For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).
Mojo package
backend
Implements the per-target halves of the data-parallel algorithms.
The entry points in the parent package select CPU or GPU at compile time and forward into the matching subpackage here. Reach for those entry points rather than these modules, whose split tracks the dispatch and shifts with it.
Host thread-pool parallelism has no GPU counterpart, so parallelize() and its
siblings reach callers only through the CPU side; the GPU side additionally
carries block- and row-level reductions, for kernel authors composing a
reduction inside a larger kernel.