Dynamic assignment of down sampling intervals for data stream processing

Patent metadata

Field	Value
Publication number	US-11798198-B2
Application number	US-202318152643-A
Country	US
Kind code	B2
Filing date	Jan 10, 2023
Priority date	Dec 30, 2017
Publication date	Oct 24, 2023
Grant date	Oct 24, 2023

Abstract

Embodiments are generally directed to compression in machine learning and deep learning processing. An embodiment of an apparatus for compression of untyped data includes a graphical processing unit (GPU) including a data compression pipeline, the data compression pipeline including a data port coupled with one or more shader cores, wherein the data port is to allow transfer of untyped data without format conversion, and a 3D compression/decompression unit to provide for compression of untyped data to be stored to a memory subsystem and decompression of untyped data from the memory subsystem.

First claim

What is claimed is:

1. An apparatus comprising:
one or more processors including a graphical processing unit (GPU); and
a memory to store data, including data for processing by the GPU;
wherein the GPU is to:
sample one or more data streams,
evaluate each of the one or more data streams for homogeneity and parameter relationship,
determine a score for each of the one or more data streams, the score for each data stream representing how much the data stream is varying, and
dynamically assign a down sampling interval for each of the one or more data streams based at least in part on the determined score for the respective data stream.

2. The apparatus of claim 1 , wherein evaluating the one or more data streams includes comparing data within each data stream of the one or more data streams.

3. The apparatus of claim 1 , wherein dynamically assigning a down sampling interval for each of the one or more data streams includes assigning a relatively longer down sampling interval to a first data stream having a score indicating a higher level of homogeneity, and assigning a relatively shorter down sampling interval to a second data stream having a indicating a lower level of homogeneity in data.

4. The apparatus of claim 1 , wherein the GPU includes a data sampling optimizer, the data sampling optimizer including:
a first portion to evaluate and score the one or more data streams regarding the homogeneity and parameter relationship for each data stream; and
a second portion to dynamically establish the down sampling interval for each data stream based on the score for each data stream.

5. The apparatus of claim 4 , wherein the data sampling optimizer further includes:
a feedback loop to adjust the down sampling intervals for the one or more data streams as the one or more data streams are sampled.

6. The apparatus of claim 4 , wherein operation of the data sampling optimizer is tunable to select how the assignment of down sampling intervals is applied, whether the assignment of down sampling intervals is enabled or disabled, or both.

7. The apparatus of claim 1 , wherein sampling the one or more data streams includes sampling of machine learning or deep learning data.

8. The apparatus of claim 1 , wherein the one or more data streams include one or more sensor streams.

9. A non-transitory computer-readable storage medium having stored thereon data representing sequences of instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
receiving one or more data streams for processing by a graphical processing unit (GPU);
sampling the one or more data streams;
evaluating each of the one or more data streams for homogeneity and parameter relationship;
determining a score for each of the one or more data streams, the score for each data stream representing how much the data stream is varying; and
dynamically assigning a down sampling interval for each of the one or more data streams based at least in part on the determined score for the respective data stream.

10. The medium of claim 9 , wherein evaluating the one or more data streams includes comparing data within each data stream of the one or more data streams.

11. The medium of claim 9 , wherein dynamically assigning a down sampling interval for each of the one or more data streams includes assigning a relatively longer down sampling interval to a first data stream having a score indicating a higher level of homogeneity, and assigning a relatively shorter down sampling interval to a second data stream having a indicating a lower level of homogeneity in data.

12. The medium of claim 9 , wherein sampling the one or more data streams includes sampling of machine learning or deep learning data.

13. The medium of claim 9 , wherein the one or more data streams include one or more sensor streams.

14. A processing system comprising:
a central processing unit (CPU);
one or more graphical processing units (GPUs) including a first GPU; and
a memory to store data, including data for machine learning processing by at least the first GPU;
wherein the first GPU is to:
sample one or more data streams,
evaluate each of the one or more data streams for homogeneity and parameter relationship,
determine a score for each of the one or more data streams, the score for each data stream representing how much the data stream is varying, and
dynamically assign a down sampling interval for each of the one or more data streams based at least in part on the determined score for the respective data stream.

15. The processing system of claim 14 , wherein evaluating the one or more data streams includes comparing data within each data stream of the one or more data streams.

16. The processing system of claim 14 , wherein dynamically assigning a down sampling interval for each of the one or more data streams includes assigning a relatively longer down sampling interval to a first data stream having a score indicating a higher level of homogeneity, and assigning a relatively shorter down sampling interval to a second data stream having a indicating a lower level of homogeneity in data.

17. The processing system of claim 14 , wherein the first GPU includes a data sampling optimizer, the data sampling optimizer including:
a first portion to evaluate and score the one or more data streams regarding the homogeneity and parameter relationship for each data stream; and
a second portion to dynamically establish the down sampling interval for each data stream based on the score for each data stream.

18. The processing system of claim 17 , wherein the data sampling optimizer further includes:
a feedback loop to adjust the down sampling intervals for the one or more data streams as the one or more data streams are sampled.

19. The processing system of claim 17 , wherein operation of the data sampling optimizer is tunable to select how the assignment of down sampling intervals is applied, whether the assignment of down sampling intervals is enabled or disabled, or both.

20. The processing system of claim 14 , wherein the one or more data streams include one or more sensor streams.

Assignees

• Intel Corp

Inventors

• Ray Joydeep
• Ashbaugh Ben
• Surti Prasoonkumar
• Ramani Pradeep
• Harihara Rama
• Justin Jerin C
• Huang Jing
• Cui Xiaoming
• Costa Timothy B
• Gong Ting
• Ould-Ahmed-Vall Elmoustapha
• Balasubramanian Kumar
• Thomas Anil
• Elibol Oguz H
• Bobba Jayaram
• Zhuang Guozhong
• Subramanian Bhavani
• Keskin Gokce
• Sakthivel Chandrasekaran
• Poornachandran Rajesh

Classifications

• G06T9/002 Primary

  using neural networks · CPC title

• G06F12/023 Primary

  Free address space management · CPC title

• G06T15/005

  General purpose rendering architectures · CPC title

• G06F2212/302

  In image processor or graphics adapter · CPC title

• G06F2212/401

  Compressed data · CPC title