tiny.simd.sumsOf8AbsDiff

1pub fn sumsOf8AbsDiff(2    comptime D: type,3    a: D.Vector,4    b: D.Vector,5) D.repartition(wideLane(wideLane(wideLane(D.Lane)))).Vector {6    requireByteLane(D, "sumsOf8AbsDiff");7    const W = wideLane(wideLane(wideLane(D.Lane)));8    const R = D.repartition(W);9    if (D.lane_count < 8) @compileError("sumsOf8AbsDiff requires at least eight lanes");10    var result: R.Vector = undefined;11    inline for (0..R.lane_count) |index| {12        var total: W = 0;13        inline for (0..8) |offset| {14            total += absoluteDifference(W, a[index * 8 + offset], b[index * 8 + offset]);15        }16        result[index] = total;17    }18    return result;19}

1pub const sumsOf8AbsDiff = reduce.sumsOf8AbsDiff;