@SeanSiddens

researcher @ amd, cs phd @ uw

Seattle, WA
Joined June 2020
Fast Kernels on AMD hardware!! Great work from @simran_s_arora @_williamhu @Drewwad and team on HipKittens (github.com/HazyResearch/HipK…).
AI has been built on one vendor’s stack for too long. AMD’s GPUs now offer state-of-the-art peak compute and memory bandwidth — but the lack of mature software / the “CUDA moat” keeps that power locked away. Time to break it and ride into our multi-silicon future. 🌊 It's been a blast working with the amazing @_williamhu, @Drewwad and team; we present HipKittens!
3
17
1
149
15,654
AI has been built on one vendor’s stack for too long. AMD’s GPUs now offer state-of-the-art peak compute and memory bandwidth — but the lack of mature software / the “CUDA moat” keeps that power locked away. Time to break it and ride into our multi-silicon future. 🌊 It's been a blast working with the amazing @_williamhu, @Drewwad and team; we present HipKittens!
14
95
20
581
229,337
From the intro of The Connection Machine (1985): "With almost no apparent difficulty a person is able to say something like 'It is a group of people and horses'. This is easy for us...Yet for a modern digital computer it is an almost impossible task."
198
Had a great time at Vulkanised! Talks are here: vulkan.org/events/vulkanised… My 2 takeaways: 1. There is huge opportunity for GPU programming education using Vulcan (and @WebGPU) 2. we need more GPU compute using Vulkan Huge thanks to the organizers: @thekhronosgroup @VulkanAPI
2
4
23
1,858
GPU Atomic Performance Modeling with Microbenchmarks, Vulkanised 2024 presentation slides vulkan.org/user/pages/09.eve…
21
95
6,550
Starting work on a ray tracer. Currently just have a basic camera system for shooting initial rays, glTF model load and BVH construction, and a UI. Right now ray tracing is on the CPU, writing to a Vulkan image which is displayed by ImGui. Next step is moving entirely to the GPU
3
365
Twitter video compression destroyed the original so trying again: turned some job-market-stress induced lack of sleep into a dynamical system integrator. You can play with it here! stevejtrettel.site/code/stra…
10
58
5
481
Real-time diffusion-limited aggregation with 256,000 particles
2
Using OpenGL compute shaders to update the positions of random walkers every frame. The random computed using a 2D hash function (pcg2d) from here: jcgt.org/published/0009/03/0…. Hashing the thread's global invocation ID and current frame seems to do the trick.
1
1
2
Can simulate way more walkers at a much higher resolution but I was having a hard time recording a GIF that would upload to Twitter
1
Some renders from a CPU path tracer written in C following @Peter_shirley's Ray Tracing in One Weekend book series. My next goal is to try to get it working on the GPU and get it to render triangles instead of just spheres. github.com/seansiddens/PathT…
3
3
1
48