A new article about TeaVM

[konsoletyper]

This article compares TeaVM and Kotlin/JS performance using as example a software 3D renderer: https://teavm.hashnode.dev/comparing-teavm-with-kotlinjs-peformance-with-software-3d-renderer

[bertung]

Interesting article Alex, I have a mac pro 2013, 12 Core, 64GB ram, AMD Fire pro dedicated GPU . My results show on all modern browsers I tested, Kotlin produced code is always 2x slower than teavm/JS and teavm/wasm. On chrome, teavm/wasm is the clear winner. I will wager that it may be Graphics card related. Graphics based benchmarking will always prove wrong. You do not mention your graphics card on your article. It should be the culprit.

[konsoletyper]

@bertung, I don't think graphics card is involved here. The 3D renderer, presented in the article, utilizes only CPU. That was the goal: to write a software 3D renderer to compare performance of different compilers.

[bertung]

Well, you still utilize GPU to write that chunk of data to the graphics card to display something even though you do not use WebGL.

[konsoletyper]

@bertung you are right, but I only measure that part of calculations that performs rasterization in main memory. Further calculations related to passing buffers between main script and worker fall into other metric. So you should not take into account "Average frame rendering time, microseconds". Instead, look into "Average by worker" metric, which is a primary metric used in "Benchmark" section in the article.