Write Mojo GPU code without CUDA assumptions.
What it does
Guides agents in writing Mojo GPU code using the correct kernel, memory, and launch conventions.
When to use it
Use when writing Mojo code for GPUs or other accelerators, especially to correct misconceptions about its GPU syntax.
How to use it
It changes how the agent writes Mojo GPU code by supplying language-specific conventions instead of CUDA assumptions.
What you provide
- An existing project
What you need · 4
The programming guidance and examples use Mojo; no language version is specified.
The examples import GPU indexing, synchronization, memory, and host APIs from max.gpu.
The tensor layout APIs come from the separate layout package, not the standard library.
Running the GPU examples requires an accelerator; the complete examples assert its availability.
Access · 0
No additional actions listed.