Jeremy Laumon

jeremy.laumon.name

15 September 2026 16:33

(I read the entire thing without your writing style coming to mind, so I'd say it's good! 😄)

Jeremy Laumon

jeremy.laumon.name

15 September 2026 16:33

Interesting to hear the CPU performance is better nowadays! I've also heard warnings about the GPU performance (concensus seems to be don't use for render targets, and maybe also not large amount of regular textures) but it feels like maybe you found a use case where it should be fine (?)

Christoph Kubisch

pixeljetstream.bsky.social

13 September 2026 11:56

Reads well. Though would add a remark that the 64 KB page sizes are less than what is deemed optimal (2 MB for NV) docs.nvidia.com/cuda/cuda-pr... Meaning there can be a performance impact on the GPU memory system depending on how the buffer is used.

Bruce "86 47" Dawson

randomascii.bsky.social

11 May 2026 03:10

I like to call this Dawson’s first law of computing: O(n^2) is the sweet spot of badly scaling algorithms: fast enough to make it into production, but slow enough to make things fall down once it gets there.

Krzysztof Narkowicz

knarkowicz.bsky.social

09 March 2026 15:59

Interesting paper showing great results for replacing standard low-res rendering + temporal upsampling (DLSS) with high-res GBuffer + adaptive sampling. Makes sense given rising costs of lighting computations vs GBuffer. arxiv.org/abs/2602.08642