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 module

apple_accelerate

Wraps the Apple Accelerate cblas_sgemm routine to provide single-precision matmul, GEMV, and batched matmul kernels for Apple CPUs.

comptime values

APPLE_ACCELERATE

comptime APPLE_ACCELERATE = _Global[StringSpan("APPLE_ACCELERATE"), _init_dylib, _on_error_msg]

Process-global lazily loaded handle to the Accelerate dynamic library.

cblas_gemm_type

comptime cblas_gemm_type = def(_CBLASOrder, _CBLASTranspose, _CBLASTranspose, Int32, Int32, Int32, Float32, Pointer[Float32, ImmutAnyOrigin], Int32, Pointer[Float32, ImmutAnyOrigin], Int32, Float32, Pointer[Float32, MutAnyOrigin], Int32) thin -> None

Function type matching the Accelerate cblas_sgemm C signature.

LIB_ACC_PATH

comptime LIB_ACC_PATH = "/System/Library/Frameworks/Accelerate.framework/Accelerate"

Filesystem path to the Apple Accelerate framework dynamic library.

Functions

• ​apple_batched_matmul: Performs a batched single-precision matmul via the Apple Accelerate library.
• ​apple_gemv: Performs a parallelized and vectorized GEMV for the M=1 case on Apple CPUs.
• ​apple_matmul: Performs a single-precision matmul via a pre-loaded cblas_sgemm function pointer.
• ​get_cblas_f32_function: Loads and returns the cblas_sgemm function pointer from the Apple Accelerate library.
• ​use_apple_accelerate_lib: Returns whether the Apple Accelerate CBLAS library should be used for this dtype combination.