US 11,886,846 B2Granted

Method for executing computation, computing device, computing system, and storage medium

AssigneeShanghai Biren Technology Co., Ltd

InventorsHaiChuan Wang, Huayuan Tian, Long Chen

Life of this patent

Estimating the term…

Loading patent…

Abstract

A method for executing computation, a computing device, a computing system, and a storage medium are provided. The method includes: confirming, via a compiler, whether there is a call instruction related to a thread block modification request in a kernel function to be compiled; in response to confirming that there is the call instruction related to the thread block modification request in the kernel function to be compiled, determining a corresponding program segment associated with the call instruction; configuring a required thread block and thread local register for the corresponding program segment; and inserting a control instruction into the corresponding program segment to enable the thread block configured for the corresponding program segment to execute relevant computation of the corresponding program segment, and an unconfigured thread block not to execute the relevant computation. The disclosure can improve overall performance, make coding and maintenance easy and reduce error rate of code.

Claims12

Dependent claims

A method for executing computation, comprising:

confirming, via a compiler, whether there is a call instruction related to a thread block modification request in a kernel function to be compiled;
in response to confirming that there is the call instruction related to the thread block modification request in the kernel function to be compiled, determining a corresponding program segment associated with the call instruction;
configuring a required thread resource for the corresponding program segment;
and
inserting a control instruction into the corresponding program segment to enable the thread resource configured for the corresponding program segment to execute relevant computation of the corresponding program segment, wherein an unconfigured thread resource does not execute the relevant computation,
wherein enabling the thread resource configured for the corresponding program segment to execute the relevant computation of the corresponding program segment comprises:
controlling an executive stream through a control flow code to encapsulate the corresponding program segment, so that when a thread identifier of a current thread is determined to belong to a thread block identifier set configured for the corresponding program segment, the current thread is used to execute the relevant computation of the corresponding program segment;
and when the thread identifier of the current thread is determined to not belong to the thread block identifier set configured for the corresponding program segment, the current thread is skipped.

5 more elements

A computing system, comprising:

a memory comprising a compiler module;
a central processor communicatively connected to the memory and comprising a driver module;
and
a graphics processor communicatively connected to central processor and comprising an execution unit computation core module;
wherein the central processor is configured to execute the compiler module to identify a call instruction related to a thread block modification request comprised in a received kernel function to be compiled, and to convert the kernel function to be compiled of the call instruction to generate a compiled kernel function, wherein a corresponding program segment comprised in the compiled kernel function is configured with a control instruction;
wherein the central processor is configured to execute the driver module to determine a maximum number of threads required by the compiled kernel function to start based on the compiled kernel function;
and
wherein the graphics processor is configured to execute the execution unit computation core module to modify a function call of a thread block in response to the control instruction of the corresponding program segment in the compiled kernel function being executed.

5 more elements

Dependencies are inferred from the claim text.

Facts

Application: 17/686,413Filed 4 Mar 2022
Priority: CN 202110244679.95 Mar 2021
Published: US 2022/0283790 A18 Sep 2022
Granted: 30 Jan 2024
Term ends: Not estimatedThe source records don’t support an estimate.

Classification10 codes

G06F 8/41
Transformation of program code: compilation
G06F 8/45
Compilation: exploiting coarse grain parallelism in compilation, i.e. parallelism between groups of instructions
G06F 9/30054
To perform operations for flow control: unconditional branch instructions
G06F 9/30087
To perform miscellaneous control operations, e.g. NOP: synchronisation or serialisation instructions