be performed by one or more sub-shaders). New work produced by the primitive ray shader 5101 or other shaders (which are executed by the cores/EUs 4001) may be distributed to the classification circuit 4008, which classifies rays into groups or bins as described herein (e.g., groups rays having similar characteristics). Scheduler 4007 then schedules new work on core/EU 4001.

Other shaders that may be executed include any hit shader 4514 and closest hit shader 4507 that process hit results (e.g., identify any hits or closest hits, respectively, for a given ray), as described above. Miss shader 4506 handles ray misses (e.g., where rays do not intersect nodes/primitives). As described above, various shaders may be referenced using shader records, which may include one or more pointers, vendor-specific metadata, and global arguments. In one embodiment, the shader records are identified by Shader Record Identifiers (SRIs). In one embodiment, each instance of execution of a shader is associated with a call stack 5203 that stores arguments passed between the parent shader and the child shader. The call stack 5121 may also store references to continuation functions that are executed on call returns.

The ray traversal circuit 5102 traverses each ray through the nodes of the BVH, working down the BVH hierarchy (e.g., through parent, child, and leaf nodes) to identify the nodes/primitives traversed by the ray. ray-BVH crossbar 5103 performs a cross test of rays, determines hit points on primitives, and generates results in response to hits. The traversal circuit 5102 and the crossbar circuit 5103 may retrieve work from one or more call stacks 5121. Within the ray trace acceleration circuit 5110, the call stack 5121 and associated ray trace data 4902 may be stored within a local Ray Trace Cache (RTC) 5107 or other local storage device for efficient access by the traversal circuit 5102 and the crossbar circuit 5103. One particular embodiment described below includes a high bandwidth ray bank (see, e.g., fig. 52).

The ray trace acceleration circuit 5110 may be a variation of the various traverse/crossbar circuits described herein, including ray-BVH traverse/crossbar circuit 4005, traverse circuits 4502 and 4503, and ray trace core 3150. Ray tracing acceleration circuitry 5110 may be used in place of ray-BVH traversal/intersection circuitry 4005, traversal circuitry 4502 and intersection circuitry 4503, and ray tracing core 3150, or any other circuitry/logic for processing a BVH stack and/or performing traversal/intersection. Accordingly, disclosure of any features in conjunction with ray-BVH traversal/intersection circuit 4005, traversal circuit 4502 and intersection circuit 4503, and ray tracing core 3150 described herein also discloses a corresponding combination with ray tracing acceleration circuit 6010, but is not limited thereto.

Referring to fig. 52, one embodiment of a ray traversal circuit 5102 includes first and second ray banks 5201 and 5202, respectively, where each bank includes a plurality of entries for storing a corresponding plurality of incident rays 5206 loaded from memory. The corresponding first and second stacks 5203 and 5204, respectively, include selected BVH node data 5290-5291, which is read from memory and stored locally for processing. As described herein, in one embodiment, the stack 5203-5204 is a "short" stack that includes a limited number of entries (e.g., six entries in one embodiment) for storing BVH node data. Although illustrated separately from the ray libraries 5201-5202, stacks 5203-5204 may also be maintained within the corresponding ray libraries 5201-5202. Alternatively, the stack 5203-5204 may be stored in a separate local memory or cache.

One embodiment of the traversal processing circuit 5210 alternates (e.g., in a ping-pong fashion) between the two banks 5201-5202 and the stacks 5203-5204 when selecting the next ray and node for processing. For example, walk processing circuit 5210 may select a new ray/BVH node from the alternate ray pool/stack on each clock cycle, thereby ensuring efficient operation. It should be noted, however, that this particular arrangement is not necessary to comply with the underlying principles of the invention.

In one embodiment, the ray distributor 5205 balances the entries of the incident rays 5206 into the first and second memory banks 5201 and 5202, respectively, based on the current relative values of a set of bank allocation counters 5220. In one embodiment, the bank allocation counter 5220 maintains a count of the number of traversals in each of the first and second memory banks 5201-5202. For example, a first bank allocation counter may be incremented when the ray allocator 5205 adds a new ray to the first bank 5201 and decremented when a ray from the first bank 5201 is processed. Similarly, the second bank allocation counter may be incremented when the ray allocator 5205 adds a new ray to the second bank 5201 and decremented when a ray from the second bank 5201 is processed.

In one embodiment, the ray distributor 5205 distributes the current ray to the bin associated with the smaller counter value. If the two counters are equal, the ray distributor 5205 may select either bank, or may select a bank different from the bank selected when the last counter was equal. In one embodiment, each ray is stored in one entry of one of the banks 5201-5202, and each bank includes 32 entries for storing up to 32 rays. However, the underlying principles of the invention are not limited to these details.

Apparatus and method for non-local mean filtering

"non-local mean" is an image processing technique for performing denoising. Unlike "local" mean filters, which smooth an image by determining the mean of blocks of pixels around a target pixel, non-local mean filters determine the mean of all pixels in the image (weighted according to their degree of similarity to the target pixel) so that post-filtering sharpness is improved and image detail loss is reduced compared to local mean algorithms.

In existing implementations, the non-local mean computation is performed in software. Thus, these implementations are not suitable for real-time rendering environments.

One embodiment of the invention uses existing motion estimation hardware blocks in the media processor pipeline to perform non-local mean filtering. Because existing motion estimation circuits are reused, only a small amount of additional circuitry is required. Using this hardware, embodiments of the invention can perform non-local mean filtering much more efficiently than current software implementations.

In addition, because the motion estimation circuitry is inherently configured to evaluate motion across image frames, one embodiment of the present invention uses this feature to perform inter-frame non-local mean filtering (rather than limiting operation to the current frame). Thus, in one implementation, efficient and accurate temporal denoising can be performed using non-local mean filtering within and/or across image frames.

One embodiment of the present invention includes a new interface that includes hardware and/or software components (e.g., programming interfaces) to make the motion estimation block of a media processor accessible to other processor components and applications. By way of example and not limitation, these may include depth estimation operations within a 3D graphics pipeline, de-interlacing of video images, and view interpolation for virtual reality implementations.

Many of the processor architectures described herein include media processors (sometimes referred to as "media engines" or "media pipelines") having dedicated circuitry for encoding and decoding video content (such as h.264 video, for example). These media processors include a media pipeline 234 (fig. 2B), a media pipeline 316 (fig. 3A, 4), a media engine 837 (fig. 8), and a media processing engine 3197 (fig. 31). Each such media processor includes circuitry to perform motion estimation for video processing. In one embodiment, instead of implementing a machine learning method for noise reduction, one embodiment of the present invention uses motion estimation hardware within the media processor(s) to perform non-local mean filtering.

Fig. 53 illustrates an embodiment of the invention in which motion estimation circuit 5320 of media processing engine 3197 performs non-local mean filtering on rendered image frame 5330 to generate denoised frame 5340. In particular, in one embodiment, media API 5302 (i.e., a set of commands/instructions capable of being executed by media processing engine 3197) is supplemented with non-local mean command 5305 to implement the techniques described herein. The non-local mean command 5305 accesses various architectural features of the motion estimation circuit 5320, including additional power provided by the motion estimation circuit 5320, to perform non-local mean filtering across the plurality of image frames 5330 (and not just within the current frame as in the current implementation). For example, when performing video encoding, one of the primary functions of the motion estimation circuit 5320 is to identify motion of blocks of pixels (e.g., macroblocks, slices, etc.) across successive frames. Thus, instead of performing non-local mean filtering on the target pixel by determining the mean of all pixels in a single image frame (weighted based on similarity to the target pixel), the motion estimation circuit 5320 is used to determine the mean of all pixels across multiple image frames.

The non-local mean filtering may be performed at the pixel or pixel block (e.g., macroblock) level. The blocks of pixels may include any arrangement of pixels supported by the motion estimation circuit 5320, including blocks having the same number of pixels in each dimension (16x16, 8x8, or 4x4) and blocks having different numbers of pixels in each dimension (e.g., 16x8, 8x8, or 4x4, etc.). Furthermore, the image frame may be encoded with an asymmetric arrangement of pixel blocks such that the central region of the image frame is encoded with higher accuracy (e.g., using 4x4 pixel blocks) and the periphery of the image frame is encoded with lower accuracy (e.g., using 16x16 pixel blocks).

Regardless of the type of encoding used, in one embodiment, the non-local mean command 5305 controls the motion estimation circuit 5320 to distinguish actual pixel data from "noise" in the image frame 5330 to produce the denoised image frame 5340. In particular, in one implementation, efficient and accurate temporal denoising is performed using non-local mean filtering within and/or across image frames using the motion estimation circuit 5320.

In operation, command streamer 5310 streams image rendering commands to ray trace circuitry 5110 (i.e., for traversal and intersection) and shader execution circuitry 4000 (i.e., for execution of shaders) to generate image frame 5330 using the various techniques described herein. Subsequently or concurrently, command streamer 5310 streams non-local mean command 5305 to media processing engine 3197, which evaluates successive image frames 5330 with motion estimation circuit 5320 executing commands to perform denoising and generate denoised image frame 5340.

One embodiment of the media API 5302 is provided with additional commands to make the motion estimation block 5320 accessible to other applications. By way of example and not limitation, these may include depth estimation operations within a 3D graphics pipeline, de-interlacing of video images, and/or view interpolation (e.g., for virtual reality implementations).

A method according to one embodiment is shown in fig. 54. The method may be implemented in the context of the architecture described herein, but is not limited to any particular processor or system architecture.

At 5401, a command stream transmitter transmits a command stream to the ray tracing circuitry and the shader execution circuitry. At 5402, the ray tracing circuitry and shader execution circuitry execute commands to render the image frame. At 5403, a command streamer transmits a non-local mean (NLM) command stream to the media processing circuit, and at 5404, executes commands using the motion estimation circuit to perform denoising by evaluating inter-frame pixel blocks.

Examples of the invention

The following are example implementations of different embodiments of the present invention.

Example 1. a processor, comprising: ray tracing circuitry to execute a first set of one or more commands to traverse a ray through a bounding volume level (BVH) to identify BVH nodes and/or primitives intersected by the ray; shader execution circuitry to execute one or more shaders, in response to a second set of one or more commands, to render a sequence of image frames based on the BVH nodes and/or primitives intersected by the ray; and a media processor including motion estimation circuitry to execute a third set of one or more commands to perform non-local mean filtering based on average pixel values collected across the image frame sequence to remove noise from the image frame sequence.

Example 2. the processor of example 1, further comprising: a command stream transmitter to stream the first set of commands to the ray tracing circuitry, to stream the second set of commands to the shader execution circuitry, and to stream the third set of commands to a media processing block.

Example 3. the processor of example 2, wherein the third set of commands includes an extension to an application programming interface, API, associated with the media processing block.

Example 4. the processor of example 1, wherein the non-local mean filtering of the target pixel comprises determining an average associated with all pixels across the sequence of images weighted by similarity to the target pixel.

5. The processor of example 1, wherein the shader execution circuitry includes a plurality of Execution Units (EUs) to execute a plurality of different shaders to render the sequence of image frames.

6. The processor of example 1, wherein the average pixel value is determined based on an evaluation of pixel blocks spread across the sequence of image frames.

7. The processor of example 6, wherein the pixel block comprises a macroblock.

8. The processor of example 7, wherein the macroblock comprises one or more of: a 16 × 16 pixel block, an 8 × 8 pixel block, a 4 × 4 pixel block, a 16 × 8 pixel block, an 8 × 4 pixel block, or a 16 × 4 pixel block.

9. A method, comprising: executing a first set of one or more commands on a ray tracing circuitry to traverse a ray through a bounding volume level (BVH) to identify BVH nodes and/or primitives intersected by the ray; executing one or more shaders, in response to a second set of one or more commands, to render a sequence of image frames based on the BVH nodes and/or primitives that intersect the ray; and executing, on a media processor comprising a motion estimation circuit, a third set of one or more commands to perform non-local mean filtering based on average pixel values collected across the image frame sequence to remove noise from the image frame sequence.

10. The method of example 9, further comprising: the first set of commands is streamed to the ray tracing circuitry, the second set of commands is streamed to the shader execution circuitry, and the third set of commands is streamed to the media processing block.

11. The method of example 10, wherein the third set of commands comprises an extension to an Application Programming Interface (API) associated with the media processing block.

12. The method of example 9, wherein the non-local mean filtering of the target pixel comprises determining an average associated with all pixels across the sequence of images weighted by similarity to the target pixel.

13. The method of example 9, wherein the one or more shaders execute on a plurality of Execution Units (EUs) to render the sequence of image frames.

14. The method of example 9, wherein the average pixel value is determined based on an evaluation of pixel blocks spread across the sequence of image frames.

15. The method of example 14, wherein the pixel block comprises a macroblock.

16. The method of example 15, wherein the macroblock comprises one or more of: a 16 × 16 pixel block, an 8 × 8 pixel block, a 4 × 4 pixel block, a 16 × 8 pixel block, an 8 × 4 pixel block, or a 16 × 4 pixel block.

17. A machine-readable medium having program code stored thereon, which when executed by a machine, causes the machine to perform operations comprising: executing a first set of one or more commands on a ray tracing circuitry to traverse a ray through a bounding volume level (BVH) to identify BVH nodes and/or primitives intersected by the ray; executing one or more shaders, in response to a second set of one or more commands, to render a sequence of image frames based on the BVH nodes and/or primitives that intersect the ray; and executing, on a media processor comprising a motion estimation circuit, a third set of one or more commands to perform non-local mean filtering based on average pixel values collected across the image frame sequence to remove noise from the image frame sequence.

18. The machine-readable medium of example 17, further comprising: the first set of commands is streamed to the ray tracing circuitry, the second set of commands is streamed to the shader execution circuitry, and the third set of commands is streamed to the media processing block.

19. The machine-readable medium of example 18, wherein the third set of commands comprises an extension to an Application Programming Interface (API) associated with the media processing block.

20. The machine-readable medium of example 17, wherein the non-local mean filtering of the target pixel comprises determining an average associated with all pixels across the sequence of images weighted by similarity to the target pixel.

21. The machine-readable medium of example 17, wherein the one or more shaders execute on a plurality of Execution Units (EUs) to render the sequence of image frames.

22. The machine-readable medium of example 17, wherein the average pixel value is determined based on an evaluation of pixel blocks spread across the sequence of image frames.

23. The machine-readable medium of example 22, wherein the pixel block comprises a macroblock.

24. The machine-readable medium of example 23, wherein the macro-block comprises one or more of: a 16 × 16 pixel block, an 8 × 8 pixel block, a 4 × 4 pixel block, a 16 × 8 pixel block, an 8 × 4 pixel block, or a 16 × 4 pixel block.

Embodiments of the invention may include various steps that have been described above. The steps may be embodied in machine-executable instructions, which may be used to cause a general-purpose or special-purpose processor to perform the steps. Alternatively, the steps may be performed by specific hardware components that contain hardwired logic for performing the steps, or by any combination of programmed computer components and custom hardware components.

As described herein, the instructions may relate to a particular configuration of hardware, such as an Application Specific Integrated Circuit (ASIC) configured to perform certain operations or having predetermined functionality, or software instructions stored in a memory embodied in a non-transitory computer readable medium. Thus, the techniques illustrated in the figures can be implemented using code and data stored and executed on one or more electronic devices (e.g., an end station, a network element, etc.). Such electronic devices use computer machine-readable media, such as non-transitory computer machine-readable storage media (e.g., magnetic disks; optical disks; random access memories; read-only memories; flash memory devices; phase change memories) and transitory computer machine-readable communication media (e.g., electrical, optical, acoustical or other form of propagated signals such as carrier waves, infrared signals, digital signals, etc.) to store and communicate (internally and/or over a network with other electronic devices) code and data.

Additionally, such electronic devices typically include a set of one or more processors coupled to one or more other components, such as one or more storage devices (non-transitory machine-readable storage media), user input/output devices (e.g., a keyboard, a touchscreen, and/or a display), and network connections. The set of processors is coupled to other components, typically through one or more buses and bridges (also referred to as bus controllers). The storage devices and the signals carrying the network traffic represent one or more machine-readable storage media and machine-readable communication media, respectively. Thus, the memory device of a given electronic device typically stores code and/or data for execution on the set of one or more processors of the electronic device. Of course, one or more portions of embodiments of the invention may be implemented using different combinations of software, firmware, and/or hardware. Throughout this detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced without some of these specific details. In certain instances, well-known structures and functions have not been described in detail so as not to obscure the subject matter of the present invention. Therefore, the scope and spirit of the present invention should be judged in terms of the claims which follow.