Kernel optimization
Antinuous writes kernel variants, compiles them, and benchmarks each one on the same inputs. It checks every variant's output against a reference before it compares speed.
A GPU die with twelve compute units coloured by how busy they are, a cache band, and memory stacks at each side
Use cases
Review published methods for a kernel
Methods for a kernel are spread across papers, repositories, and vendor notes.
You ask AntinuousFind papers and open-source implementations of fused attention kernels. Summarize each method, the hardware it targets, and where its code lives.
- What it does
- Antinuous searches arXiv, the web, and GitHub for papers and implementations.
- It reads the full papers and the linked code.
- It records each method with its target hardware and a source link.
- You get
A table with one cited row per method.
How each method is added to the table.
- Search papers and code
- Read each method
- Note the target hardware
- Link the source
Who it's for
- Performance engineers
Benchmark many kernel variants on the same inputs.
- ML infrastructure engineers
Compare kernel options before you ship one.
- Compiler engineers
Test code generation settings across many workloads.
Works with
GitHub
arXiv
SlackMore in compute and hardware
A code change with removed lines in red and added lines in green, beside a branch graph
Software engineering
Run your kernel benchmarks on Antinuous.
Start with our agent, or connect your own to the same compute, orchestration, and simulation.