garyexplains / examples

Example code used in my videos

Examples

Welcome to the Gary Explains GitHub repository!

Example code used in my videos

You can find my videos here: https://www.youtube.com/c/GaryExplains

examples's Issues

Taking it farther and netbooting from iSCSI

Hey Gary!

I've been following a guide for setting up the PI to boot from an iSCSI drive instead of SD card.
There were a few interesting benchmarks about read/write performance that you or your viewers may find interesting.
Sašo Stanovnik's Guide on XLAB for Netbooting a PI from an iSCSI drive

ORTE has lost communication with a remote daemon.

^Zmpiexec: Forwarding signal 20 to job

ORTE has lost communication with a remote daemon.

HNP daemon : [[64112,0],0] on node wmaster-desktop
Remote daemon: [[64112,0],1] on node 10.126.10.10

This is usually due to either a failure of the TCP network
connection to the node, or possibly an internal failure of
the daemon itself. We cannot recover from this failure, and
therefore will terminate the job.

Log for Compiler Tests

Hi Gary,

I recently watched your video on Windows vs Linux C++ compiler video. I reproduced the test on my hardware and have got different results. My results and workflow with the hardware and software version details are listed below.
I appreciate your detailed video breakdowns and information. This particular video should also touch upon the wider topic of C++ compilers. Microsoft has been releasing newer version of Visual Studio with significant performance upgrades and better compliance towards Standard C++. Healthy competition between compilers is necessary otherwise C++ compilers would look like Web Engine landscape with dominance of one particular Toolchain.

Results -

Visual Studio
subleq() took 21.78s
subleq3() took 8.27s
subleq_g3() took 9.02s

GCC 10
subleq() took 10.36s
subleq3() took 10.35s
subleq_g3() took 10.55s

GCC 9.3
subleq() took 10.27s
subleq3() took 9.79s
subleq_g3() took 10.28s

Specs -

CPU  -  Intel® Core™  i7-9750H CPU @ 2.60GHz
RAM -  32 GB @ 2.667GHz
OS - Windows 10 Pro with 2004 Update
Visual Studio 2019 16.7.1 
GCC  9 Version on WSL Ubuntu 20.04.1 LTS  - 9.3.0 (installed using apt)
GCC 10 Version on WSL Ubuntu 20.04.1 LTS - 10.0.1 (installed using apt)
Target - x64 for both the compilers (kindly ensure that you are using x64 target for Visual Studio Developer tools)

Commands executed -

Visual Studio Developer Prompt with [vcvarsall.bat] Environment initialized for: 'x64'
cl /O2 compiler_test1.cpp
compiler_test1.exe

GCC 10
g++-10 -O3 compiler_test1.cpp -o compiler_test1
./compiler_test1

GCC 9.3
g++ -O3 compiler_test1.cpp -o compiler_test1
./compiler_test1

simpleMPI problem Jetson Nano

In https://github.com/garyexplains/examples/blob/master/how_to_build_nvidia_jetson_gpu_cluster.md

you say:

Edit simpleMPI.cpp and alter int blockSize = 256; to:
int blockSize = 384 / commSize;

bit in simpleMPI.cpp block size is initialized after this.

if i change the order to

int commSize, commRank;
MPI_CHECK(MPI_Comm_size(MPI_COMM_WORLD, &commSize));
MPI_CHECK(MPI_Comm_rank(MPI_COMM_WORLD, &commRank));

int blockSize = 384 / commSize;
int gridSize = 10000;
int dataSizePerNode = gridSize * blockSize;

it compiles but i cannot execute it

Running on 1 nodes
CUDA error calling "cudaMemcpy(hostData, deviceOutputData, dataSize *sizeof(float), cudaMemcpyDeviceToHost)", code is 702
Test FAILED

MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
with errorcode 702.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.

any idea? the vanilla version is working fine

Changing the URL name

Hi, is it possible to add an extra name to the back of the URL to differentiate the video stream and still image?
e.g. http://192.168.5.5:8000/image & http://192.168.5.5:8000/video

if it is possible, which lines can I add/change in the code?

Thank you!

vASM STH elif typo

b = [0x14, r, 0, r2] should be b = [0x15, r, 0, r2] otherwise the *.bin will use STL op code for either STL or STH.

About ARM ISA

Sorry Gary to use this channel to get contact with you I couldn't find your email..

I am Salvador from Pi Labs YouTube channel. I would like to know your thoughts
About the ISA impact on efficiency.. I saw one old paper about that claiming
Than the ISA have almost no impact on efficiency, and great ARM devs
Suggested thab the ISA have impact on the efficiency...
While ARM soc's are made targeting low power I believe than if we make
An ARM chip of 65W TDP we should win on most escenarios against a similar lithography and power consumption x86 chip. And not only bc of the microarch, bc of the ISA itself.
I would love a video on your thoughts about it...

Salvador.
Also, my mail is [email protected] if you want to close an answer there.
Best regards

Please release the SHA256 benchmark

It matters but not that much how optimised a benchmark is as it isn't optimised for any comparing object.
I'd like to run this benchmark on my AMD 3700X to get a better understanding of the performance of different ISAs and I want to evaluate if I should get a Nvidia Jetson NX cluster.
So please release the source code with the compiler version and optimization options you used.

ublox_tinygps_ntp_server_pico - dealing with malformed udp

Thanks for the ublox_tinygps_ntp_server_pico. Made tiny mods to run on ESP8266. Would run for 40-60 mins then crash having received a udp packet from 71.69.84.31 of packet size 18516. Couldn't fathom how this was getting on my LAN. Don't accept anything incoming. Possible corruption in udp stack? Minor mod to your code to only process udp packets if ((packetSize > 40) && (packetSize < UDP_TX_PACKET_MAX_SIZE)). Trust of interest.

typedef __fp16 ggml_fp16_t; on Raspberrypi 4

for solve ;-)

ggerganov/whisper.cpp#1771

E: Unable to locate package tkinter

I'm not really familiar with linux, and stuck in getting this error when trying to install stuff from point Get the compilers etc in examples/Raspberry Pi Pico/C/ readme instructions.

I have fresh installed WSL Ubuntu 20.04 LTS and also tried on WSL Ubuntu 18.04 LTS

me@my:~$ sudo apt install gcc-arm-none-eabi libnewlib-arm-none-eabi build-essential libssl-dev tk tkinter python3-tk

Output:

Reading package lists... Done
Building dependency tree
Reading state information... Done
E: Unable to locate package tkinter

Jetson cluster does not work

I have tried your steps many times and hours and it does not work can you please kindly help me
I get running on 2 nodes but in reality it is not working

Documentation issue

Command wrong in documentation

wget https://huggingface.co/TheBloke/Llama-2-7b-Chat-GGUF/resolve/main/llama-2-7b-chat.Q4_K_S.gguf)

Removed the ")"

Binary Clock with Pico W - code error

Location: Line 100 of main.py in this https://github.com/garyexplains/examples/tree/master/Raspberry%20Pi%20Pico/How-to-Make-a-Binary-Clock-Using-a-Raspberry-Pi-Pico-W

Content: "Y, M, D, W, H, M, S, SS = rtc.datetime()"

Issue: The letter "M" is used twice, once to mean "month" and again to mean "minute"

Comment: It took me ages to realise that my code derived from yours was not working because one variable was being used for two different things!

Piccolo OS source code

I wasn't able to find the source code for the OS in examples/Raspberry Pi Pico/Piccolo_OS/. Could you please let me know if I'm looking in the wrong place?

Ethernet connection is working as expected. but wifi is not.

The connection with the eth1 one is working as it should but wlan0 is not.

after booting the Pi, I connect to the access point "PiNet". im able to use the internet for around a minute but after that no internet. not able to ping the Pi via the wlan0 ip (192.168.17.1).

same behavior after multiple reboots

Small problem with this code: primality_cluster_test2.py

Thank you for your nice program located here:

https://github.com/garyexplains/examples/blob/master/primality_cluster_test2.py

I have been using this program as a basis for some of my own exploration of MPI and parallel processing. I found a small error in the program that causes it to occassionally miss prime numbers. Note the following code:

for p in range(me[0], me[1]):
if isPrimeTrialDiv(p) == True:
results.append(p)

The second number in the range should be me[1] + 1. As written the code will not check me[1] and miss any prime number that occurs at the end of the range.

Thank you for your attention to this matter,
Alex Ambrioso (https://github.com/aambrioso1)

Discussion about (missing) compiler optimizations and wrong conclusions

https://www.cnx-software.com/2018/11/29/amazon-ec2-a1-arm-cost-savings/#comment-558960