正在加载视频...

视频加载失败

So I updated "Water" ! The sim is running in a single CPU thread. It should work faster now, rendering optimized & WebAssembly SIMD support! But seems like the speedup with wasm is only limited because of the heavily scattered memory access pattern 😇

40,920 次观看 • 1 年前 •via X (Twitter)

10 条评论

saharan / さはら 的头像
saharan / さはら1 年前

context: also the source code updated. e.g. you can see the wasm part here:

SoyDegurime 的头像
SoyDegurime1 年前

a SINGLE CPU thread? How is this so well optimized?

saharan / さはら 的头像
saharan / さはら1 年前

Well, I tried hard! See also:

matsuoka-601 的头像
matsuoka-6011 年前

自分もいま MLS-MPM の AVX2 化に取り組んでいるので,参考になります! このシミュレーションについて,いくつか知りたい点があります.

Elliot Walmsley 的头像
Elliot Walmsley1 年前

The velocity needs to be constant wrt to frame time. The sim shouldn't slow down on higher particle count, the FPS should just lower.

saharan / さはら 的头像
saharan / さはら1 年前

Ideally so, but in this sim the spatial resolution is fixed and instead the view is scaled down. So physically you "just have more particles" in higher settings and nothing else is changed (except the boundaries). Have to subdivide time too when aim to a pure resolution increase.

Perro Faldero 的头像
Perro Faldero1 年前

Impressive! One question: Is it on purpose that it does not really behave like water? Can one tweak parameters to make it more realistic? OK, two qustions: Can it be extended to 3D? Thanks!

saharan / さはら 的头像
saharan / さはら1 年前

It's more like "ocean" than water considering all the parameters set. Maybe that's why you feel it a bit off. It's not on purpose, and increasing gravity would make it more realistic, at an additional computation cost. For the second question: Yes!

Lazarz 的头像
Lazarz1 年前

dope

Gustav Ekerot 的头像
Gustav Ekerot1 年前

insane! so cool, well done

相关视频