Tutorial for parallel_for_ and Universal Intrinsic (GSoC '21) by r0hit05 · Pull Request #20361 · opencv/opencv

r0hit05 · 2021-07-05T16:28:45Z

This pull request is a part of the Google Summer of Code 2021 project. Here is an overview of the project:

Tutorial for parallel_for_:
- Updated the parallel_for_ tutorial. The tutorial includes a simple function and benchmarks it's performance after using parallel_for_
- Files Updated:
  - Added: how_to_use_OpenCV_parallel_for_new.markdown
  - Added: how_to_use_OpenCV_parallel_for_new.cpp
Tutorial for Universal Intrinsic:
- Added a new tutorial for Universal Intrinsic. The tutorial consists of two parts:
  - Basics of how to use universal intrinsic
  - Demonstration on how to vectorize convolution
- Files Updated:
  - Added: univ_intrin.markdown
  - Added: univ_intrin.cpp

Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

I agree to contribute to the project under Apache 2 License.
To the best of my knowledge, the proposed patch is not based on a code under GPL or other license that is incompatible with OpenCV
The PR is proposed to proper branch
There is reference to original bug report and related work
There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
The feature is well documented and sample code can be built with the project CMake

force_builders=linux,docs

doc/tutorials/core/univ_intrin/univ_intrin.markdown

...pp/tutorial_code/core/how_to_use_OpenCV_parallel_for_/how_to_use_OpenCV_parallel_for_new.cpp

doc/tutorials/core/univ_intrin/univ_intrin.markdown

* Added first half of universal intrinsic tutorial * Fixed warnings in documentation and sample code for parallel_for_new tutorial * Restored original parallel_for_ tutorial and table_of_content_core * Minor changes

* Minor changes in vectorized implementation of 1-D and 2-D convolution

doc/tutorials/core/univ_intrin/univ_intrin.markdown

terfendail · 2021-08-21T11:17:35Z

Could you please add new tutorials to the table of content? I mean table_of_content_core.markdown

r0hit05 · 2021-08-21T14:42:09Z

Could you please add new tutorials to the table of content? I mean table_of_content_core.markdown

Do I remove the old parallel_for_ tutorial from the table of contents? Also, the new parallel_for_ tutorial has the 'new' suffix in the name. Do I need to change that?

terfendail · 2021-08-22T20:47:18Z

I think we should start with both tutorials. Old one could be removed later

…le of contents

...utorials/core/how_to_use_OpenCV_parallel_for_new/how_to_use_OpenCV_parallel_for_new.markdown

...pp/tutorial_code/core/how_to_use_OpenCV_parallel_for_/how_to_use_OpenCV_parallel_for_new.cpp

...utorials/core/how_to_use_OpenCV_parallel_for_new/how_to_use_OpenCV_parallel_for_new.markdown

doc/tutorials/core/univ_intrin/univ_intrin.markdown

terfendail · 2021-08-23T15:38:45Z

I think tutorials are finished so turn the PR to "ready for review"

alalek · 2021-08-30T18:57:31Z

@r0hit2005 Need to fix build errors:

/build/precommit_macosx/opencv/samples/cpp/tutorial_code/core/univ_intrin/univ_intrin.cpp:131:27: error: variable-sized object may not be initialized
                float ans[cols] = {0};
                          ^~~~
/build/precommit_macosx/opencv/samples/cpp/tutorial_code/core/univ_intrin/univ_intrin.cpp:181:15: error: variable-sized object may not be initialized
    float ans[vsrc.cols] = {0};
              ^~~~~~~~~
2 errors generated.

C:\build\precommit_opencl\opencv\samples\cpp\tutorial_code\core\univ_intrin\univ_intrin.cpp(131,27): error C2131: expression did not evaluate to a constant

"variable-sized arrays" is GNU extension and not a part of ISO C++11.

alalek · 2021-08-31T05:01:11Z

Please use static_cast<T>() or cv::saturate_cast<T>() to resolve these warnings:

C:\build\precommit_windows64\opencv\samples\cpp\tutorial_code\core\univ_intrin\univ_intrin.cpp(11): warning C4244: 'initializing': conversion from 'double' to 'int', possible loss of data [C:\build\precommit_windows64\build\samples\cpp\example_tutorial_univ_intrin.vcxproj]
C:\build\precommit_windows64\opencv\samples\cpp\tutorial_code\core\univ_intrin\univ_intrin.cpp(63): warning C4244: '=': conversion from 'double' to 'uchar', possible loss of data [C:\build\precommit_windows64\build\samples\cpp\example_tutorial_univ_intrin.vcxproj]

alalek · 2021-09-04T18:27:39Z

...utorials/core/how_to_use_OpenCV_parallel_for_new/how_to_use_OpenCV_parallel_for_new.markdown

+In OpenCV 4.5, the following parallel frameworks are available in that order:
+
+*   Intel Threading Building Blocks (3rdparty library, should be explicitly enabled)
+*   C= Parallel C/C++ Programming Language Extension (3rdparty library, should be explicitly enabled)


C= Parallel C/C++ Programming Language Extension (3rdparty library, should be explicitly enabled)

It is deprecated and dropped.

alalek · 2021-09-04T18:28:52Z

...utorials/core/how_to_use_OpenCV_parallel_for_new/how_to_use_OpenCV_parallel_for_new.markdown

+![Convolution Animation](images/convolution-example-matrix.gif)
+
+
+For more information about different kernels and what they do, look [here](https://docs.opencv.org/4.5.2/d7/da8/tutorial_table_of_content_imgproc.html).


https://docs.opencv.org/...

Don't put direct links on docs site.
Use Doxygen references instead.

alalek · 2021-09-04T18:30:43Z

...utorials/core/how_to_use_OpenCV_parallel_for_new/how_to_use_OpenCV_parallel_for_new.markdown

+Based on that, we can broadly classify algorithms into two categories:-
+1. Algorithms in which only a single thread writes data to a particular memory location.
+    * In *convolution*, for example, even though multiple  threads may read from a pixel at a particular time, only a single thread *writes* to a particular pixel.
+<!-- <br> -->




needed?

alalek · 2021-09-04T18:32:00Z

...utorials/core/how_to_use_OpenCV_parallel_for_new/how_to_use_OpenCV_parallel_for_new.markdown

+
+In the tutorial, we used a horizontal gradient filter(as shown in the animation above), which produces an image highlighting the vertical edges.
+
+![result image](images/resimg.png)


resimg.png

We don't need lossless format for documentation. Please use .jpg instead.

alalek · 2021-09-04T18:32:46Z

...utorials/core/how_to_use_OpenCV_parallel_for_new/how_to_use_OpenCV_parallel_for_new.markdown

+
+        Sequential Implementation: 0.0953564s
+        Parallel Implementation: 0.0246762s
+        Parallel Implentatation(Row Split): 0.0248722s


Implentatation

typo

alalek · 2021-09-04T18:33:29Z

doc/tutorials/core/univ_intrin/univ_intrin.markdown

+
+The goal of this tutorial is to provide a guide to using the Universal Intrinsics feature to vectorize your C++ code for a faster runtime.
+We'll briefly look into _SIMD intrinsics_ and how to work with wide _registers_, followed by a tutorial on the basic operations using wide registers.
+The tutorial will only demonstrate basic operations. To know more about Universal Intrinsics, visit the [documentation](https://docs.opencv.org/4.5.3/df/d91/group__core__hal__intrin.html).


https://docs.opencv.org/...

ditto

How do I refer to the documentation?

Use @ref <id>, where <id>:

is a tutorial identifier (see the first line of tutorial page)

or identifier of code group, for this case it is here

alalek · 2021-09-04T18:34:44Z

...pp/tutorial_code/core/how_to_use_OpenCV_parallel_for_/how_to_use_OpenCV_parallel_for_new.cpp

+namespace
+{
+    //! [convolution-sequential]
+    void conv_seq(Mat src, Mat &dst, Mat kernel)
+    {
+        //![convolution-make-borders]
+        int rows = src.rows, cols = src.cols;
+        dst = Mat(rows, cols, src.type());


Please avoid indentation in namespaces

r0hit05 · 2021-09-07T14:43:44Z

Should I change the previous tutorial(file_input_output_with_xml_yml) to point to the new parallel_for tutorial?

alalek · 2021-09-07T15:34:36Z

Yes, please keep next/prev links up-to-date.

terfendail · 2021-09-13T13:16:30Z

As far as I could understand resimg.png was just renamed. Have you committed converted version?

Tutorial for parallel_for_ and Universal Intrinsic (GSoC '21) * New parallel_for tutorial * Universal Intrinsics Draft Tutorial * Added draft of universal intrinsic tutorial * * Added final markdown for parallel_for_new * Added first half of universal intrinsic tutorial * Fixed warnings in documentation and sample code for parallel_for_new tutorial * Restored original parallel_for_ tutorial and table_of_content_core * Minor changes * Added demonstration of 1-D vectorized convolution * * Added 2-D convolution implementation and tutorial * Minor changes in vectorized implementation of 1-D and 2-D convolution * Minor changes to univ_intrin tutorial. Added new tutorials to the table of contents * Minor changes * Removed variable sized array initializations * Fixed conversion warnings * Added doxygen references, minor fixes * Added jpg image for parallel_for_ doc

r0hit05 changed the title ~~Tutorial for parallel_for_ and Universal Intrinsic, First Evaluation PR, GSoC '21~~ Tutorial for parallel_for_ and Universal Intrinsic (GSoC '21) Jul 9, 2021

r0hit05 marked this pull request as draft July 9, 2021 11:16

terfendail reviewed Jul 23, 2021

View reviewed changes

doc/tutorials/core/univ_intrin/univ_intrin.markdown Outdated Show resolved Hide resolved

terfendail reviewed Jul 23, 2021

View reviewed changes

doc/tutorials/core/univ_intrin/univ_intrin.markdown Outdated Show resolved Hide resolved

terfendail reviewed Jul 23, 2021

View reviewed changes

doc/tutorials/core/univ_intrin/univ_intrin.markdown Outdated Show resolved Hide resolved

terfendail reviewed Jul 23, 2021

View reviewed changes

doc/tutorials/core/univ_intrin/univ_intrin.markdown Outdated Show resolved Hide resolved

terfendail reviewed Jul 23, 2021

View reviewed changes

doc/tutorials/core/univ_intrin/univ_intrin.markdown Outdated Show resolved Hide resolved

terfendail reviewed Jul 23, 2021

View reviewed changes

...pp/tutorial_code/core/how_to_use_OpenCV_parallel_for_/how_to_use_OpenCV_parallel_for_new.cpp Outdated Show resolved Hide resolved

r0hit05 added 3 commits July 30, 2021 20:25

New parallel_for tutorial

6d989c9

Universal Intrinsics Draft Tutorial

f64ba2c

Added draft of universal intrinsic tutorial

ac27752

terfendail reviewed Aug 11, 2021

View reviewed changes

doc/tutorials/core/univ_intrin/univ_intrin.markdown Outdated Show resolved Hide resolved

r0hit05 force-pushed the master branch 3 times, most recently from 9ea74ea to 715b9d6 Compare August 12, 2021 15:54

* Added final markdown for parallel_for_new

2e98a7c

* Added first half of universal intrinsic tutorial * Fixed warnings in documentation and sample code for parallel_for_new tutorial * Restored original parallel_for_ tutorial and table_of_content_core * Minor changes

r0hit05 force-pushed the master branch from 23bfd0c to 2e98a7c Compare August 14, 2021 18:46

r0hit05 added 2 commits August 19, 2021 23:14

Added demonstration of 1-D vectorized convolution

e906078

* Added 2-D convolution implementation and tutorial

104558e

* Minor changes in vectorized implementation of 1-D and 2-D convolution

terfendail reviewed Aug 21, 2021

View reviewed changes

doc/tutorials/core/univ_intrin/univ_intrin.markdown Outdated Show resolved Hide resolved

terfendail reviewed Aug 21, 2021

View reviewed changes

doc/tutorials/core/univ_intrin/univ_intrin.markdown Outdated Show resolved Hide resolved

terfendail reviewed Aug 21, 2021

View reviewed changes

doc/tutorials/core/univ_intrin/univ_intrin.markdown Outdated Show resolved Hide resolved

Minor changes to univ_intrin tutorial. Added new tutorials to the tab…

87b287f

…le of contents

terfendail reviewed Aug 23, 2021

View reviewed changes

Minor changes

73e23ce

terfendail marked this pull request as ready for review August 23, 2021 15:38

Removed variable sized array initializations

ce89a62

Fixed conversion warnings

297fbf5

alalek reviewed Sep 4, 2021

View reviewed changes

Added doxygen references, minor fixes

607b484

Added jpg image for parallel_for_ doc

e8e0c3f

terfendail approved these changes Sep 15, 2021

View reviewed changes

alalek assigned terfendail Sep 15, 2021

alalek merged commit 41a2eb5 into opencv:master Sep 15, 2021

alalek mentioned this pull request Oct 15, 2021

(5.x) Merge 4.x #20886

Merged

mshabunin mentioned this pull request Jan 19, 2024

Removed all pre-C++11 code, workarounds, and branches #23736

Merged

		![Convolution Animation](images/convolution-example-matrix.gif)


		For more information about different kernels and what they do, look [here](https://docs.opencv.org/4.5.2/d7/da8/tutorial_table_of_content_imgproc.html).


		In the tutorial, we used a horizontal gradient filter(as shown in the animation above), which produces an image highlighting the vertical edges.

		![result image](images/resimg.png) No newline at end of file

Uh oh!

Conversation

r0hit05 commented Jul 5, 2021 • edited by alalek Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pull Request Readiness Checklist

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

terfendail commented Aug 21, 2021

Uh oh!

r0hit05 commented Aug 21, 2021

Uh oh!

terfendail commented Aug 22, 2021

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

terfendail commented Aug 23, 2021

Uh oh!

alalek commented Aug 30, 2021

Uh oh!

alalek commented Aug 31, 2021

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

r0hit05 commented Sep 7, 2021

Uh oh!

alalek commented Sep 7, 2021

Uh oh!

terfendail commented Sep 13, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

r0hit05 commented Jul 5, 2021 •

edited by alalek

Loading