Nvidia Corporation

Google Inc.

Jesse David Hall

Jerome Francis Duluk, Jr

Patrick Richard Brown

Henry Packard Moreton

Jeffrey A Bolz

Philip Alexander Cuadra

Gregory Scott Palmer

Andrew J Tao

Eric Sovelen Werness

Mark Dennis Stadler

Ignacio Llamas

Karim Abdalla

Timothy John Purcell

Daniel Elliot Wexler

Gernot Schaufler

Naveen Leekha

A method for providing state inheritance across command lists in a multi-threaded processing environment. The method includes receiving an application program that includes a plurality of parallel threads; generating a command list for each thread of the plurality of parallel threads; causing a first command list associated with a first thread of the plurality of parallel threads to be executed by a processing unit; and causing a second command list associated with a second thread of the plurality of parallel threads to be executed by the processing unit, where the second command list inherits from the first command list state associated with the processing unit.

Hardware for parallel command list generation

One embodiment of the present invention sets forth a technique for error-checking a compute task. The technique involves receiving a pointer to a compute task, storing the pointer in a scheduling queue, determining that the compute task should be executed, retrieving the pointer from the scheduling queue, determining via an error-check procedure that the compute task is eligible for execution, and executing the compute task.

Error checking in out-of-order task scheduling

Systems and techniques are provided that allow for installation of an application on a mobile device using a shader executable suitable for execution by the mobile device. When a request to install an application is received by an application distribution platform, a shader configured for the device is generated by the platform and provided with the application package for installation on the device. Thus, launch-time compilation and just-in-time re-compilation may be avoided.

Offline shader compilation

Systems and methods for auto-throttling encapsulated compute tasks. A device driver may configure a parallel processor to execute compute tasks in a number of discrete throttled modes. The device driver may also allocate memory to a plurality of different processing units in a non-throttled mode. The device driver may also allocate memory to a subset of the plurality of processing units in each of the throttling modes. Data structures defined for each task include a flag that instructs the processing unit whether the task may be executed in the non-throttled mode or in the throttled mode. A work distribution unit monitors each of the tasks scheduled to run on the plurality of processing units and determines whether the processor should be configured to run in the throttled mode or in the non-throttled mode.

Methods and apparatus for auto-throttling encapsulated compute tasks

One embodiment of the present invention sets forth a technique for tracking and filtering state change methods provided to a graphics pipeline. State shadow circuitry at the start of the graphics pipeline may be configured in different modes. A track mode is used to capture the current state by storing state change methods that are transmitted to the graphics pipeline. A passthrough mode is used to provide different state data to the graphics pipeline without updating the current state stored in the state shadow circuitry. A replay mode is used to restore the current state to the graphics pipeline using the state shadow circuitry. Additionally, the state shadow circuitry may also be configured to filter the state change methods that are transmitted to graphics pipeline by removing redundant state change methods.

State shadowing to support a multi-threaded driver environment

One embodiment of the present invention sets forth a method for generating work to be processed by a graphics pipeline residing within a graphics processor. The method includes the steps of receiving an indication that a first graphics workload is to be submitted to a command queue associated with the graphics processor, allocating a first portion of shader accessible memory for one or more units of state information that are necessary for processing the first graphics workload, populating the first portion of shader accessible memory with the one or more units of state information, and transmitting to the command queue of the graphics processor the one or more units of state information stored within the first portion of shader accessible memory, wherein the first graphics workload is processed within the graphics pipeline based on the one or more units of state information.

GPU work creation and stateless graphics in OPENGL

One embodiment of the present invention sets for a method for accessing data objects stored in a memory that is accessible by a graphics processing unit (GPU). The method comprises the steps of creating a data object in the memory based on a command received from an application program, transmitting a first handle associated with the data object to the application program such that data associated with different graphics commands can be accessed by the GPU, wherein the first handle includes a memory address that provides access to only a particular portion of the data object, receiving a first graphics command as well as the first handle from the application program, wherein the first graphics command includes a draw command or a compute grid launch, and transmitting the first graphics command and the first handle to the GPU for processing.

Bindless memory access in direct 3D

Techniques are disclosed for dispatching pixel information in a graphics processing pipeline. A fragment processing unit in the graphics processing pipeline generates a pixel that includes multiple samples based on a portion of a graphics primitive received by a thread. The fragment processing unit calculates a set of source values, where each source value corresponds to a different sample of the pixel. The fragment processing unit retrieves a set of destination values from a render target, where each destination value corresponds to a different source value. The fragment processing unit blends each source value with a corresponding destination value to create a set of final values, and creates one or more dispatch messages to store the set of final values in a set of output registers. One advantage of the disclosed techniques is that pixel shader programs perform per-sample operations with increased efficiency.

Programmable blending via multiple pixel shader dispatches

One embodiment of the present invention sets forth a technique for performing low latency computation on a parallel processing subsystem. A low latency functional node is exposed to an operating system. The low latency functional node and a generic functional node are configured to target the same underlying processor resource within the parallel processing subsystem. The operating system stores low latency tasks generated by a user application within a low latency command buffer associated with the low latency functional node. The parallel processing subsystem advantageously executes tasks from the low latency command buffer prior to completing execution of tasks in the generic command buffer, thereby reducing completion latency for the low latency tasks.

Low latency concurrent computation

One embodiment of the present invention sets forth a technique for storing only the enabled components for each enabled vector and writing only enabled components to one or more specified render targets. A shader program header (SPH) file provides per-component mask bits for each render target. Each enabled mask bit indicates that the pixel shader generates the corresponding component as an output to the raster operations unit. In the hardware, the per-component mask bits are combined with the applications programming interface (API)-level per-component write masks to determine the components that are updated by the shader program. The combined mask is used as the write enable bits for components in one or more render targets. One advantage of the combined mask is that the components that are not updated are not forwarded from the pixel shader to the ROP, thereby saving bandwidth between those processing units.

Pixel shader output map

Systems and methods for texture processing are presented. In one embodiment a texture method includes creating a sparse texture residency translation map; performing a probe process utilizing the sparse texture residency translation map information to return a finest LOD that contains the texels for a texture lookup operation; and performing the texture lookup operation utilizing the finest LOD. In one exemplary implementation, the finest LOD is utilized as a minimum LOD clamp during the texture lookup operation. A finest LOD number indicates a minimum resident LOD and a sparse texture residency translation map includes one finest LOD number per tile of a sparse texture. The sparse texture residency translation can indicate a minimum resident LOD.

Sparse texture systems and methods

One embodiment of the present invention sets forth a technique for configuring a graphics processing pipeline (GPP) to process data according to one or more shader programs. The method includes receiving a plurality of pointers, where each pointer references a different shader program header (SPH) included in a plurality of SPHs, and each SPH is associated with a different shader program that executes within the GPP. For each SPH included in the plurality of SPHs, one or more GPP configuration parameters included in the SPH are identified, and the GPP is adjusted based on the one or more GPP configuration parameters.

Shader program headers

One embodiment of the present invention sets forth a technique for providing state information to one or more shader engines within a processing pipeline. State information received from an application accessing the processing pipeline is stored in constant buffer memory accessible to each of the shader engines. The shader engines can then retrieve the state information during execution.

Providing pipeline state through constant buffers

One embodiment of the present invention sets forth a method macro expander (MME) coupled to a driver and the processing pipeline of a graphics processing unit. In operation, the MME receives, from the driver, a first packet of work indicating a macro stored in an instruction memory that is to be executed. The MME then executes the commands of the macro in the instruction memory to generate a second packet of work, and the second packet of work is then transmitted to the processing pipeline for further execution.

Growing community of inventors

Santa Clara, CA, United States of America

Jesse David Hall