Intel Corporation

Jayanth N Rao

Murali Sundaresan

Ali-Reza Adl-Tabatabai

Ben J Ashbaugh

Tatiana Shpeisman

Rajkishore Barik

Tomasz Janczak

Yang Ni

Ronald W Silvas

Michal Andrzej Mrozek

Pavan K Lanka

Ankur N Shah

Brandon L Fliflet

John A Tsakok

A method and system for sharing memory between a central processing unit (CPU) and a graphics processing unit (GPU) of a computing device are disclosed herein. The method includes allocating a surface within a physical memory and mapping the surface to a plurality of virtual memory addresses within a CPU page table. The method also includes mapping the surface to a plurality of graphics virtual memory addresses within an I/O device page table.

Memory sharing via a unified memory architecture

A method and system are described herein for an optimization technique on two aspects of thread scheduling and dispatch when the driver is allowed to pick the scheduling attributes. The present techniques rely on an enhanced GPGPU Walker hardware command and one dimensional local identification generation to maximize thread residency.

Scheduling and dispatch of GPGPU workloads

An apparatus and method are described for executing workloads without host intervention. For example, one embodiment of an apparatus comprises: a host processor; and a graphics processor unit (GPU) to execute a hierarchical workload responsive to one or more commands issued by the host processor, the hierarchical workload comprising a parent workload and a plurality of child workloads interconnected in a logical graph structure; and a scheduler kernel implemented by the GPU to schedule execution of the plurality of child workloads without host intervention, the scheduler kernel to evaluate conditions required for execution of the child workloads and determine an order in which to execute the child workloads on the GPU based on the evaluated conditions; the GPU to execute the child workloads in the order determined by the scheduler kernel and to provide results of parent and child workloads to the host processor following execution of all of the child workloads.

Method and apparatus for a highly efficient graphics processing unit (GPU) execution model

A mechanism is described for facilitating dynamic pipelining of workload executions at graphics processing units on computing devices. A method of embodiments, as described herein, includes generating a command buffer having a plurality of kernels relating to a plurality of workloads to be executed at a graphics processing unit (GPU), and pipelining the workloads to be processed at the GPU, where pipelining includes scheduling each kernel to be executed on the GPU based on at least one of availability of resource threads and status of one or more dependency events relating to each kernel in relation to other kernels of the plurality of kernels.

Facilitating dynamic pipelining of workload executions on graphics processing units on computing devices

According to some embodiments, a graphics processor may abort a workload without requiring changes to the kernel code compilation or intruding upon graphics processing unit execution. Instead, it is possible to only read the predicate state once before starting and once before restarting a workload that has been preempted because the user wishes to abort the work. This avoids the need to read from each execution unit, reducing the drain on memory bandwidth and increasing power and performance in some embodiments.

Aborting graphics processor workload execution

A method and system for shared virtual memory between a central processing unit (CPU) and a graphics processing unit (GPU) of a computing device are disclosed herein. The method includes allocating a surface within a system memory. A CPU virtual address space may be created, and the surface may be mapped to the CPU virtual address space within a CPU page table. The method also includes creating a GPU virtual address space equivalent to the CPU virtual address space, mapping the surface to the GPU virtual address space within a GPU page table, and pinning the surface.

Shared virtual memory

A method and apparatus to facilitate shared pointers in a heterogeneous platform. In one embodiment of the invention, the heterogeneous or non-homogeneous platform includes, but is not limited to, a central processing core or unit, a graphics processing core or unit, a digital signal processor, an interface module, and any other form of processing cores. The heterogeneous platform has logic to facilitate sharing of pointers to a location of a memory shared by the CPU and the GPU. By sharing pointers in the heterogeneous platform, the data or information sharing between different cores in the heterogeneous platform can be simplified.

Method and apparatus to facilitate shared pointers in a heterogeneous platform

A method for sharing memory between a central processing unit (CPU) and an input/output (I/O) device of a computing device is described. The method may include creating an allocation of memory for the I/O device to operate on. The method includes detecting whether the allocation is not page-aligned, wherein an allocation is page-aligned when its base address and size be evenly divisible by the applicable page-size. The allocation may be successfully shared, even if not page-aligned, even if an operating system of the computing device doesn't support sharing of non-page-aligned allocations.

Sharing non-page aligned memory

Apparatuses, systems, and methods may sample a texture, manage a page fault, and/or switch a context associated with the page fault. A three-dimensional (3D) graphics pipeline may provide texture sample location data corresponding to a texture, wherein sampling of the texture is to be executed external to the 3D graphics pipeline. A compute pipeline may execute sampling of the texture utilizing the texture sample location data and provide texture sample result data corresponding to the texture, wherein the 3D graphics pipeline may composite a frame utilizing the texture sample result data. The compute pipeline may manage a page fault, wherein the page fault and/or management of the page fault may be hidden from a graphics application. In addition, the compute pipeline may switch a compute context associated with the page fault to allow a graphics task not associated with the page fault to be executed and/or to prevent a stall.

Growing community of inventors

Folsom, CA, United States of America

Jayanth N Rao