~/talk

Cyfra: Scala on GPU

Talkat Scala Days 2026 · 2026-10-12..10-13· [scala, machine-learning, webassembly]
links:none on file
links:none on file
slides:none on file

Cyfra is a Scala GPGPU framework that compiles plain Scala data structures and higher‑order functions to a GPU IR and runs on a Vulkan runtime with backends for AMD, NVIDIA, Apple and Intel. It exposes GFunction and GProgram/GExecution models, a GSequence abstraction, a GIO monad for mutable shader code and supports shared memory via SPIR‑V. Real projects include a ray tracer and an LLM inference engine that composes 200+ GPU programs and runs a 3B‑parameter LLaMA, beating TornadoVM by over 2x and trailing llama.cpp.

AI-written summary · about

  1. 0:00Introduction and motivation
  2. 2:21GPU programming challenges
  3. 4:19GPU landscape and trade-offs
  4. 7:02Getting started with Cyfra
  5. 8:33Cyfra features and syntax
  6. 12:39Architecture and design
  7. 19:43Real-world applications
  8. 22:57Inference engine performance
  9. 24:52Q&A and conclusion
~/search
type to search 6,518 conferences, 143,918 speakers and 192,133 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
← →
on a talk: 5 seconds back / forward
0–9
on a talk: jump to 0%–90% of the recording
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.