vbmatmul2x4+ ( a b accumulator -- result )

Vocabulary

math.vectors

Inputs

a	a sequence
b	a sequence
accumulator	a sequence

Outputs

result

a sequence

Word description

Multiplies a row-major 2 by 4 BF16 matrix A by a column-major 4 by 2 BF16 matrix B, adding four row-major float-4 accumulators. Each output applies two adjacent-pair BFDotAdd operations in order.

Definition

IN: math.vectors

GENERIC: vbmatmul2x4+ ( a b accumulator -- result )

Methods

USING: kernel math.vectors math.vectors.simd
math.vectors.simd.extensions
math.vectors.simd.extensions.private ;

M: simd-128 vbmatmul2x4+
    3dup check-bfloat-accumulate (vbmatmul2x4+) ;