cudf VS PSRayTracing

The interesting thing about Polars is that it does not try to be a drop-in replacement to pandas, like Dask, cuDF, or Modin, and instead has its own expressive API. Despite being a young project, it quickly got popular thanks to its easy installation process and its “lightning fast” performance.

Perhaps the largest for package size is the NVIDIA developed rapids toolkit https://rapids.ai/ . Even still adding things like pandas and some geospatial tools, you rapidly end up with an image well over a gigabyte, despite following cutting edge best practice with docker and python.

Yes sure, that is how OpenMP does; but on the other side: you seem to already do some basic type inference, and building an AST, no? Then you know as well the size and type of your vectors, and can execute actions in parallel if there is enough data to be worth parallelizing. Is there anyone who don't want their code to execute faster if it is possible? Those that do work in big data domain do use threads and vectorized instructions without user having to type in any directive; just import different library. Example, numpy or numpy with cuda backend, or similar GPU accelerated libraries like cudf.

Not the answer about Ray: you could use rapids.ai. I'm using it for for dataframe manipulation on GPU

To put Data Analytics on GPU Steroids, Try RAPIDS cudf https://rapids.ai/

You can scope out https://rapids.ai/. Nvidia's AI toolkits. They have some handy notebooks to poke at to get you started.

try https://rapids.ai/

As a companion to the other comments, I'd like to mention that the RAPIDS library cuML provides GPU-accelerated versions of quite a few of the algorithms mentioned in this thread (HDBSCAN, UMAP, SVM, PCA, {Exact, Approximate} Nearest Neighbors, DBSCAN, KMeans, etc.).

I wouldn't say that CMake isn't that painful for the deployment stage. I have successfully deployed an open source project on Windows, Mac, Linux, Android, and iOS.

It's a very old project and a VERY basic engine. TBH, engines like godot do a much better job, have their own scripting language (to hide away C++), but still let you write some native code if need be. For TeaScript, I'd be more interested in using it to have a more dynamic pipeline for this project, but performance there is absolutely critical since it can mean the difference between 2 minutes an 2 hours.

Does it work as a drop-in replacement? I've got this project here where I'm looking to squeeze more perf from: https://github.com/define-private-public/PSRayTracing

While working on a ray tracing implementation, I was interested in replacing out the usage of std::shared_ptr with something else. I've always been told that shared pointers are slow, and this is due to things such as reference counting. The original implementation of this ray tracer used shared pointers quite extensively in the rendering (hot path) code. I didn't want to deviate from the architecture for my implementaiton . Mostly, the pointers are passed around as const-ref objects.

Last year I was working on an implementation of the Ray Tracing in one Weekend book series. I noticed there was a fair amount of sub-optimal code in it, so I took the opportunity to rewrite parts of it to be better optimized. One of the other things I added to the CMake configuration were some compile time flags that could be toggled ON/OFF, as to use either the books code, or my code. e.g. WITH_BOOK_AABB_HIT=True, it would use the books' method AABB-Ray intersection. False, it would use my (faster) one. This allowed anyone else to download the code, toggle the change and easily see the effect it had on performance.

Here's my implementation: https://github.com/define-private-public/PSRayTracing It's different, mainly in structure (cleaner) and that it's much more performant. It also allows you to either use the code I wrote (typically faster) or the book's methods with the flip of a compiler flag.