SVET 2022.3 release

Main chanages:
1.Upgrade MediaSDK to OneVPL dispather v2022.0.3 and Onevpl GPU 22.3.2.
  The libva, gmmlib and media-driver versions are also upgrade
2.Upgrade OpenVINO version to 2022.1
3.Add support for Intel ADL platforms
4.Drop the support for old Intel platforms like Skylake, Kabylake, Apollo Lake
  and Coffee Lake

Signed-off-by: Elaine Wang  <[email protected]>

Author: dwang26

LICENSE

@@ -1,5 +1,5 @@
 /******************************************************************************\
-Copyright (c) 2005-2021, Intel Corporation
+Copyright (c) 2005-2022, Intel Corporation
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

README.md

@@ -38,15 +38,16 @@ See [user guide](./doc/concurrent_video_analytic_sample_application_user_guide.p
 # System requirements
 
 **Operating System:**
-* Ubuntu 20.04
+* Ubuntu 20.04.03
 
 **Software:**
-* [MediaSDK 21.1.3](https://github.com/Intel-Media-SDK/MediaSDK/releases/tag/intel-mediasdk-21.1.3)
-* [OpenVINO™ 2021.4](https://software.intel.com/en-us/openvino-toolkit)
+* [OneVPL dispather v2022.0.3] (https://github.com/oneapi-src/oneVPL/archive/refs/tags/v2022.0.3.tar.gz)
+* [OneVPL GPU 22.3.2] (https://github.com/oneapi-src/oneVPL-intel-gpu/archive/refs/tags/intel-onevpl-22.3.2.tar.gz)
+* [OpenVINO™ 2022.1](https://software.intel.com/en-us/openvino-toolkit)
 
 **Hardware:** 
-* Intel® platforms supported by the MediaSDK 21.1.3 and OpenVINO 2021.4. 
-* For Media SDK, the major platform dependency comes from the back-end media driver. https://github.com/intel/media-driver
+* Intel® platforms supported by the OneVPL GPU 22.3.2 and OpenVINO 2022.1. 
+* For OneVPL GPU, the major platform dependency comes from the back-end media driver. https://github.com/intel/media-driver
 * For OpenVINO™, see details from here: https://software.intel.com/en-us/openvino-toolkit/documentation/system-requirements
 
 # How to build
@@ -55,8 +56,50 @@ Run build_and_install.sh to install dependent software packages and build sample
 
 Please refer to ”Installation Guide“ in [user guide](./doc/concurrent_video_analytic_sample_application_user_guide.pdf) for details.
 
-## Build steps
+Below are the main steps
+
+## Install OpenVINO 2022.1, OpenCV and NEO 
+Install OpenVINO 2022.1 from https://software.intel.com/en-us/openvino-toolkit
+
+Then install OpenCV:
+```sh
+cd /opt/intel/openvino_2022/extras/scripts/
+sudo -E download_opencv.sh
+```
+
+If you see errors, you can also try below commands:
+```sh
+su
+cd /opt/intel/openvino_2022/extras/scripts/
+wget https://storage.openvinotoolkit.org/repositories/openvino/packages/2022.1/opencv/openvino_opencv_ubuntu20.tgz
+mv openvino_opencv_ubuntu20.tgz ../../
+cd ../../
+tar zxvf openvino_opencv_ubuntu20.tgz
+```
 
+Install NEO manually according to instruction on https://github.com/intel/compute-runtime/releases/tag/22.20.23198
+
+```sh
+mkdir neo; cd neo
+wget https://github.com/intel/intel-graphics-compiler/releases/download/igc-1.0.11222/intel-igc-core_1.0.11222_amd64.deb
+wget https://github.com/intel/intel-graphics-compiler/releases/download/igc-1.0.11222/intel-igc-opencl_1.0.11222_amd64.deb
+wget https://github.com/intel/compute-runtime/releases/download/22.20.23198/intel-level-zero-gpu-dbgsym_1.3.23198_amd64.ddeb
+wget https://github.com/intel/compute-runtime/releases/download/22.20.23198/intel-level-zero-gpu_1.3.23198_amd64.deb
+wget https://github.com/intel/compute-runtime/releases/download/22.20.23198/intel-opencl-icd-dbgsym_22.20.23198_amd64.ddeb
+wget https://github.com/intel/compute-runtime/releases/download/22.20.23198/intel-opencl-icd_22.20.23198_amd64.deb
+wget https://github.com/intel/compute-runtime/releases/download/22.20.23198/libigdgmm12_22.1.2_amd64.deb
+
+sudo dpkg -i *.deb
+```
+Add user to group video and render:
+```sh
+usermod -a -G video $USER
+usermod -a -G render $USER
+```
+
+At last, log out and log in again to make above change take effect.
+
+## Build steps
 Get sources with the following git command: 
 ```sh
 git clone https://github.com/intel-iot-devkit/concurrent-video-analytic-pipeline-optimization-sample-l.git cva_sample 
@@ -66,21 +109,25 @@ git clone https://github.com/intel-iot-devkit/concurrent-video-analytic-pipeline
 cd cva_sample 
 ./build_and_install.sh
 ```
-This script will install the dependent software packages by running command "apt install". So it will ask for sudo password. Then it will download libva, libva-util, media-driver and MediaSDK source code and install these libraries. It might take 10 to 20 minutes depending on the network bandwidth.
+This script will install the dependent software packages by running command "apt install". So it will ask for sudo password. Then it will download libva, libva-util, gmmlib, media-driver, onevpl and onevpl_gpu source code and install these libraries. It might take 10 to 20 minutes depending on the network bandwidth.
 
 After the script finishing, the sample application video_e2e_sample can be found under ./bin.
 
-In order to enable the media SDK installed by SVET to coexist with different versions of media SDK installed on the same computer, we suggest (we have done so in our build script) that the media SDK environment variables of SVET should only be set in the current bash, not saved to the global system environment.
+In order to enable the libva/media-driver/onevpl installed by SVET to coexist with different versions of libva/media-driver/onevpl installed on the same computer, we suggest (we have done so in our build script) that the media SDK environment variables of SVET should only be set in the current bash, not saved to the global system environment.
 So please run 'source ./svet_env_setup.sh' first when you start a new shell (or change user in shell such as run 'su -') to run ./bin/video_e2e_sample".
 
 ```sh
 cd cva_sample
 source ./svet_env_setup.sh
+./bin/video_e2e_sample -par sanity_test/n1nodsp.par
+./bin/video_e2e_sample -par sanity_test/n2fd.par
 ```
-Please refer to "Run sample application" in [user guide](./doc/concurrent_video_analytic_sample_application_user_guide.pdf) for details.
+n1nodsp.par is a basic test to check if the media stack work correctly.
+n2fd.par runs two video decode and face detection inference without display. It can be used to check if the OpenVINO and NEO installed correctly.
+Please refer to "Run sample application" for more use cases in [user guide](./doc/concurrent_video_analytic_sample_application_user_guide.pdf) for details.
 
 # Known limitations
 
-The sample application has been validated on Intel® platforms Skylake(i7-6770HQ), Coffee Lake(i7-8559U i7-8700), Whiskey Lake(i7-8665UE) and Tiger Lake U(i7-1185G7E, i5-1135G7E).
+The sample application has been validated on Intel® platforms Tiger Lake U(i7-1185G7E, i5-1135G7E, Celeron 6305E) and Adler Lake (i5-12400) 
 
 

build_and_install.sh

@@ -4,182 +4,76 @@ root_path=$PWD
 #download_method="git"
 download_method="wget"
 built_with_ocv=true
-LIBVA_INSTALL_PREFIX="/opt/intel/svet/msdk"
-LIBVA_INSTALL_PATH="${LIBVA_INSTALL_PREFIX}/lib"
-LIBVA_DRIVERS_PATH="${LIBVA_INSTALL_PATH}/dri"
-LIBVA_DRIVER_NAME="iHD"
 
-while getopts b: flag
-do
+install_overwrite=False
+
+while getopts 'b:i:' flag; do
     case "${flag}" in
         b)
-            if [[ "no_ocv" == ${OPTARG} ]]
-            then
+        if [[ "no_ocv" == ${OPTARG} ]]
+        then
                 echo "Build SVET without OpenVINO"
                 built_with_ocv=false
-            fi
-            ;;
+        fi
+        ;;
+        i)
+        if [[ "overwrite" == ${OPTARG} ]]
+        then
+                echo "Overwrite the default Intel media stack library installed."
+                install_overwrite=True
+                echo 'export SVET_INSTALL_OVERWRITE=True' >> ~/.bashrc
+                export SVET_INSTALL_OVERWRITE=True
+        fi
+        ;;
     esac
 done
 
+
 if [ ! -d "${INTEL_OPENVINO_DIR}" ] && [ $built_with_ocv = true ];
 then echo "Please make sure openvino has been installed and enviroment variables has been set by "
     echo "source intel/openvino/bin/setupvars.sh"
     exit -1;
 fi
 
-sudo apt install python
+mkdir -p vpl
+cd vpl
+rm $root_path/svet_build.log
+touch $root_path/svet_build.log
+python3 ../onevpl_pre_install.py | tee $root_path/svet_build.log
 
-if [[ $download_method == "git" ]]
+mkdir -p ~/cl_cache
+export cl_cache_dir=~/cl_cache
+
+if [[ -d $cl_cache_dir ]];
 then
-python msdk_pre_install_internal.py | tee svet_download.log
+	echo "cl_cache is enabled. \$cl_cache_dir : $cl_cache_dir"
+	echo "Clean previous cl kernel cache"
+	sudo rm -f $cl_cache_dir/*
 else
-python msdk_pre_install.py | tee svet_download.log
+	echo "Add enabling cl_cache commands to ~\.bashrc"
+	mkdir -p ~/cl_cache
+	export cl_cache_dir=~/cl_cache
+	echo "mkdir -p ~/cl_cache" >> ~/.bashrc
+	echo "export cl_cache_dir=~/cl_cache" >> ~/.bashrc
 fi
 
-git_projects="MediaSDK media-driver libva"
-for i in $git_projects;
-do
-    if [[ ! -d $i ]];
-    then
-        echo "Directory $i doesn't exist. Please run ./msdk_pre_install.py manually to
-    download the code"
-    exit -1;
-    fi
-done
+cd $root_path
+source ./svet_env_setup.sh
+rm -rf video_e2e_sample/build
 
-#check md5sum if wget is used to download source code packages
+mkdir -p video_e2e_sample/build
 
-if [[ $download_method == "wget" ]]
+if [ $built_with_ocv != true ]
 then
-    echo "Checking the md5sum of downloaded packages"
-    declare -A  media_packages_md5sum
-    media_packages_md5sum[intel-gmmlib-21.1.1.tar.gz]=100ff57bff1cb19763f729fcd00269d1
-    media_packages_md5sum[intel-media-21.1.3.tar.gz]=98f59d7fd2c6c08808acf0ba741216b0
-    media_packages_md5sum[intel-mediasdk-21.1.3.tar.gz]=a4747e315448ded38171b245232a5a7c
-    media_packages_md5sum[libva.tar.gz]=ba136b5a02ee4c7bf093d09bc1635560
-    media_packages_md5sum[libva-utils.tar.gz]=010243cc8217fd455d0ad11f2025d43c
-
-    for c in ${!media_packages_md5sum[@]}; do
-        md5_res=`md5sum $c`
-        if [[ ${media_packages_md5sum[$c]} ==  ${md5_res%% *} ]]
-        then
-            echo "$c" "${media_packages_md5sum[$c]}" correct
-        else
-            echo "$c" "md5sum is wrong " ${md5_res%% *}  vs ${media_packages_md5sum[$c]}
-            echo "Pleaes remove libva/media-driver/MediaSDK re-run this script"
-            echo "If you still see md5sum not match, please refer to msdk_pre_install.py and download, uncomparess the package manually"
-            echo "Or you can modify line 5 in this scirpt and change the download_method to \"git\". Then re-run this script."
-            exit
-        fi
-    done
-fi
-
-cd "MediaSDK"
-patch -N --no-backup-if-mismatch -p1 < ../patches/0001-MSDK-Enable-AVC-decode-SFC-RGB4.patch
-patch -N --no-backup-if-mismatch -p1 < ../patches/0002-sample_common-Add-support-to-multiple-displays.patch
-
-if [ $? != 0 ]; then
-    echo "Apply patch failed!"
+	echo "Use the CMakefile.txt that excludes inference source code"
+	cp script/CMakeLists_woOpenVINO.txt video_e2e_sample/CMakeLists.txt
 fi
 
-cd ../
-
-if [[ $download_method == "git" ]]
-then
-python msdk_pre_install_internal.py -b cfl| tee svet_build.log
-else
-python msdk_pre_install.py -b cfl | tee svet_build.log
-fi
-
-#video_e2e_sample depends on libsample_common.a
-if [[ -f MediaSDK/build/__lib/release/libsample_common.a ]];
-then
-    rm -f bin
-    ln -s MediaSDK/build/__bin/release/ bin
-
-
-    #va_env_set=`grep "LIBVA_DRIVERS_PATH=/usr/lib/x86_64-linux-gnu/dri" ~/.bashrc -c`
-    set -x
-    export LD_LIBRARY_PATH=$LIBVA_INSTALL_PATH:$LD_LIBRARY_PATH
-    export LIBRARY_PATH=$LIBVA_INSTALL_PATH:$LIBRARY_PATH
-    export LIBVA_DRIVERS_PATH=$LIBVA_DRIVERS_PATH
-    export LIBVA_DRIVER_NAME=$LIBVA_DRIVER_NAME
-    export MFX_HOME=$LIBVA_INSTALL_PREFIX
-    set +x
-
-    #if (( va_env_set == 0 ))
-    #then
-        #set -x
-        #export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib/x86_64-linux-gnu:/usr/lib
-        #export LIBVA_DRIVERS_PATH=/usr/lib/x86_64-linux-gnu/dri
-        #export LIBVA_DRIVER_NAME=iHD
-        #export MFX_HOME=/opt/intel/mediasdk
-        #set +x
-        #echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib/x86_64-linux-gnu:/usr/lib' >> ~/.bashrc
-        #echo "export LIBVA_DRIVERS_PATH=/usr/lib/x86_64-linux-gnu/dri" >> ~/.bashrc
-        #echo "export LIBVA_DRIVER_NAME=iHD" >> ~/.bashrc
-        #echo "export MFX_HOME=/opt/intel/mediasdk" >> ~/.bashrc
-
-    #fi
-
-    #if [[ -d "/opt/intel/mediasdk/lib64" ]]
-    #then
-    #    echo "A old MediaSDK version is found under folder /opt/intel/mediasdk/lib64"
-    #    echo "Rename /opt/intel/mediasdk/lib64 as /opt/intel/mediasdk/oldlib64"
-    #    mv  /opt/intel/mediasdk/lib64 /opt/intel/mediasdk/oldlib64
-    #fi
-
-    #if [[ $LD_LIBRARY_PATH != *"/opt/intel/mediasdk/lib/"* ]]
-    #then
-    #    export LD_LIBRARY_PATH="/opt/intel/mediasdk/lib/:$LD_LIBRARY_PATH"
-    #    echo 'export LD_LIBRARY_PATH="/opt/intel/mediasdk/lib/:$LD_LIBRARY_PATH"' >> ~/.bashrc
-    #fi
-
-    if [[ -d $cl_cache_dir ]];
-    then
-        echo "cl_cache is enabled. \$cl_cache_dir : $cl_cache_dir"
-        echo "Clean previous cl kernel cache"
-        sudo rm -f $cl_cache_dir/*
-    else
-        echo "Add enabling cl_cache commands to ~\.bashrc"
-        set -x
-        mkdir -p ~/cl_cache
-        export cl_cache_dir=~/cl_cache
-        echo "mkdir -p ~/cl_cache" >> ~/.bashrc
-        echo "export cl_cache_dir=~/cl_cache" >> ~/.bashrc
-    fi
-
-    #install library for metric_monitor
-    sudo cp MediaSDK/build/__bin/release/libcttmetrics.so /usr/lib/
-
-    #Start to build SVET sample application
-    #Copy MediaSDK sample headers and libraries that weren't installed by default
-    sudo -E mkdir -p $MFX_HOME/include/sample_common
-    sudo -E mkdir -p $MFX_HOME/include/vpp_plugin
-    sudo -E mkdir -p $MFX_HOME/include/rotate_cpu
-    sudo -E cp -rf MediaSDK/samples/sample_common/include/ $MFX_HOME/include/sample_common/
-    sudo -E cp -rf MediaSDK/samples/sample_plugins/vpp_plugin/include/ $MFX_HOME/include/vpp_plugin/
-    sudo -E cp -rf MediaSDK/samples/sample_plugins/rotate_cpu/include/ $MFX_HOME/include/rotate_cpu/
-    sudo -E cp -f MediaSDK/build/__lib/release/libsample_common.a  $LIBVA_INSTALL_PATH/
-    sudo -E cp -f MediaSDK/build/__lib/release/libvpp_plugin.a $LIBVA_INSTALL_PATH/
-
-    rm -rf video_e2e_sample/build
-
-    mkdir -p video_e2e_sample/build
-
-    if [ $built_with_ocv != true ]
-    then
-        echo "Use the CMakefile.txt that excludes inference source code"
-        cp script/CMakeLists_woOpenVINO.txt video_e2e_sample/CMakeLists.txt
-    fi
-
-    cd video_e2e_sample/build
-    cmake ../ | tee /tmp/svet_build.log
-    cat /tmp/svet_build.log >> svet_build.log
-    make -j4 | tee /tmp/svet_build.log
-    cat /tmp/svet_build.log >> svet_build.log
-fi
+cd video_e2e_sample/build
+cmake ../ | tee /tmp/svet_build.log
+cat /tmp/svet_build.log >> $root_path/svet_build.log
+make -j4 | tee /tmp/svet_build.log
+cat /tmp/svet_build.log >> $root_path/svet_build.log
 
 cd $root_path
 if [ $built_with_ocv = true ]
@@ -218,25 +112,19 @@ then
     echo "************************************************************************"
     echo "SVET sample application building has completed!"
 
-    gen12=`sudo cat /sys/kernel/debug/dri/0/i915_capabilities | grep "gen: 12"`
-    if [[ ! -z $gen12 ]];
+    if [[ $install_overwrite == True ]]
     then
-        ver=`uname -a`
-        if [[ $ver != *5.4.102* ]] ;
-        then
-            echo "For TigerLake CPU, please refer to user guide chapter 1.3 to upgrade the kernel with https://github.com/intel/linux-intel-lts/releases/tag/lts-v5.4.102-yocto-210310T010318Z";
-        fi
+        echo "Please run ./bin/video_e2e_sample for testing"
+    else
+        echo "Please run the cmd below to setup running environment:"
+        echo " "
+        echo "source ./svet_env_setup.sh"
+        echo " "
+        echo "Then use ./bin/video_e2e_sample for testing"
+        echo "IMPORTANT NOTICE: please run 'source ./svet_env_setup.sh' first when you start a new shell (or change user in shell such as run 'su -') to run ./bin/video_e2e_sample"
+        echo "************************************************************************"
     fi
-
-    echo "Please run the cmd below to setup running environment:"
-    echo " "
-    echo "source ./svet_env_setup.sh"
-    echo " "
-    echo "Then use ./bin/video_e2e_sample for testing"
-    echo "IMPORTANT NOTICE: please run 'source ./svet_env_setup.sh' first when you start a new shell (or change user in shell such as run 'su -') to run ./bin/video_e2e_sample"
-    echo "************************************************************************"
 else
     echo "SVET sample application building failed!"
-    echo "Please create an IPS with SVET in title on premiersupport.intel.com and upload ./svet_build.log ./svet_download.log to IPS."
+    echo "Please create an IPS with SVET in title on premiersupport.intel.com and upload ./svet_build.log to IPS."
 fi
-

doc/FAQ.md

@@ -34,9 +34,6 @@ Add "-fps 30" to every decoding session.
 ## How to limit the frame number of input to 1000?
 Add "-n 1000" to every decoding dessions. But please not add "-n" to encode, display and fake sink session. These sink sessions will automatically stop when the source session stops. Note, this option won't work if both "-vpp_comp_only" and "-vpp_comp" are set. 
 
-## Where can I find tutorials for inference engine?
-Please refer to https://docs.openvinotoolkit.org/latest/_docs_IE_DG_Deep_Learning_Inference_Engine_DevGuide.html
-
 ## Why HDDL card usage ratio is low for face detection inference?
 It can be caused by the decoded frames aren't fed to inference engine efficiently. The default inference interval of face detection is 6. You try to set the inference interval to a lower valuer when using HDDL as inference target device. For example, with 3 HDDL L2 card, adding "-infer::inverval 1" to 16-channel face detection par file can increase the HDDL usage ratio to 100%.  
 
@@ -46,7 +43,3 @@ face-detection-retail-0004, human-pose-estimation-0001, vehicle-attributes-recog
 
 ## Can I use other OpenVINO version rather than 2021.3 ?
 Yes, but you have to modify some code due to interfaces changing. And also you need to download the IR files and copy them to ./model manually. Please refer to script/download_and_copy_models.sh for how to download the IR files.
-
-## When run 4 channel decode plus inference and display on APL, the CPU occupy ratio is very high and fps is low
-Please refer to par file par_file/inference/n4_face_detection_rgbp.par. It uses option "-dc::rgbp" that make the SFC outputs RGB4 for display and VPP outputs RGBP for inference input. Then there is no need to use OpenCV for resizing and color conversion which consume much CPU time on APL.
-Note, "-dc::rgbp" only works with "-infer::fd". Will support more inference types in the future.