agostini01

#/bin/bash

set -o pipefail
set -e 

# Verbose commands for now
set -x

# Check if the first argument is passed
if [ "$#" -eq 0 ]; then

agostini01

#!/usr/bin/python3
# Author: Nicolas Bohm Agostini
# Date: 2022-Dec-12
# Version: 1.1
# License: MIT

# Changelog:
# 1.1 - added alternative way of printing

# @brief: This script parses a bib file and prints it in a specific format

agostini01

Links for Nisght tools and tutorials

Download center

• https://developer.nvidia.com/gameworksdownload

Installing Nsight compute from command-line:

# Download .run file from website and accept the terms and conditions.
sudo ./nsight-compute-linux-2020.1.1.8-28506821.run --nox11

agostini01

241   execve("/usr/bin/htop", ["htop"], 0x7fffeaaf9b78 /* 24 vars */) = 0
241   brk(NULL)                         = 0x7fffd27d9000
241   arch_prctl(0x3001 /* ARCH_??? */, 0x7fffdaf59710) = -1 EINVAL (Invalid argument)
241   access("/etc/ld.so.preload", R_OK) = -1 ENOENT (No such file or directory)
241   openat(AT_FDCWD, "/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 3
241   fstat(3, {st_mode=S_IFREG|0644, st_size=27438, ...}) = 0
241   mmap(NULL, 27438, PROT_READ, MAP_PRIVATE, 3, 0) = 0x7facd63d6000
241   close(3)                          = 0
241   openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libncursesw.so.6", O_RDONLY|O_CLOEXEC) = 3
241   read(3, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\20\221\0\0\0\0\0\0"..., 832) = 832

agostini01

Extracted with:

for var in synthesizer._model.model.all_vars:
    print('{:15} {} {}'.format(
        '{}'.format(var.get_shape().as_list()),
        var.dtype,
        var.name))

agostini01

I have been working on TF 2.0 and the custom op tutorial presented here does allow execution of the op.

This is mainly due to eager execution and different custom op handling.

In order to call a custom op in python one must:

Start docker image

docker pull tensorflow/tensorflow:nightly-custom-op-ubuntu16

agostini01

#install ubuntu 18.04
apt update && apt dist-upgrade -y
apt install vim tmux htop git tree curl wget ctags linux-tools-common linux-tools-generic linux-tools-`uname -r` valgrind kcachegrind -y

# nvidia drivers
cd # changes to root's home folder
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-ubuntu1804.pin
mv cuda-ubuntu1804.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget http://developer.download.nvidia.com/compute/cuda/10.1/Prod/local_installers/cuda-repo-ubuntu1804-10-1-local-10.1.243-418.87.00_1.0-1_amd64.deb
dpkg -i cuda-repo-ubuntu1804-10-1-local-10.1.243-418.87.00_1.0-1_amd64.deb

agostini01

You can remove these apps (or re-install them if needed) using the XIAOMI ADB/Fastboot Tools.

https://github.com/Saki-EU/XiaomiADBFastbootTools

Alternatively it is possible to use the commandline:

pm uninstall -k --user 0 com.mi.android.globalpersonalassistant
pm uninstall -k --user 0 com.xiaomi.mipicks
pm uninstall -k --user 0 com.android.browser

agostini01

Connecting to the Zedboard

We will go through the steps of communicating to a digilent Zedboard using the UART and the Ethernet port.

This tutorial assumes that the zedboard SD card was flashed with the Xillinux distribution for Zynq-7000 EPP and the host system is running a ubuntu 14.04 or newer (tested on 16.04).

Using UART port

Connect the host (USB) to the zedboard's UART port (micro USB) and execute on the host:

agostini01

Followed this tutorial:

http://www.bnikolic.co.uk/blog/hpc-howto-measure-flops.html



    Get the latest version of perfmon2/libpfm (h/t this developerworks article):

    git clone git://perfmon2.git.sourceforge.net/gitroot/perfmon2/libpfm4
    cd libpfm4