Darknet on OpenCL on Windows 11 x64

Hi, there are many challenges on Windows 11 x64 with AMD Radeon RX 6900 XT. The last few months from time to time I started to try to build on Windows 10 x64 first and later I updated to Windows 11 x64. There are the following open issues I faced: 1) Time spent on detection is much longer and 10 times more counts than on GNU/Linux or macOS. 2) Detections are wrong Continue ReadingDarknet on OpenCL on Windows 11 x64

GPU-Computing macOS Big Sur vs iBuntu Lux Sur

Hello, Let’s test this beauty from the last post on… LuxMark ;-). The thing is that I wanted to compare OpenCL. I know that OpenCL in both systems is not working fully right. I found on macOS that (multi-GPU case) clFinish(q) is slow down each time, on iBuntu -cl-fast-relaxed-math cause a black-scenes on Lux Ball… and on Darknet on OpenCL it is more things wrong, but maybe there are my bugs, so… Continue ReadingGPU-Computing macOS Big Sur vs iBuntu Lux Sur

Mac Moria Miner Project Failure

Hello, every failure is a lesson, so let me describe the experiment I did and the story behind it… I dream about the most robust possible GPU-computing macOS. Let’s start with hardware… design, the way it looks like, is not a design, so let’s name it the look and feel ;-). I am sorry to Apple, please forgive me this was only for test and future, again try, to Ph.D. studies. Do Continue ReadingMac Moria Miner Project Failure

GPU-Computing for Vision Recognition Platform as IoT

Hi my dear readers, today I would like to tell you three technical stories. It will be stories with a happy ending… and you know me..? I am not sure if that story happens precisely this way, but the story is true. IoT Idea Story ;-). First of all, you probably know my passion for IoT? And also, as a tool builder, sorry to say, I do not like to buy the Continue ReadingGPU-Computing for Vision Recognition Platform as IoT

DarkNet-vNext on CUDA

Hello, I know that many people like my DarkNet on OpenCL. Recently I also made an improved version of the original DarkNet on CUDA, and I called it DarkNet-vNext, and it has all improvements that probably you were looking for. All Yolo models are supported from the 1st to the 4th version. It supports OpenCV 4 and, for example, benchmark possibility. The good news is it is high-speed. As you probably may Continue ReadingDarkNet-vNext on CUDA

The Multi-GPU-SET Idea

Hello, I am a Ph.D. student in Poland at the Silesian University of Technology. And I wanted to start a general discussion on the classifier on the video stream improvement idea. I called it the “Multi-GPU & Multi-SET” or, in short, the “Multi-GPU-SET” idea. Basically, people use Multi-GPU and the “Syncing” for Convolutional Neural Networks. But nobody, in my opinion, tried multi GPU to classification. So what is it about? Well, it Continue ReadingThe Multi-GPU-SET Idea

GPU OpenCL Fine-Tuning Problem Solution

Hello, after about half-year I at last found the solution of the biggest optimization issue I had so far in the Darknet on OpenCL solution. It was really hard to solve and I even wrote at the AMD Community Post. Today very early morning I post on that post solution that is shown below. CPU CODE THAT INVOKES THE GPU CODE AND USES BOTH GLOBAL AND LOCAL THREAD SPACES! GPU CODE THAT Continue ReadingGPU OpenCL Fine-Tuning Problem Solution

PhD Progress from May 27th 2020 Update Keynote

Hi, This video was recorded today, and it shows in detail my Ph.D. first year of study effects. The first on the planet… Multi-GPU & Multi-SET image classification pattern… so, I wish you the happy watch! Thanks for watching! p ;).

PhD OpenCL Challenges

Hi, today I want to show you code samples for OpenCL that may be important when you start with these amazing graphics cards computing (GPU-computing) library. The goal is to make code as good as possible on any platform that supports OpenCL and on GPUs from AMD, Intel, NVidia, Mali on CPUs from AMD, Intel, ARM. All examples are from the Darknet on OpenCL port I did some time ago. The first Continue ReadingPhD OpenCL Challenges

PhD Progress from March 8th 2020 Update Keynote

This is the status update of my private “after work” progress on Ph.D. studies at the Silesian University in Gliwice, Poland. I think it may interest all AI/ML researchers around the world. Basically, I covered in 30 minutes the Conventional Neural Network (CNN) engine used to classify the objects on video frames. I also cover my implementation of the Darknet on OpenCL that you can find at https://github.com/sowson/darknet including the goal, idea, Continue ReadingPhD Progress from March 8th 2020 Update Keynote

GPU OpenCL Fine-Tuning Problem

Hi, today I have a C and OpenCL code fine-tuning problem to think about. This problem is quite complex to solve and truth be told I do not know how and why it does not work. The example is from real project at https://github.com/sowson/darknet. Let me first show you slow to compute version of example method in OpenCL. To give you all the aspects of computation you need to see also C Continue ReadingGPU OpenCL Fine-Tuning Problem

DarkNet CUDA vs OpenCL and CPU vs NVIDIA vs AMD

Hi, today I will show you some measurements results for my PhD. I am working of the first publication about DarkNet on OpenCL, the source code of this project you can find at https://github.com/sowson/darknet. The IEEE publication has to be consistent and smart. I cannot put on it too much graphics and big tables… but waits, I have a public blog site. So, I can post it here. First things first the Continue ReadingDarkNet CUDA vs OpenCL and CPU vs NVIDIA vs AMD

GPU Computing on OpenCL

Hi, today I have something very nice for all readers. The subject is the GPU Computing on OpenCL for all readers. Abstract: Have you ever wanted to sum 10 million random numbers in array in C? Do you know that on GPU it can be faster about 20k times than on CPU? This post is about GPU Computing and all examples were written in C language with OpenCL 1.2+. Basic understanding of Continue ReadingGPU Computing on OpenCL