blas 3 matrix multiplication kernel build error when called from ArrayFire

I am indirectly calling the clBLAS gemm function from ArrayFire. I found that ArrayFire's matmul function crashes when it calls clBLAS's gemmS function. So, I activated build error log in clBLAS (https://github.com/clMathLibraries/clBLAS/blob/54e949e2db8ee49a170b1bbc829ceed07f16533f/src/library/blas/generic/common.c) and noticed there is build error in clBLAS.

When I run the matrix multiplication example in clBLAS repo directly (https://github.com/clMathLibraries/clBLAS/blob/54e949e2db8ee49a170b1bbc829ceed07f16533f/src/samples/example_sgemm.c), everything works just fine.

What could be an issue here? (I am posting this as an ArrayFire issue as well.)

========================================================

AN INTERNAL KERNEL BUILD ERROR OCCURRED!
device name = HD Graphics 4000
error = 0
memory pattern = Cached global memory based block gemm, computing kernel generator
Subproblem dimensions: dims[0].itemY = 8, dims[0].itemX = 8, dims[0].y = 8, dims[0].x = 8, dims[0].bwidth = 1; ; dims[1].itemY = 1, dims[1].itemX = 1, dims[1].y = 1, dims[1].x = 1, dims[1].bwidth = 1; ; 
Parallelism granularity: pgran->wgDim = 2, pgran->wgSize[0] = 8, pgran->wgSize[1] = 8, pgran->wfSize = 64
Kernel extra flags: 31508
Source:

typedef union GPtr {
    __global float *f;
    __global float2 *f2v;
    __global float4 *f4v;
    __global float8 *f8v;
    __global float16 *f16v;
} GPtr;

typedef union LPtr {
    __local float *f;
    __local float2 *f2v;
    __local float4 *f4v;
    __local float8 *f8v;
    __local float16 *f16v;
} LPtr;

typedef union PPtr {
    float *f;
    float2 *f2v;
    float4 *f4v;
    float8 *f8v;
    float16 *f16v;
} PPtr;

__attribute__((reqd_work_group_size(8, 8, 1)))
void __kernel
sgemmBlock(
    uint M,
    uint N,
    uint K,
    const float alpha,
    const float beta,
    const __global float *restrict A,
    const __global float *restrict B,
    __global float *C,
    uint lda,
    uint ldb,
    uint ldc)
{
    float a0;
    float b0;
    float c0;
    uint4 coord = 0u; /* contains coordB, coordA, k */

    uint kif;
    uint get_group_id_1;
    uint get_global_id_1;
    A += (uint)get_global_id(0);
    get_group_id_1 = (get_group_id(0) + get_group_id(1))% get_num_groups(1);
    get_global_id_1 = get_group_id_1 * get_local_size(1) + get_local_id(1);
    kif = (N % 512 != 0);
    get_global_id_1 = (kif*(uint)get_global_id(1)) + ((1-kif)*get_global_id_1);
    B += get_global_id_1;
    coord.y = 1u * (uint)get_global_id(0);
    coord.x = 1u * (uint)get_global_id_1;
    if ((coord.y >= M) || (coord.x >= N)) {
        return;
    }

    c0 = 0;

    for (uint k1 = 0; k1 < K; k1 += 1) {
        /* -- Tiles multiplier -- */
        b0 = B[0];

        a0 = A[0];

        c0 += a0 * b0;

        A += lda;
        B += ldb;
        /* ---------------------- */
    }

    GPtr uC;

    uC.f = C + coord.x * ldc + coord.y;

    __global float *pC = uC.f;

    float tempC0;

    tempC0 = c0 * alpha + 0;
    pC[0] = tempC0;
}

--------------------------------------------------------

Build log:

========================================================

The program has unexpectedly finished.

clMathLibraries / clBLAS

blas 3 matrix multiplication kernel build error when called from ArrayFire #59