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 function

enqueue_apple_gemv_config

def enqueue_apple_gemv_config[c_type: DType, in_type: DType, //, *, tile_m: Int, rows_per_warp: Int, tile_k: Int = Int(8), unroll: Int = Int(2), warps_per_block: Int = Int(8), elementwise_lambda_fn: Optional[def[dtype: DType, width: SIMDLength, *, alignment: Int = Int(1)](IndexList[Int(2)], SIMD[dtype, width]) capturing thin -> None] = None](c: TileTensor[c_type, Engine=c.Engine, address_space=c.address_space, linear_idx_type=c.linear_idx_type], a: TileTensor[in_type, Engine=a.Engine, address_space=a.address_space, linear_idx_type=a.linear_idx_type], weight: TileTensor[in_type, Engine=weight.Engine, address_space=weight.address_space, linear_idx_type=weight.linear_idx_type], ctx: DeviceContext)

Enqueues apple_gemv_kernel with an explicit launch configuration.

Parameters:

• ​c_type (DType): Output element type. Accumulation is fp32.
• ​in_type (DType): Activation and weight element type (bf16 or fp16).
• ​tile_m (Int): Largest activation row count the launch supports.
• ​rows_per_warp (Int): Weight rows (output columns) per warp.
• ​tile_k (Int): Elements per lane per K chunk.
• ​unroll (Int): K chunks per loop iteration.
• ​warps_per_block (Int): Warps per threadgroup.
• ​elementwise_lambda_fn (Optional[def[dtype: DType, width: SIMDLength, *, alignment: Int = Int(1)](IndexList[Int(2)], SIMD[dtype, width]) capturing thin -> None]): Optional epilogue applied to each output.

Args:

• ​c (TileTensor[c_type, Engine=c.Engine, address_space=c.address_space, linear_idx_type=c.linear_idx_type]): Output [M, N].
• ​a (TileTensor[in_type, Engine=a.Engine, address_space=a.address_space, linear_idx_type=a.linear_idx_type]): Activation [M, K] with M <= tile_m and K % tile_k == 0.
• ​weight (TileTensor[in_type, Engine=weight.Engine, address_space=weight.address_space, linear_idx_type=weight.linear_idx_type]): Weight [N, K].
• ​ctx (DeviceContext): Device context to enqueue on.