arthurcerveira / Video-Memory-Analysis-Environment

A Docker container to perform memory analysis on HEVC and VVC Test Models

Memory Analysis Environment for HEVC and VVC Test Models

A Docker container to perform memory analyses on an isolated environment.

Instructions

Prerequisites

The only tools needed to run this container is docker-compose and git.

Running the container

First you need to clone the repository to your local machine

$ git clone https://github.com/arthurcerveira/Video-Memory-Analysis-Environment.git

Then you'll need to create a directory video_sequences to store the videos(.yuv).

$ cd Video-Memory-Analysis-Environment
$ mkdir video_sequences

To build the container and start the analysis you need to run the following commands

$ docker-compose build
$ docker-compose run -d --name running-mem mem-analysis

Once the execution is over, the results will be available in the Custom-Simulator directory on the host machine and the container can be removed.

$ docker rm running-mem

Settings

The settings used for the analysis can be found in Custom-Simulator/custom_simulator.py.

The default values are:

# Routines
AUTOMATE_TRACE = True
GENERATE_TRACE_GRAPH = False
GENERATE_BLOCK_GRAPH = False

AUTOMATE_VTUNE = False
GENERATE_VTUNE_GRAPH = False

# Trace Reader
AUTOMATE_TRACE_OUTPUT = "automate_trace_output.txt"

# Vtune Reader
AUTOMATE_VTUNE_OUTPUT = "automate_vtune_output.txt"

# Encoder Paths
HM = "../hm-videomem/"
VTM = "../vtm-mem/"

HEVC = True
VVC = True

CONFIG = {"HEVC": {"Low Delay": HM + "cfg/encoder_lowdelay_main.cfg",
                   "Random Access": HM + "cfg/encoder_randomaccess_main.cfg"},
          "VVC": {"Low Delay": VTM + "cfg/encoder_lowdelay_vtm.cfg",
                  "Random Access": VTM + "cfg/encoder_randomaccess_vtm.cfg"}}
VIDEO_CFG_PATH = {"HEVC": HM + "cfg/per-sequence/",
                  "VVC": VTM + "cfg/per-sequence/"}

VIDEO_SEQUENCES_PATH = "../video_sequences"

# Parameters
FRAMES = '17'
SEARCH_RANGE = ['96']

Running the image on DockerHub

The image on DockerHub uses the default settings and the video sequence BQSquare.

$ docker container run arthurcerveira/video-memory-analysis:1.0

Results

Example of results achieved by this container are available in this repository.