For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo function

vpdpwssd

def vpdpwssd[width: SIMDLength, a_type: DType, b_type: DType, c_type: DType](src: SIMD[c_type, width], a: SIMD[a_type, (width * SIMDLength(2))], b: SIMD[b_type, (width * SIMDLength(2))]) -> SIMD[c_type, width]

Computes a multiply-accumulate of signed 16-bit integers using the VPDPWSSD Intel AVX-512 VNNI instruction.

Multiplies pairs of adjacent signed 16-bit integers from a and b, accumulates the 32-bit products into src, and returns the result.

Parameters:

width (SIMDLength): Number of int32 output elements (4, 8, or 16).
a_type (DType): DType of the a operand (int16).
b_type (DType): DType of the b operand (int16).
c_type (DType): DType of the accumulator; must be int32.

Args:

src (SIMD[c_type, width]): Int32 accumulator vector.
a (SIMD[a_type, (width * SIMDLength(2))]): Int16 input vector, twice the output width.
b (SIMD[b_type, (width * SIMDLength(2))]): Int16 input vector, twice the output width.

Returns:

SIMD[c_type, width]: Updated int32 accumulator after multiply-add.