WO2019183861A1

WO2019183861A1 - Method, device, and machine readable storage medium for task processing

Info

Publication number: WO2019183861A1
Application number: PCT/CN2018/080970
Authority: WO
Inventors: 李庆; 夏昌奇; 张晓炜
Original assignee: 深圳市大疆创新科技有限公司
Priority date: 2018-03-28
Filing date: 2018-03-28
Publication date: 2019-10-03
Also published as: CN110494848A

Abstract

Provided are a method, a device, and a machine readable storage medium for task processing. The method comprises: selecting a task to be processed from a processing queue; selecting at least one target graphics processor corresponding to the task from a plurality of graphics processors; allocating the task to the target graphics processor; and processing the task by means of the target graphics processor. By use of embodiments of the present invention, real-time data processing is ensured, multiple graphics processors can be reasonably managed, multiple tasks to be processed are effectively scheduled, the utilization rates of the graphics processors are maximized, the processing speed is optimized, accuracy and effectiveness of a processing result are effectively ensured, and reliability is improved.

Description

Task processing method, device and machine readable storage medium

Technical field

The present invention relates to the field of image processing technologies, and in particular, to a task processing method, device, and machine readable storage medium.

Background technique

At present, the processing of data such as images or radars can be based on platforms such as ARM (Advanced RISC Machines, Advanced Reduced Instruction Set Processor), DSP (Digital Signal Processing), or CPU (Central Processing Unit). The amount of data that can be processed is affected by factors such as processor frequency, memory size, and transmission bandwidth. In artificial intelligence applications such as autopilot, sensor data and acquisition frequency make the data level reach GB/s or higher, resulting in Platforms such as ARM, DSP or CPU are no longer able to meet the needs of real-time processing.

Summary of the invention

The present invention provides a task processing method, apparatus, and machine readable storage medium.

A first aspect of the present invention provides a task processing method for a device including a plurality of graphics processors, the method comprising: selecting a task to be processed from a processing queue;

Selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors; assigning the to-be-processed task to the target graphics processor;

The to-be-processed task is processed by the target graphics processor.

A second aspect of the present invention provides a task processing apparatus including a scheduler and a plurality of graphics processors, wherein the scheduler is configured to select a task to be processed from a processing queue and select from the plurality of graphics processors At least one target graphics processor corresponding to the to-be-processed task, and assigning the to-be-processed task to the target graphics processor;

The target graphics processor is configured to process the to-be-processed task.

According to a third aspect of the present invention, a computer readable storage medium is stored, the computer readable storage medium storing a plurality of computer instructions, and the task processing method is implemented when the computer instructions are executed.

Based on the foregoing technical solution, in the embodiment of the present invention, a target graphics processor corresponding to a task to be processed may be selected from a plurality of graphics processors, a task to be processed is allocated to the target graphics processor, and the target graphics processor is adopted The task to be processed is processed, so that the real-time processing of the data is performed based on the graphics processor, and the real-time performance of the data processing is ensured, and the data of the GB/s level can be processed in real time. In addition, multiple graphics processors can be reasonably managed, and multiple tasks to be processed can be effectively scheduled, so that the utilization of the graphics processor can be utilized to the maximum extent, and the processing speed can be optimized to ensure the accuracy of the processing results. And effectiveness, increasing reliability.

DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings to be used in the embodiments of the present invention or in the description of the prior art will be briefly described below. Obviously, the drawings in the following description It is merely some of the embodiments described in the present invention, and those skilled in the art can also obtain other drawings according to the drawings of the embodiments of the present invention.

1 is a schematic diagram of an application scenario of an embodiment of the present invention;

2 is a flow chart of an embodiment of a task processing method of the present invention;

3 is a schematic diagram of processing a task to be processed by a target graphics processor of the present invention;

4 is a schematic diagram of data sharing by different processing threads of the present invention;

Figure 5 is a block diagram of one embodiment of a task processing device of the present invention.

detailed description

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention. Further, the features of the following embodiments and examples may be combined with each other without conflict.

The terminology used herein is for the purpose of describing particular embodiments, The singular forms "a", "the" and "the" It will be understood that the term "and/or" as used herein refers to any and all possible combinations of one or more of the associated listed items.

Although the terms first, second, third, etc. may be used to describe various information in the present invention, such information should not be limited to these terms. These terms are used to distinguish the same type of information from each other. For example, the first information may also be referred to as the second information without departing from the scope of the invention. Similarly, the second information may also be referred to as the first information. Depending on the context, in addition, the word "if" may be interpreted as "when", or "when", or "in response to determination."

A method for processing a task is provided in the embodiment of the present invention. The method can be applied to a processing device including a plurality of graphics processing units (GPUs). The type of the processing device is not limited, as long as there are multiple graphics. The processor is fine. The task processing method may be a multi-task real-time processing method based on multiple graphics processors, and the multi-channel sensor data is used as an input, which can be applied in an application scenario such as automatic driving, assisted driving, indoor and outdoor working robots. In addition, the above sensor data may be image data, such as camera to captured image data, image data collected by radar (such as laser radar, millimeter wave radar, etc.), and the data type is not limited.

In artificial intelligence application scenarios, such as artificial intelligence application scenarios where data needs to be processed in real time, as the number of sensors increases and the acquisition frame rate increases, the amount of data increases exponentially, and ARM, DSP, or CPU cannot implement real-time processing of data. Therefore, the present embodiment proposes a multi-task real-time processing method based on multiple graphics processors, which can ensure real-time processing of data and accuracy and effectiveness of processing results.

In this embodiment, in consideration of the need to process multiple tasks in parallel, and different tasks have different processing requirements, in this embodiment, by using the powerful computing performance of the graphics processor, real-time task processing based on multiple graphics processors is proposed, and multi-task can be realized. Parallel scheduling, and achieve tasks such as fairness, priority, independence, interruption and recovery, can manage multiple graphics processors reasonably, effectively schedule multiple tasks to be processed, and make use of graphics processors The utilization rate is maximized, the processing speed is optimized, the accuracy and effectiveness of the processing result are effectively ensured, and the reliability is increased.

FIG. 1 is a schematic diagram of an application scenario according to an embodiment of the present invention. The multi-task scheduling module is configured to cache a task, complete task scheduling, and apply a response task processing result to an upper layer. The processing module is composed of a plurality of graphics processors, and the real-time processing of tasks is completed by the graphics processor. The statistical monitoring module is configured to monitor and collect the status information of the task and the resource information of the graphics processor, and feed back the status information of the task and the resource information of the graphics processor to the multi-task scheduling module, and the multi-task scheduling module utilizes the status information of the task. Task scheduling with resource information of the graphics processor.

Embodiment 1:

As shown in FIG. 2, it is a flowchart of a task processing method in an embodiment of the present invention, where the method includes:

Step 201: Select a task to be processed from the processing queue.

Specifically, after receiving the task, the task may be stored in the processing queue. When the task needs to be processed, the task may be selected from the processing queue, and the selection manner is not limited. For the sake of convenience, the task selected from the processing queue may be referred to as a pending task.

Step 202: Select at least one target graphics processor corresponding to the to-be-processed task from a plurality of graphics processors (ie, all graphics processors). Wherein, for convenience of distinction, a graphics processor selected from a plurality of graphics processors may be referred to as a target graphics processor corresponding to a task to be processed.

Specifically, after the task to be processed is selected from the processing queue, the status information of the to-be-processed task and/or the resource information of each graphics processor may be selected, and all the graphics processors are selected to correspond to the to-be-processed task. At least one target graphics processor, there is no limit to this selection.

Step 203: Assign the to-be-processed task to the target graphics processor.

Specifically, after selecting a task to be processed from the processing queue and selecting a target graphics processor for the to-be-processed task, the to-be-processed task may be allocated to the target graphics processor, that is, the to-be-processed The task is assigned to the idle processing thread of the target graphics processor. For example, a pending task can be assigned to all or part of the idle processing thread of the target graphics processor.

Step 204: Process the to-be-processed task by the target graphics processor.

Specifically, after the task to be processed is allocated to the idle processing thread of the target graphics processor, the to-be-processed task can be processed by the idle processing thread, and the processing is not limited. For example, the pending task may include data (such as sensor data, etc.) and a task type, and the idle processing thread of the target graphics processor may perform processing corresponding to the task type based on the data.

Embodiment 2:

Before the step 201, after the task to be processed is received, if the task to be processed is a static scheduling task whose time series and/or data has a dependency, the task to be processed may be cached into the static processing queue; The task is a time series and/or data independent dynamic scheduling task, and the pending task can be cached into the dynamic processing queue. On this basis, in step 201, selecting a task to be processed from the processing queue may include: if the processing queue is a static processing queue, the time series and/or the data may be selected from the static processing queue. A task to be processed; if the processing queue is a dynamic processing queue, a time series and/or data independent pending task can be selected from the dynamic processing queue.

In this embodiment, the processing queue can be divided into a static processing queue (Static) and a dynamic processing queue (Dynamic). The static processing queue may be a task for storing a time series having a dependency relationship, or a task for storing data having a dependency relationship, or a task for storing a time series and data having a dependency relationship; A task with dependencies is called a static scheduled task. In addition, the dynamic processing queue may be used to store time series independent tasks, or to store data independent tasks, or to store time series and data independent tasks; for the convenience of distinction, the above independent tasks may be referred to as dynamic Schedule tasks.

Moreover, all tasks stored in the static processing queue can be processed in the same graphics processor, or can be processed in the associated graphics processor, that is, there are restrictions on the graphics processor; The correspondence between the static processing queue and the graphics processor (such as graphics processor 1), therefore, all tasks of the static processing queue need to be processed by the graphics processor 1. In addition, all tasks stored in the dynamic processing queue can be processed in the same graphics processor or processed in different graphics processors. That is, there is no limit to the graphics processor, and the dynamic processing queue can be processed. All tasks are assigned to any graphics processor.

Among them, the time series has a dependency task: the processing of one task depends on the previous task. For example, in H264 encoding, non-I frame tasks depend on the previous I frame task. Based on this, after receiving the task 1 and the task 2, if the task 1 and the task 2 have a dependency in the time series, the task 1 and the task 2 can be stored in the static processing queue. In addition, the task of data dependency is that the processing of one data depends on the previous data, and there is no limitation on this. Moreover, all tasks stored in the static processing queue can be processed in the same graphics processor or can be processed in the associated graphics processor.

Among them, the time series independent task means that the processing of one task does not depend on the previous task, that is, the task is an independent task. For example, tasks such as image conversion and point cloud algorithm processing, each task is an independent task, and does not depend on the previous task. Based on this, after receiving the task 3, if the task 3 is independent in time series, the task 3 can store the task 3 to the dynamic processing queue. In addition, the data-independent task means that the processing of one data is independent of other data, and there is no limitation on this. Moreover, all tasks stored in the dynamic processing queue can be processed in the same graphics processor or in different graphics processors.

For each pending task in the static processing queue, the multi-task scheduling module may select a task to be processed from the static processing queue, and may select another pending task from the static processing queue when the pending task is processed or not processed. . For each pending task in the dynamic processing queue, the multi-task scheduling module may select a to-be-processed task from the dynamic processing queue, and may select another pending processing from the dynamic processing queue when the pending task is processed or not processed. task.

Embodiment 3:

Before the step 201, after the task to be processed is received, the task to be processed may be cached into a task queue (which may also be referred to as a large queue or a Task Queue), and the task queue may include multiple tasks to be processed. For each pending task in the task queue, if the pending task is a static scheduled task whose time series and/or data has a dependency, the pending task may be cached into a static processing queue; if the pending task is Time-series and/or data-independent dynamic scheduling tasks can cache pending tasks to a dynamic processing queue. On this basis, in step 201, selecting a task to be processed from the processing queue may include: if the processing queue is a static processing queue, the time series and/or the data may be selected from the static processing queue. A task to be processed; if the processing queue is a dynamic processing queue, a time series and/or data independent pending task can be selected from the dynamic processing queue. The processing procedure of the third embodiment is similar to that of the second embodiment, and details are not described herein again.

Embodiment 4:

In step 201, selecting a task to be processed from the processing queue may include, but is not limited to, acquiring a priority of each to-be-processed task in the processing queue; and selecting a high-priority priority from the processing queue based on the priority priority. Handling tasks. The obtaining the priority of the to-be-processed task in the processing queue may include: obtaining, for each pending task in the processing queue, a task type of the to-be-processed task; and then querying the mapping by using the task type The table obtains a priority corresponding to the task type, where the mapping table is used to record a correspondence between the task type and the priority.

The multi-task scheduling module can be configured with a mapping table. As shown in Table 1, the mapping table is used to record the correspondence between the task type and the priority. On the basis of this, it is assumed that the task to be processed 1 includes the task type A, the task to be processed 2 includes the task type B, and the task to be processed 3 includes the task type C, and the multitask scheduling module queries the mapping table through the task type A to obtain the task to be processed. Priority 5 of 1 is queried by the task type B to obtain the priority 3 of the task 2 to be processed, and the mapping table is queried by the task type C to obtain the priority 1 of the task 3 to be processed. Obviously, since the priority 5 of the task 1 to be processed is higher than the priority 3 of the task 2 to be processed, the priority 3 of the task 2 to be processed is higher than the priority 1 of the task 3 to be processed, and therefore, the multitask scheduling module first processes Select pending task 1 in the queue, then select pending task 2, then select pending task 3, and so on.

Table 1

任务类型Task type	优先级priority
任务类型ATask type A	55
任务类型BTask type B	33
任务类型CTask type C	11

Embodiment 5: In order to select a target graphics processor, state information of a task to be processed and/or resource information of a graphics processor may be acquired first, and the acquisition process of the information is described in detail below.

Case 1: The multi-task scheduling module obtains status information of the task to be processed.

During the processing of the to-be-processed task, the statistical monitoring module may obtain the status information of the to-be-processed task that has been processed, and record the correspondence between the task type of the to-be-processed task and the status information of the to-be-processed task in the status information table. Send the status information table to the multitasking scheduling module. On the basis of the task, the multi-task scheduling module obtains the task type of the task to be processed after selecting the task to be processed from the processing queue, and queries the state information table through the task type to obtain state information corresponding to the task type. The obtained status information is determined as status information of the task to be processed.

For example, after the processing task 1, the pending task 2, and the to-be-processed task 3 have been processed, the statistical monitoring module can obtain the status information table shown in Table 2, and send the status information table to the multi-task scheduling module. On the basis of this, after the multi-task scheduling module selects the to-be-processed task 4 from the processing queue, if the task type of the task to be processed 4 is the task type A, the state information A corresponding to the task type A can be obtained, that is, The status information of the task 4 to be processed is status information A.

Table 2

任务类型Task type	状态信息status information
任务类型ATask type A	状态信息AStatus information A
任务类型BTask type B	状态信息BStatus information B
任务类型CTask type C	状态信息CStatus information C

Case 2: The multitasking scheduling module acquires resource information of the graphics processor.

During the processing of the to-be-processed task, the statistical monitoring module may acquire (eg, periodically acquire) resource information of each graphics processor, and send the resource information of each graphics processor to the multi-task scheduling module, so that multi-tasking The scheduling module can obtain resource information for each graphics processor.

In the above embodiment, the status information may include a task processing time, which is a time difference between the task completion time and the task reception time. Further, the resource information may include: the number of idle processing threads; and/or the state of the processing thread (eg, occupied state or idle state).

The statistical monitoring module may perform statistics on the status information of the processing task and notify the multi-task scheduling module of the correspondence between the task type and the status information. Specifically, the processing of the to-be-processed task may go through three stages, a task receiving buffer stage, a task scheduling distribution stage, and a task processing stage. Therefore, the statistical monitoring module may perform statistics on the time consumed by the above three stages, and the three The time consumed by the phase is also the state information of the task to be processed, that is, the task processing time of the task to be processed. In addition, the task processing time may be the time difference between the task completion time and the task receiving time. Therefore, the statistical monitoring module may also perform statistics on the task receiving time and the task completion time of the processing task, and calculate the time difference between the task completion time and the task receiving time. The time difference is the task processing time of the to-be-processed task.

When the statistical monitoring module sends the corresponding relationship between the task type and the task processing time to the multi-task scheduling module, each task to be processed has one task processing time and corresponding to the different tasks to be processed. Task processing times can be the same or different. Based on this, the maximum value is selected from the task processing time of the plurality of to-be-processed tasks corresponding to the task type, and the maximum value is used as the task processing time corresponding to the task type. Or, select a minimum value from the task processing time of the plurality of to-be-processed tasks corresponding to the task type, and use the minimum value as the task processing time corresponding to the task type. Or, the average value of the task processing time of the plurality of to-be-processed tasks corresponding to the task type is calculated, and the average value is taken as the task processing time corresponding to the task type. Alternatively, the maximum value, the minimum value, and the average value may be used as the task processing time corresponding to the task type, and no limitation is imposed thereon.

The statistical monitoring module may perform statistics on the resource information of the graphics processor and notify the multi-task scheduling module of the resource information of the graphics processor during the processing of each to-be-processed task. Specifically, the graphics processor can support multi-task parallel processing, each graphics processor has multiple processing threads, and the to-be-processed tasks are allocated to the processing thread for processing. Therefore, the statistical monitoring module can process threads for each graphics processor. Monitoring is performed, such as monitoring the number of idle processing threads of the graphics processor, monitoring the state of the processing thread of the graphics processor (eg, occupied state or idle state).

Example 6:

In step 202, selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors may include: selecting and waiting from the plurality of graphics processors according to the state information of the to-be-processed task The target graphics processor corresponding to the processing task. Alternatively, a target graphics processor corresponding to the to-be-processed task is selected from a plurality of graphics processors according to resource information of each graphics processor. Alternatively, the target graphics processor corresponding to the to-be-processed task is selected from the plurality of graphics processors according to the status information of the to-be-processed task and the resource information of each graphics processor.

In the first case, the target graphics processor corresponding to the to-be-processed task is selected from the plurality of graphics processors according to the status information of the to-be-processed task, and may include: determining task processing time according to status information of the to-be-processed task, and processing time for the task A plurality of pending tasks that are greater than the time threshold (indicating that their task processing time is relatively long), and different target graphics processors are selected for different pending tasks.

The time threshold may be set according to experience. When the task processing time is greater than the time threshold, the processing time is relatively long. Therefore, by assigning multiple to-be-processed tasks whose task processing time is greater than the time threshold to different graphics processors, it is ensured. Each task to be processed is scheduled in time, and all the resources of the graphics processor are allocated and utilized reasonably, to prevent the task processing from timing out, and to ensure that tasks with relatively large processing time do not occupy multiple processing threads of the same graphics processor at the same time.

In the second case, the target graphics processor corresponding to the to-be-processed task is selected from the plurality of graphics processors according to the resource information of the graphics processor, and may include: determining the number of idle processing threads according to the resource information of the graphics processor, and Among the graphics processors, the graphics processor with the most idle processing threads is selected as the target graphics processor. Since the graphics processor with the largest number of idle processing threads is the least utilized graphics processor, the graphics processor with the largest number of idle processing threads is selected as the target graphics processor, and the resources of all graphics processors can be reasonably allocated and use.

In the third case, the target graphics processor corresponding to the to-be-processed task is selected from the plurality of graphics processors according to the resource information of the graphics processor, and may include: determining an idle processing thread according to the resource information of the graphics processor, and setting the idle processing thread The corresponding graphics processor is determined to be the target graphics processor. Since the graphics processor corresponding to the idle processing thread is determined as the target graphics processor, after the task to be processed is allocated to the target graphics processor, the target graphics processor has an idle processing thread that processes the to-be-processed task, avoiding the target graphics processor. There are no idle processing threads that handle pending tasks.

Wherein, when the state of the processing thread is an idle state, it indicates that the processing thread is an idle processing thread, and when the state of the processing thread is an occupied state, it indicates that the processing thread is an occupied processing thread. Obviously, since the graphics processor corresponding to the idle processing thread is determined as the target graphics processor, the pending task is allocated when the processing thread is idle, so that the pending task can be processed in time.

Case 4: selecting, according to the status information of the task to be processed and the resource information of each graphics processor, a target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors, including: for the task processing time being greater than a time threshold Multiple pending tasks, selecting different target graphics processors for different pending tasks; selecting a graphics processor with more idle processing threads from multiple graphics processors when selecting the target graphics processor, or selecting to have idle A graphics processor that processes threads.

Example 7:

In step 202, selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors may include: selecting, for a plurality of to-be-processed tasks for parallel processing, corresponding to each of the plurality of to-be-processed tasks The target graphics processor is the same or different from the target graphics processor corresponding to different pending tasks. For example, when different to-be-processed tasks correspond to different target graphics processors, multiple to-be-processed tasks processed in parallel may be separately distributed to multiple graphics processors, thereby rationally allocating and utilizing resources of all graphics processors to ensure that each The pending tasks are processed in a timely manner.

In step 202, selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors may include: if the processing queue is a static processing queue, querying from the plurality of graphics processors The graphics processor corresponding to the static processing queue (in the above embodiment, the correspondence between the static processing queue and the graphics processor may be pre-configured, and based on the correspondence, the graphics processing corresponding to the static processing queue may be queried. And determining the queried processor to be the target graphics processor corresponding to the to-be-processed task. If the processing queue is a dynamic processing queue, the target graphics processor is selected from all graphics processors, see the above embodiment.

Example 8:

In step 204, processing the to-be-processed task by the target graphics processor may include: the target graphics processor may utilize image processing, feature point tracking, semi-global stereo block matching, radar and camera self-calibration, point cloud tracking, local Algorithms such as maps and deep learning, which deal with processing tasks, do not limit this process. For example, when the image processing algorithm is used to process the processing task, the source image may be subjected to format conversion, distortion correction, and image data of an expected format or a special format may be output. For example, when the feature point tracking algorithm is used to process the processing task, the correspondence between the previous frame and the current frame can be obtained by using the change of the pixel in the image sequence in the time domain and the correlation between adjacent frames. , thereby calculating motion information of an object between adjacent frames. For another example, when the processing task is processed by the deep learning algorithm, the depth of the target object relative to the camera can be calculated according to the positional relationship between the parallax and the binocular in the binocular imaging of the same target object.

The following describes the task to be processed by the target graphics processor in combination with specific situations.

Case 1: When a pending task (such as a low priority pending task) is processed by the target graphics processor, if there is a pending task with a higher priority than the low priority pending task, the low priority is interrupted. The pending task is to process the higher priority pending task through the target graphics processor; after the higher priority pending task is processed, the low priority pending task is restored.

Referring to FIG. 3, it is assumed that the target graphics processor is processing the pending task 1, and a new pending task 2 is received. The target graphics processor determines if there is an idle processing thread. If there is one, the pending task 2 is assigned to the idle processing thread. If not, compare the priority of the task 1 to be processed with the priority of the task 2 to be processed. If the priority of the task to be processed 1 is high, the task 1 to be processed is continuously processed, and the task 2 to be processed is in a waiting state; after the processing of the task 1 to be processed is completed, the task 2 to be processed is processed. If the priority of the task to be processed 2 is high, the task 1 to be processed can be interrupted, and the task 2 to be processed can be processed; after the processing of the task 2 to be processed is completed, the task 1 to be processed is restored.

Case 2: When the pending task is processed by the target graphics processor, if the pending task is abnormal, the pending task is interrupted, the priority of the pending task is increased, and the to-be-processed task is cached in the processing queue. The priority of the to-be-processed task is increased, so that the to-be-processed task can be preferentially selected from the processing queue to prevent the pending task from waiting for timeout.

Case 3: A plurality of to-be-processed tasks assigned to the target graphics processor are processed in parallel by the target graphics processor; wherein the parallel processing may include synchronous serial processing and kernel asynchronous processing.

Wherein, when the target graphics processor processes a plurality of to-be-processed tasks, the synchronous processing or the kernel asynchronous processing may be used to perform parallel processing on the plurality of to-be-processed tasks, and the processing manner is not limited. It can effectively save processing time and improve processing efficiency.

Case 4: When the target graphics processor processes the processing task, the address of the central processing unit is latched, and the interaction data between the target graphics processor and the central processing unit is transmitted through the DMA controller. For example, when the amount of data of the task to be processed is large, the address of the central processing unit is page-locked by calling the cudaHostRegister interface, and the data interaction between the central processing unit and the graphics processor is realized by the DMA controller, thereby significantly improving Bandwidth, reducing unnecessary copies of data.

Case 5: When the target graphics processor processes the processing task, data sharing is performed through different processing threads. Specifically, if multiple to-be-processed tasks are completed by multiple processing threads, data sharing between processing threads can be performed, and after multiple copies of memory and video memory are avoided, sharing of memory and memory can be realized in the processing thread, and the elimination is omitted. The copy process saves a lot of processing time.

In this embodiment, the target graphics processor can process the processing task by using image processing, feature point tracking, semi-global stereo block matching, radar and camera self-calibration, point cloud tracking, local map, deep learning, etc., therefore, The processing thread may include: an image processing thread, a feature point tracking processing thread, a semi-global block matching processing thread (SBM), a radar and a camera self-calibration processing thread (hereinafter referred to as a self-calibration processing thread), Point cloud tracking processing thread, map processing thread, deep learning processing thread (Deep Learning).

The image processing thread is configured to receive the original data and process the original data to obtain a grayscale image and an RGB image after the distortion and epipolar correction. The feature point tracking processing thread is used to perform feature point detection and tracking on the image. The semi-global stereo block matching processing thread is used to match the binocular image, obtain a disparity map, and calculate a three-dimensional point cloud. The self-calibration processing thread is used to calibrate the external parameters between the camera and the camera, radar and camera. The point cloud tracking processing thread is used to split the 3D point cloud into different objects, perform target tracking and area detection, and send the result to the map processing thread. The deep learning processing thread is used to detect and track the RGB image and send the result to the map processing thread. The map processing thread is configured to receive the processing result of the point cloud tracking processing thread and the processing result of the deep learning processing thread, and generate the partial map by using the received information.

See Figure 4 for a schematic diagram of data sharing through different processing threads.

Wherein, when the image processing thread processes the processing task, the output data may also be provided to the self-calibration processing thread, and the self-calibration processing thread processes the processing task according to the output data; or, the output data may be provided to the feature point. Tracking the processing thread, wherein the feature point tracking processing thread processes the processing task according to the output data; or, the output data may be provided to the semi-global stereo block matching processing thread, and the semi-global stereo block matching processing thread according to the output data The processing task is processed; or the output data can be provided to a deep learning processing thread, and the deep learning processing thread processes the processing task according to the output data.

Wherein, when the semi-global stereo block matching processing thread processes the processing task, the output data may also be provided to the point cloud tracking processing thread, and the point cloud tracking processing thread processes the processing task according to the output data. Further, when the point cloud tracking processing thread processes the processing task, the output data may also be provided to the map processing thread; in addition, when the deep learning processing thread processes the processing task, the output data may also be provided to the map processing. Threading; based on this, the map processing thread may process the processing task according to the input data provided by the point cloud tracking processing thread and the input data provided by the deep learning processing thread.

Example 8:

Based on the same inventive concept as the above method, the embodiment of the present invention further provides a task processing device. As shown in FIG. 5, the task processing device includes a scheduler and a plurality of graphics processors; and the scheduler is configured to Selecting a task to be processed in the processing queue, and selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors, and assigning the to-be-processed task to the target graphics processor; The target graphics processor is configured to process the to-be-processed task.

When the scheduler selects a task to be processed from the processing queue, it is specifically used to: if the processing queue is a static processing queue, select a to-be-processed task whose time series and/or data has a dependency from the static processing queue.

When the scheduler selects a task to be processed from the processing queue, it is specifically used to: if the processing queue is a dynamic processing queue, select a time series and/or data independent pending task from the dynamic processing queue.

The scheduler is further configured to receive a task to be processed; if the task to be processed is a static scheduling task with a time series and/or data having a dependency, the task to be processed is cached into a static processing queue; The to-be-processed task is a time-series and/or data-independent dynamic scheduling task, and the to-be-processed task is cached into a dynamic processing queue.

The scheduler is further configured to receive a task to be processed, and cache the task to be processed into a task queue. For the task to be processed in the task queue, if the task to be processed is a static sequence task with time series and/or data having dependencies, The pending task is cached to the static processing queue; if the pending task is a time series and/or data independent dynamic scheduling task, the pending task is cached into the dynamic processing queue.

When the scheduler selects a task to be processed from the processing queue, it is used to: obtain a priority of a task to be processed in the processing queue; and select a high priority task to be processed from the processing queue based on the priority priority.

When the scheduler obtains the priority of the task to be processed in the processing queue, the method is specifically configured to: obtain a task type of the task to be processed in the processing queue; and query the mapping table by using the task type to obtain a priority corresponding to the task type. Level; the mapping table is used to record the correspondence between task types and priorities.

In an example, when the scheduler selects at least one target graphics processor corresponding to the to-be-processed task, the scheduler is specifically configured to: according to state information of the to-be-processed task and/or The resource information of the graphics processor, selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors.

The scheduler is further configured to acquire a task type of the to-be-processed task, and query the state information table by using the task type to obtain state information corresponding to the task type, where the state information table is used to record a task type. Corresponding relationship with the status information; determining the obtained status information as status information of the to-be-processed task.

In an example, as shown in FIG. 4, the task processing device further includes: a monitor, configured to acquire state information of the to-be-processed task that has been processed, and record the task of the to-be-processed task in the state information table. Corresponding relationship between the type and the status information of the to-be-processed task, and sending the status information table to the scheduler.

In the above embodiment, the status information includes a task processing time; the task processing time is a time difference between the task completion time and the task receiving time.

The monitor is further configured to acquire resource information of the graphics processor, and send the resource information of the graphics processor to the scheduler; the scheduler is further configured to acquire resource information of the graphics processor;

The resource information includes: the number of idle processing threads; and/or the state of the processing thread; the state is an occupied state or an idle state.

The scheduler specifically uses at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors according to state information of the to-be-processed task and/or resource information of the graphics processor The task processing time is determined according to the state information, and different target graphics processors are selected for different to-be-processed tasks for the plurality of to-be-processed tasks whose task processing time is greater than the time threshold.

The scheduler specifically uses at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors according to state information of the to-be-processed task and/or resource information of the graphics processor And determining the number of idle processing threads according to the resource information, and selecting a graphics processor with the largest idle processing thread from the plurality of graphics processors as the target graphics processor.

The scheduler specifically uses at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors according to state information of the to-be-processed task and/or resource information of the graphics processor And determining, by the resource information, an idle processing thread, and determining, by the graphics processor corresponding to the idle processing thread, a target graphics processor corresponding to the to-be-processed task.

In an example, when the scheduler selects at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors, specifically, the plurality of to-be-processed tasks for parallel processing are The plurality of to-be-processed tasks respectively select corresponding target graphics processors, and the target graphics processors corresponding to different to-be-processed tasks are the same or different.

When the scheduler selects at least one target graphics processor corresponding to the to-be-processed task, the scheduler is specifically configured to: if the processing queue is a static processing queue, from the multiple graphics The processor is configured to query the graphics processor corresponding to the static processing queue, and determine the queried processor as the target graphics processor corresponding to the to-be-processed task.

When the target graphics processor processes the to-be-processed task, the method is specifically configured to: if there is a task to be processed that has a higher priority than the to-be-processed task, interrupt the to-be-processed task, and process the priority A high to-be-processed task; after the processing of the higher-priority pending task is completed, the to-be-processed task is restored. The target graphics processor is configured to: when the task to be processed is abnormal, interrupt the to-be-processed task, and increase the priority of the to-be-processed task, and The processing task is cached to the processing queue. The target graphics processor is specifically configured to: parallel process a plurality of to-be-processed tasks allocated to the target graphics processor; wherein the parallel processing includes synchronous serial processing and kernel asynchronous processing.

When the target graphics processor processes the to-be-processed task, the method is specifically configured to: when processing the to-be-processed task, latching an address of the central processing unit, and transmitting, by using a DMA controller, the target graphics processor and Interaction data between central processors.

When the target graphics processor processes the to-be-processed task, the data processing is performed by using different processing threads when processing the to-be-processed task.

When the target graphics processor performs data sharing through different processing threads, it is specifically used for:

When the image processing thread processes the processing task, the output data is provided to the self-calibration processing thread, and the self-calibration processing thread processes the processing task according to the output data; or

Providing the output data to the feature point tracking processing thread, wherein the feature point tracking processing thread processes the processing task according to the output data; or

Providing output data to a semi-global stereo block matching processing thread, wherein the semi-global stereo block matching processing thread processes the processing task according to the output data; or

The output data is provided to a deep learning processing thread, and the deep learning processing thread processes the processing task according to the output data.

When the target graphics processor performs data sharing through different processing threads, the method is specifically configured to: when the semi-global stereo block matching processing thread processes the processing task, provide the output data to the point cloud tracking processing thread, and the point cloud tracking processing is performed. The thread processes the processing task according to the output data. When the target graphics processor performs data sharing through different processing threads, the specific data is used to: when the point cloud tracking processing thread processes the processing task, the output data is provided to the map processing thread; and the deep learning processing thread performs the processing task. During processing, the output data is provided to the map processing thread; the map processing thread processes the processing task according to the input data provided by the point cloud tracking processing thread and the input data provided by the deep learning processing thread.

Example 9:

Based on the same inventive concept as the above method, the embodiment of the present invention further provides a computer readable storage medium, where the computer readable storage medium stores a plurality of computer instructions, and when the computer instructions are executed, implementing the above claims Task processing method.

The system, apparatus, module or unit set forth in the above embodiments may be implemented by a computer chip or an entity, or by a product having a certain function. A typical implementation device is a computer, and the specific form of the computer may be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email transceiver, and a game control. A combination of a tablet, a tablet, a wearable device, or any of these devices.

For the convenience of description, the above devices are described separately by function into various units. Of course, the functions of the various units may be implemented in one or more software and/or hardware in the practice of the invention.

Those skilled in the art will appreciate that embodiments of the invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or a combination of software and hardware. Moreover, embodiments of the invention may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) including computer usable program code.

The present invention has been described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (system), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or FIG. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing device to produce a machine for the execution of instructions for execution by a processor of a computer or other programmable data processing device. Means for implementing the functions specified in one or more of the flow or in a block or blocks of the flow chart.

Moreover, these computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture comprising the instruction device. The instruction means implements the functions specified in one or more blocks of the flowchart or in a flow or block diagram of the flowchart.

These computer program instructions can also be loaded onto a computer or other programmable data processing device such that a series of operational steps are performed on a computer or other programmable device to produce computer-implemented processing for execution on a computer or other programmable device. The steps are provided to implement the functions specified in one or more blocks of the flowchart or in a block or blocks of the flowchart.

Those skilled in the art will appreciate that embodiments of the present invention can be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or a combination of software and hardware. Moreover, the invention can be embodied in the form of a computer program product embodied on one or more computer-usable storage media (which may include, but not limited to, disk storage, CD-ROM, optical storage, etc.) including computer usable program code.

The above is only the embodiments of the present invention and is not intended to limit the present invention. It will be apparent to those skilled in the art that various modifications and changes can be made in the present invention. Any modifications, equivalents, and improvements made within the spirit and scope of the invention are intended to be included within the scope of the appended claims.

Claims

A task processing method, characterized by being applied to a device including a plurality of graphics processors, the method comprising:

Select a pending task from the processing queue;

Selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors; assigning the to-be-processed task to the target graphics processor;

The to-be-processed task is processed by the target graphics processor.
The method of claim 1 wherein

The selecting a task to be processed from the processing queue includes:

If the processing queue is a static processing queue, a task to be processed whose time series and/or data has a dependency is selected from the static processing queue.
The method of claim 1 wherein

The selecting a task to be processed from the processing queue includes: selecting a time series and/or data independent pending task from the dynamic processing queue if the processing queue is a dynamic processing queue.
Method according to claim 2 or 3, characterized in that

Before the task to be processed is selected from the processing queue, the method further includes:

Receiving a pending task;

If the to-be-processed task is a static scheduling task whose time series and/or data has a dependency relationship, the to-be-processed task is cached into a static processing queue;

If the to-be-processed task is a time-series and/or data-independent dynamic scheduling task, the to-be-processed task is cached into a dynamic processing queue.
Method according to claim 2 or 3, characterized in that

Before the task to be processed is selected from the processing queue, the method further includes:

Receiving a task to be processed, and buffering the to-be-processed task to a task queue;

For the task to be processed in the task queue, if the to-be-processed task is a static scheduling task with a time series and/or data having a dependency, the task to be processed is cached into a static processing queue;

If the to-be-processed task is a time-series and/or data-independent dynamic scheduling task, the to-be-processed task is cached to a dynamic processing queue.
The method of claim 1 wherein

The selecting a task to be processed from the processing queue includes:

Obtaining a priority of the to-be-processed task in the processing queue;

A high priority pending task is selected from the processing queue based on the priority priority.
The method of claim 6 wherein:

The obtaining the priority of the to-be-processed task in the processing queue includes:

Obtaining a task type of the to-be-processed task in the processing queue;

Querying the mapping table by using the task type to obtain a priority corresponding to the task type;

The mapping table is used to record the correspondence between the task type and the priority.
The method according to claim 1, wherein the selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors comprises:

And selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors according to status information of the to-be-processed task and/or resource information of the graphics processor.
The method according to claim 8, wherein the selecting and the to-be-processed task are selected from the plurality of graphics processors according to status information of the to-be-processed task and/or resource information of a graphics processor Before corresponding to at least one target graphics processor, the method further includes:

Obtaining a task type of the to-be-processed task;

Obtaining state information corresponding to the task type by using the task type to query the state information table, where the state information table is used to record a correspondence between the task type and the state information;

The obtained status information is determined as status information of the to-be-processed task.
The method according to claim 9, wherein before the obtaining the status information corresponding to the task type by using the task type to query the status information table, the method further includes:

The status information of the to-be-processed task that has been processed is obtained, and the correspondence between the task type of the to-be-processed task and the status information of the to-be-processed task is recorded in the status information table.
The method according to claim 8, wherein the status information comprises a task processing time; the task processing time is a time difference between the task completion time and the task receiving time.
The method according to claim 8, wherein the selecting and the to-be-processed task are selected from the plurality of graphics processors according to status information of the to-be-processed task and/or resource information of a graphics processor Before corresponding to at least one target graphics processor, the method further includes:

Obtaining resource information of the graphics processor; wherein the resource information includes: a number of idle processing threads; and/or a state of processing a thread; the state is an occupied state or an idle state.
The method according to claim 8, wherein the selecting and the to-be-processed task are selected from the plurality of graphics processors according to status information of the to-be-processed task and/or resource information of a graphics processor Corresponding at least one target graphics processor, including:

The task processing time is determined according to the state information, and for the plurality of to-be-processed tasks whose task processing time is greater than the time threshold, different target graphics processors are selected for different to-be-processed tasks.
The method according to claim 8, wherein the selecting and the to-be-processed task are selected from the plurality of graphics processors according to status information of the to-be-processed task and/or resource information of a graphics processor Corresponding at least one target graphics processor, including:

Determining the number of idle processing threads according to the resource information, and selecting a graphics processor with the largest idle processing thread from the plurality of graphics processors as the target graphics processor.
The method according to claim 8, wherein the selecting and the to-be-processed task are selected from the plurality of graphics processors according to status information of the to-be-processed task and/or resource information of a graphics processor Corresponding at least one target graphics processor, including:

Determining an idle processing thread according to the resource information, and determining a graphics processor corresponding to the idle processing thread as a target graphics processor corresponding to the to-be-processed task.
The method according to claim 1, wherein the selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors comprises:

For the plurality of to-be-processed tasks to be processed in parallel, the corresponding target graphics processors are respectively selected for the plurality of to-be-processed tasks, and the target graphics processors corresponding to the different to-be-processed tasks are the same or different.
The method according to claim 1, wherein the selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors comprises:

If the processing queue is a static processing queue, query a graphics processor corresponding to the static processing queue from the plurality of graphics processors, and determine the queried graphics processor as corresponding to the to-be-processed task Target graphics processor.
The method of claim 1, wherein when the object to be processed is processed by the target graphics processor, the method further comprises:

If there is a task to be processed that has a higher priority than the to-be-processed task, interrupting the to-be-processed task, and processing the higher-priority to-be-processed task by the target graphics processor;

After the processing of the higher priority pending task is completed, the to-be-processed task is restored.
The method of claim 1, wherein when the object to be processed is processed by the target graphics processor, the method further comprises:

And if the pending task is abnormal, interrupting the to-be-processed task, and increasing a priority of the to-be-processed task, and buffering the to-be-processed task into the processing queue.
The method according to claim 1, wherein the processing the to-be-processed task by the target graphics processor comprises:

And processing, by the target graphics processor, a plurality of to-be-processed tasks allocated to the target graphics processor; wherein the parallel processing comprises synchronous serial processing and kernel asynchronous processing.
The method according to claim 1, wherein the processing the to-be-processed task by the target graphics processor comprises:

When processing the to-be-processed task, the address of the central processing unit is latched, and the interaction data between the target graphics processor and the central processing unit is transmitted by the DMA controller.
The method according to claim 1, wherein the processing the to-be-processed task by the target graphics processor comprises:

When the pending task is processed, data sharing is performed through different processing threads.
The method according to claim 22, wherein said sharing data by different processing threads comprises:

When the image processing thread processes the processing task, the output data is provided to the self-calibration processing thread, and the self-calibration processing thread processes the processing task according to the output data; or

Providing the output data to the feature point tracking processing thread, wherein the feature point tracking processing thread processes the processing task according to the output data; or

Providing the output data to the semi-global stereo block matching processing thread, where the semi-global stereo block matching processing thread processes the processing task according to the output data; or

The output data is provided to a deep learning processing thread, and the deep learning processing thread processes the processing task according to the output data.
The method according to claim 22, wherein said sharing data by different processing threads comprises: providing output data to point cloud tracking processing when the semi-global stereo block matching processing thread processes the processing task a thread, by which the point cloud tracking processing thread processes the processing task according to the output data.
The method according to claim 22, wherein said sharing data by different processing threads comprises:

When the point cloud tracking processing thread processes the processing task, the output data is provided to the map processing thread; when the deep learning processing thread processes the processing task, the output data is provided to the map processing thread; the map processing thread according to the The point cloud tracking processing thread provides input data and input data provided by the deep learning processing thread to process the processing task.
A task processing device, comprising: a scheduler and a plurality of graphics processors;

The scheduler is configured to select a task to be processed from a processing queue, and select at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors, and allocate the to-be-processed task Giving the target graphics processor;

The target graphics processor is configured to process the to-be-processed task.
The device according to claim 26, wherein the scheduler selects a task to be processed from the processing queue, specifically: if the processing queue is a static processing queue, selecting a time series from the static processing queue / or data to be processed with dependencies.
The device according to claim 26, wherein the scheduler selects a task to be processed from the processing queue, specifically: if the processing queue is a dynamic processing queue, selecting a time series from the dynamic processing queue / or data independent pending tasks.
Device according to claim 27 or 28, characterized in that

The scheduler is further configured to receive a task to be processed; if the task to be processed is a static scheduling task with a time series and/or data having a dependency, the task to be processed is cached into a static processing queue; The to-be-processed task is a time-series and/or data-independent dynamic scheduling task, and the to-be-processed task is cached into a dynamic processing queue.
The device according to claim 27 or 28, wherein the scheduler is further configured to receive a task to be processed, and cache the to-be-processed task to a task queue; The to-be-processed task is a static scheduling task whose time series and/or data has a dependency, and caches the to-be-processed task to a static processing queue; if the to-be-processed task is a time-series and/or data-independent dynamic scheduling task, The to-be-processed task is cached to a dynamic processing queue.
The device according to claim 26, wherein the scheduler selects a task to be processed from a processing queue, specifically: acquiring a priority of a task to be processed in the processing queue; prioritizing based on the priority A high priority pending task is selected from the processing queue.
The device according to claim 31, wherein the scheduler obtains a priority of a task to be processed in the processing queue, and is configured to: acquire a task type of a task to be processed in the processing queue; The task type query mapping table obtains a priority corresponding to the task type; wherein the mapping table is used to record a correspondence between a task type and a priority.
The device according to claim 26, wherein the scheduler selects at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors, specifically for: according to the waiting Processing status information of the task and/or resource information of the graphics processor, and selecting at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors.
The device according to claim 33, wherein the scheduler is further configured to acquire a task type of the task to be processed, and query the state information table by using the task type to obtain state information corresponding to the task type. The status information table is used to record the correspondence between the task type and the status information; and the obtained status information is determined as the status information of the to-be-processed task.
The device according to claim 34, further comprising:

a monitor, configured to acquire status information of the to-be-processed task that has been processed, and record a correspondence between the task type of the to-be-processed task and the status information of the to-be-processed task in the status information table, and the status is The information table is sent to the scheduler.
The device according to claim 33, wherein said status information comprises a task processing time; and the task processing time is a time difference between the task completion time and the task receiving time.
The device according to claim 33, further comprising: a monitor, configured to acquire resource information of the graphics processor, and send resource information of the graphics processor to the scheduler;

The scheduler is further configured to acquire resource information of the graphics processor;

The resource information includes: the number of idle processing threads; and/or the state of the processing thread; the state is an occupied state or an idle state.
The device according to claim 33, wherein the scheduler selects and waits from the plurality of graphics processors according to status information of the to-be-processed task and/or resource information of a graphics processor. When processing at least one target graphics processor corresponding to the task, it is specifically used to:

The task processing time is determined according to the state information, and for the plurality of to-be-processed tasks whose task processing time is greater than the time threshold, different target graphics processors are selected for different to-be-processed tasks.
The device according to claim 33, wherein the scheduler selects and waits from the plurality of graphics processors according to status information of the to-be-processed task and/or resource information of a graphics processor. When processing at least one target graphics processor corresponding to the task, it is specifically used to:

Determining the number of idle processing threads according to the resource information, and selecting a graphics processor with the largest idle processing thread from the plurality of graphics processors as the target graphics processor.
The device according to claim 33, wherein the scheduler selects and waits from the plurality of graphics processors according to status information of the to-be-processed task and/or resource information of a graphics processor. When processing at least one target graphics processor corresponding to the task, it is specifically used to:

Determining an idle processing thread according to the resource information, and determining a graphics processor corresponding to the idle processing thread as a target graphics processor corresponding to the to-be-processed task.
The device according to claim 26, wherein the scheduler selects at least one target graphics processor corresponding to the to-be-processed task from the plurality of graphics processors, specifically for: parallel processing The plurality of to-be-processed tasks respectively select corresponding target graphics processors for the plurality of to-be-processed tasks, and the target graphics processors corresponding to the different to-be-processed tasks are the same or different.
The device according to claim 26, wherein

When the scheduler selects at least one target graphics processor corresponding to the to-be-processed task, the scheduler is specifically configured to: if the processing queue is a static processing queue, from the multiple graphics The processor is configured to query the graphics processor corresponding to the static processing queue, and determine the queried processor as the target graphics processor corresponding to the to-be-processed task.
The device according to claim 26, wherein the target graphics processor is configured to: when the task to be processed is processed, if there is a task to be processed having a higher priority than the task to be processed, interrupting And the task to be processed is processed, and the task to be processed with a higher priority is processed; after the processing of the higher priority task is completed, the task to be processed is restored.
The device according to claim 26, wherein

The target graphics processor is configured to: when the task to be processed is abnormal, interrupt the to-be-processed task, and increase the priority of the to-be-processed task, and The processing task is cached into the processing queue.
The device according to claim 26, wherein the target graphics processor is configured to: process, in parallel, a plurality of to-be-processed tasks assigned to the target graphics processor when processing the to-be-processed task; Parallel processing includes synchronous serial processing and kernel asynchronous processing.
The device according to claim 26, wherein

When the target graphics processor processes the to-be-processed task, the method is specifically configured to: when processing the to-be-processed task, latching an address of the central processing unit, and transmitting, by using a DMA controller, the target graphics processor and Interaction data between central processors.
The device according to claim 26, wherein

When the target graphics processor processes the to-be-processed task, the data processing is performed by using different processing threads when processing the to-be-processed task.
The device according to claim 47, wherein

When the target graphics processor performs data sharing through different processing threads, it is specifically used for:

When the image processing thread processes the processing task, the output data is provided to the self-calibration processing thread, and the self-calibration processing thread processes the processing task according to the output data; or

Providing the output data to the feature point tracking processing thread, wherein the feature point tracking processing thread processes the processing task according to the output data; or

Providing the output data to the semi-global stereo block matching processing thread, where the semi-global stereo block matching processing thread processes the processing task according to the output data; or

The output data is provided to a deep learning processing thread, and the deep learning processing thread processes the processing task according to the output data.
The device according to claim 47, wherein the target graphics processor performs data sharing through different processing threads, and is specifically configured to: when the semi-global stereo block matching processing thread processes the processing task, output data Provided to the point cloud tracking processing thread, the point cloud tracking processing thread processes the processing task according to the output data.
The device according to claim 47, wherein

When the target graphics processor performs data sharing through different processing threads, it is specifically used for:

When the point cloud tracking processing thread processes the processing task, the output data is provided to the map processing thread; when the deep learning processing thread processes the processing task, the output data is provided to the map processing thread; the map processing thread according to the point The cloud traces the input data provided by the processing thread and the input data provided by the deep learning processing thread to process the processing task.
A computer readable storage medium, wherein the computer readable storage medium stores a plurality of computer instructions, and when the computer instructions are executed, the task processing method of any one of claims 1-25 is implemented.