`fastmath` option is not supported

[akharche]

Numba keyword argument fastmath is not supported in @dppy.kernel. It can cause performance degradation in some cases.

import dpctl
import numba_dppy
import numpy as np
import numpy.random as rnd

import dpctl.tensor as dpt

work_dtype = np.float64

SEED_TEST = 777777
SEED_TRAIN = 0


def __gen_data(ip_size, dtype, seed, dim):
    rnd.seed(seed)
    data = rnd.rand(ip_size, dim)
    return data.astype(dtype)


@numba_dppy.func
def get_dist(arr1, arr2, dims):
    result = 0
    for i in range(dims):
        result += (arr1[i] - arr2[i])*(arr1[i] - arr2[i])

    return result


@numba_dppy.kernel(fastmath=True)
def dist_kernel(test_data,test_size,
                    train_data, train_size,
                    distance_m,
                    dim):

    g_test_id = numba_dppy.get_global_id(0)
    g_train_id = numba_dppy.get_global_id(1)

    if (g_train_id < train_size and g_test_id < test_size):
    
        test_v = test_data[g_test_id]
        train_v = train_data[g_train_id]

        dist = 0
        for i in range(dim):
            dist += (train_v[i] - test_v[i])*(train_v[i] - test_v[i])

        dist_idx = g_train_id 
        distance_m[g_test_id * train_size + g_train_id] = dist


def run_kernel(
    train_data,
    test_data, 
    ntrain, 
    ntest, 
    dim, 
):

    dist_matrix = np.empty(ntest * ntrain, dtype=work_dtype)

    gws = [ntest, ntrain]
    lws = [2, 2]


    with dpctl.device_context("opencl:gpu") as gpu_queue:
        dist_kernel[gws, lws](test_data, ntest,
                       train_data, ntrain,
                        dist_matrix,
                        dim)
 
def main():
    dim = 16
    ntrain = 32
    ntest = 8

    test = __gen_data(ntest, work_dtype, SEED_TEST, dim)
    train = __gen_data(ntrain, work_dtype, SEED_TRAIN, dim)

    run_kernel(train, test, ntrain, ntest, dim)


if __name__ == "__main__":
    main()

The code runs with the error: TypeError: kernel() got an unexpected keyword argument 'fastmath'