Blog / Learning in public
I’m studying Aalto University’s open Programming Parallel Computers course. This is my visual learning log: start with a correct answer, then learn to move less data and do more work at once.
The optimization lab
Aalto’s running example finds the cheapest route using at most two edges. Change the version to see how the work moves through memory and execution lanes.
Strided reads
One output, one reduction chain. A row scan meets a strided column scan.
Read this version at Aalto ↗One chain depends on its previous minimum.
R[0, 1] = min over k of
D[0, k] + D[k, 1]
At k = 0: 0 + 8 = 8
Final minimum 8
The result stays the same across versions. Only how we organize the work changes.A small illustration of the access and reuse ideas, not an execution trace of Aalto’s C++ or CUDA kernels. Tile shapes and lane counts are schematic. CPU OpenMP parallelism is an additional layer across the case study.
Put the ideas to work
Every task in Open 2026, grouped by problem family and the hardware idea it develops. Topic tags are my reading of the published instructions.
Task names and requirements: Aalto PPC · Open 2026. Checked October 8, 2026. Follow the official task pages for restrictions, precision requirements, hardware, and deadlines.
Watch. Read. Try.
The lectures introduce the tools; the case studies show how they change a real implementation.
The work behind the notes
Source-linked moments from my streams. Caption excerpts retain the auto-generated wording; they can contain transcription errors.
Watch the session ↗Course study
My first learning log: why parallelism matters, how the exercises fit together, and where to start.
Download timestamped captions ↓English auto-generated captions · retrieved October 8, 2026
is programming parallel computers. So why not say programming computers? Why say programming parallel computers? So all computers are massively parallel processors. The CPU of a normal desktop or a laptop computer like what you might be watching this on easily may have hundreds of arithmetic operations in progress. simultaneously.YouTube auto-generated caption excerpt
correlated pairs and then we'll probably you know quickly understand why each of these are important or maybe like briefly try to understand the concept. Then we have image segmentation, median filter. [snorts] So I'm going to say image segmentation. Then I'm going to say median filter.YouTube auto-generated caption excerpt
the correlated pairs. So then we're going to have the CPU baseline. So then I'm going to say problem two, the first one, it's going to be CPU baseline. And then let's get a note on this.YouTube auto-generated caption excerpt
Browser captures of this blog’s caption reader, built from the timestamped source captions.
Open the course caption capture ↗Open the WebGPU caption capture ↗Keep learning
This guide is an independent learning companion by Jai Bhagat. The original course, lectures, examples, and exercise instructions are by Aalto’s Programming Parallel Computers team. Use their material alongside these notes.