NDRnage Kernels Global Memory Write Pattern

Here is the kernel I compile:

__attribute__((num_compute_units(5)))
__attribute__((num_simd_work_items(16)))
__attribute__((reqd_work_group_size(256,1,1)))
__kernel void WGSXMAPIXLLXOPS16(const __global float * restrict GIn, __global float * restrict GOut, const float M, const float N, const float P) {
    const int XGL = get_global_id(0);
    const int XGRid = get_group_id(0);
    const int XGRnum = get_num_groups(0);
    const int XLSize = get_local_size(0);
    const int XLid = get_local_id(0);
    // Just a private variable
    float MF = (float) M;
    float NF = (float) N;
    float PF = (float) P;
    // Start of a new level of for loop
    long baseIndex1 = XGRid*XLSize*2+XLid;
    float temp1 = 1.0;
    float temp2 = 1.0;
    float temp3 = 1.0;
    float temp4 = 1.0;
    float tempOut;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    temp1 += temp1 * MF;
    tempOut = temp1 + temp2 + temp3 + temp4;
    GOut = tempOut;
}