CN-121985108-A - GPU hardware decoding fault detection method and device, electronic equipment and storage medium

CN121985108ACN 121985108 ACN121985108 ACN 121985108ACN-121985108-A

Abstract

The disclosure provides a method and a device for detecting decoding faults of GPU hardware, electronic equipment and a storage medium, and relates to the technical field of GPU testing. The method comprises the steps of collecting audio output signals of an audio and video under the condition that the audio and video adopts a hardware decoding function of a graphics processor GPU, carrying out audio anomaly analysis on a frequency spectrum of the audio output signals to obtain an audio analysis result, determining video state data corresponding to video frames in the audio and video, wherein the video state data represent video processing stability of the video frames in a GPU hardware decoding process, and carrying out fault location on the GPU hardware decoding process based on the audio analysis result and the video state data to determine GPU hardware decoding fault points. According to the scheme, the fault location can be accurately carried out on the GPU hardware decoding process, and further the GPU hardware decoding fault point can be accurately determined.

Inventors

Request for anonymity

Assignees

摩尔线程智能科技(上海)有限责任公司

Dates

Publication Date: 20260505
Application Date: 20251223

Claims (14)

1. A method for detecting a decoding failure of GPU hardware, the method comprising: Under the condition that the audio and video adopts the hardware decoding function of the graphic processor GPU, acquiring an audio output signal of the audio and video; performing audio anomaly analysis on the frequency spectrum of the audio output signal to obtain an audio analysis result; determining video state data corresponding to video frames in the audio and video, wherein the video state data is used for representing video processing stability of the video frames in a GPU hardware decoding process; And based on the audio analysis result and the video state data, performing fault positioning on the GPU hardware decoding process, and determining GPU hardware decoding fault points.
2. The GPU hardware decoding failure detection method of claim 1, wherein the performing audio anomaly analysis on the spectrum of the audio output signal to obtain an audio analysis result comprises: Determining a target spectral energy of the spectrum from the spectrum of the audio output signal; And comparing the target spectrum energy with preset reference spectrum energy to obtain the audio analysis result.
3. The method of claim 2, wherein the target spectral energy is characterized by a first spectral energy within a target frequency band, wherein determining the target spectral energy of the spectrum from the spectrum of the audio output signal comprises: for at least one target frequency band within a spectral range of the audio output signal, acquiring first spectral energy within the target frequency band; The step of comparing the target spectrum energy with preset reference spectrum energy to obtain the audio analysis result comprises the following steps: And obtaining second spectrum energy of the reference audio signal in the target frequency band, and comparing the first spectrum energy and the second spectrum energy corresponding to at least one target frequency band to obtain the audio analysis result.
4. The GPU hardware decoding failure detection method according to claim 3, wherein comparing the first spectral energy and the second spectral energy corresponding to at least one target frequency band to obtain the audio analysis result comprises: determining a ratio between the first spectral energy and the second spectral energy for at least one of the target frequency bands; if the ratio is smaller than the energy fluctuation lower limit threshold or larger than the energy fluctuation upper limit threshold, judging that the audio analysis sub-result of the target frequency band is abnormal in frequency spectrum; if the ratio is greater than or equal to the energy fluctuation lower limit threshold and less than or equal to the energy fluctuation upper limit threshold, judging that the audio analysis sub-result of the target frequency band is normal in frequency spectrum; And determining the audio analysis result according to the audio analysis sub-result of each target frequency band.
5. The GPU hardware decoding failure detection method of claim 4, wherein the spectral anomalies include one of a frequency component missing anomaly and an energy lifting anomaly; If the ratio is smaller than the lower energy fluctuation threshold or larger than the upper energy fluctuation threshold, determining that the audio analysis sub-result of the target frequency band is abnormal in frequency spectrum, including: if the ratio is smaller than the energy fluctuation lower limit threshold, judging that the audio analysis sub-result of the target frequency band is abnormal in frequency component deletion; And if the ratio is larger than the energy fluctuation upper limit threshold, judging that the audio analysis sub-result of the target frequency band is abnormal in energy lifting.
6. The GPU hardware decoding failure detection method of claim 4, wherein the determining the audio analysis result from the audio analysis sub-results for each of the target frequency bands comprises: If the audio analysis sub-results of each target frequency band are all frequency band normal, determining that the audio analysis results are frequency spectrum normal; if at least one audio analysis sub-result of the target frequency band is abnormal, determining that the audio analysis result is abnormal in frequency spectrum.
7. The GPU hardware decoding failure detection method of claim 4, wherein the performing failure localization on the GPU hardware decoding process based on the audio analysis result and the video status data, determining a GPU hardware decoding failure point, comprises: If the audio analysis result is that the frequency spectrum is abnormal, judging that the GPU hardware decoding process has faults; If the audio analysis result is that the frequency spectrum is normal and the video state data indicates that the video output signal is stuck, the GPU hardware decoding process is judged to be normal, and the video processing link is judged to be faulty.
8. The method for detecting a failure in decoding GPU hardware according to claim 7, wherein the hardware decoding function is implemented by a hardware structure including a video decoding unit and an audio decoding unit, and wherein if the audio analysis result is a spectrum anomaly, determining that the GPU hardware decoding process has a failure comprises: If the audio analysis result is that the frequency spectrum is abnormal and the video state data represents that the video decoding is in an unstable state, judging that a video decoding unit in the GPU hardware decoding process has a fault; and if the audio analysis result is that the frequency spectrum is abnormal and the video state data represents that the video decoding is in a stable state, judging that an audio decoding unit in the GPU hardware decoding process has a fault.
9. The GPU hardware decoding failure detection method of claim 2, wherein the determining the reference spectral energy comprises: Calling a pre-constructed baseline model, and acquiring frequency domain statistical characteristics of the baseline model, wherein the baseline model is constructed based on a reference audio signal without faults; a reference spectral energy is determined based on the frequency domain statistics.
10. The GPU hardware decoding failure detection method according to any one of claims 1 to 9, wherein the acquiring the audio output signal of the audio-video comprises: and collecting the audio output signals of the audio and video at a final node of the audio rendering flow, wherein the audio output signals are pulse code modulation format audio signals.
11. The GPU hardware decoding failure detection method according to any one of claims 1-9, wherein the determining video state data corresponding to a video frame in the audio-video includes: Calling a graphic rendering interface through an interception application program, and extracting context information of at least one video frame in the audio and video; and acquiring the hardware decoding state of the GPU associated with at least one video frame, and determining video state data corresponding to the video frame according to the hardware decoding state and the context information associated with at least one video frame.
12. A GPU hardware decoding failure detection apparatus, comprising: The audio/video signal acquisition module is used for acquiring audio output signals of the audio/video under the condition that the audio/video adopts the hardware decoding function of the GPU; The audio frequency spectrum analysis module is used for carrying out audio frequency anomaly analysis on the frequency spectrum of the audio output signal to obtain an audio frequency analysis result; The video state data determining module is used for determining video state data corresponding to video frames in the audio and video, and the video state data represents video processing stability of the video frames in the GPU hardware decoding process; And the fault positioning module is used for performing fault positioning on the GPU hardware decoding process based on the audio analysis result and the video state data, and determining GPU hardware decoding fault points.
13. An electronic device, comprising: processor, and A memory having stored thereon computer readable instructions that when executed by the processor implement a GPU hardware decoding failure detection method as in any of claims 1 to 11.
14. A computer readable storage medium, having stored thereon a computer program which, when executed by a processor, implements a GPU hardware decoding failure detection method as claimed in any of claims 1 to 11.

Description

GPU hardware decoding fault detection method and device, electronic equipment and storage medium Technical Field The disclosure relates to the technical field of GPU testing, in particular to a GPU hardware decoding fault detection method, a device, an electronic device and a storage medium. Background With the rapid development of high-definition video, immersive media and real-time streaming media services, the hardware decoding technology of the graphics processor GPU (Graphics Processing Unit) decodes high-resolution and high-bitrate video streams with high efficiency through a special hardware unit, and becomes the core supporting capability of a modern multimedia playing system. With the popularity of high resolution, high rate video content, GPU hardware decoding has become a standard configuration for player software. However, in the process of decoding GPU hardware, the stability problem of the GPU is increasingly prominent, so that phenomena such as blocking, screen-printing, asynchronous sound and picture, sound explosion and the like occur in the playing process of the audio and video. The traditional GPU hardware decoding fault detection method mainly relies on monitoring performance counters (such as decoder utilization rate, frame processing time consumption, temperature and the like) exposed by a GPU driving layer, and indirectly deducing by combining peripheral indexes such as network throughput, system load and the like, so that fault positioning cannot be performed on GPU hardware decoding accurately. It should be noted that the information disclosed in the above background section is only for enhancing understanding of the background of the present disclosure and thus may include information that does not constitute prior art known to those of ordinary skill in the art. Disclosure of Invention The embodiment of the disclosure aims to provide a GPU hardware decoding fault detection method, a device, electronic equipment and a storage medium, which can accurately locate faults in a GPU hardware decoding process and further accurately determine GPU hardware decoding fault points. Other features and advantages of the present disclosure will be apparent from the following detailed description, or may be learned in part by the practice of the disclosure. According to a first aspect of an embodiment of the present disclosure, there is provided a GPU hardware decoding failure detection method, the method including: Under the condition that the audio and video adopts the hardware decoding function of the graphic processor GPU, acquiring an audio output signal of the audio and video; performing audio anomaly analysis on the frequency spectrum of the audio output signal to obtain an audio analysis result; determining video state data corresponding to video frames in the audio and video, wherein the video state data represents video processing stability of the video frames in a GPU hardware decoding process; And based on the audio analysis result and the video state data, performing fault positioning on the GPU hardware decoding process, and determining GPU hardware decoding fault points. In some example embodiments of the present disclosure, based on the foregoing solution, the performing audio anomaly analysis on the spectrum of the audio output signal to obtain an audio analysis result includes: Determining a target spectral energy of the spectrum from the spectrum of the audio output signal; And comparing the target spectrum energy with preset reference spectrum energy to obtain the audio analysis result. In some example embodiments of the present disclosure, based on the foregoing, the target spectral energy is characterized by a first spectral energy within a target frequency band, the determining the target spectral energy of the spectrum from the spectrum of the audio output signal comprises: for at least one target frequency band within a spectral range of the audio output signal, acquiring first spectral energy within the target frequency band; The step of comparing the target spectrum energy with preset reference spectrum energy to obtain the audio analysis result comprises the following steps: And obtaining second spectrum energy of the reference audio signal in the target frequency band, and comparing the first spectrum energy and the second spectrum energy corresponding to at least one target frequency band to obtain the audio analysis result. In some example embodiments of the present disclosure, based on the foregoing solution, the comparing the first spectral energy and the second spectral energy corresponding to at least one target frequency band to obtain the audio analysis result includes: Determining a ratio between the first spectral energy and the second spectral energy for at least one of the target frequency bands; if the ratio is smaller than the energy fluctuation lower limit threshold or larger than the energy fluctuation upper limit threshold, judging that the audio anal