I would like to perform a summation operation on a polynomial inside a cuda kernel which contains coefficients and function as given

y = a1f1(x) + a2f2(x) + a3f3(x) + a4f4(x);

My first idea is to loop over the three terms and use switch statements

float y = 0;
    for (int i = 0; i < 4; i++) {
        if (a[i] != 0) {

            switch (i) {
            case 0: y += a[i] * f1(x); break;
            case 1: y += a[i] * f2(x); break;
            case 2: y += a[i] * f3(x); break;
            case 3: y += a[i] * f4(x); break;
            }
        }
    }

The above solution seems to be very basic and slow when number of terms in equation can be greater than 30. Now, I am thinking of creating an array of functions to remove switch statement.

float y = 0;
    for (int i = 0; i < 4; i++) {
        if (a[i] != 0) {
            y += a[i] * f[i](x);
        }
    }```





still, I feel it can be better optimized. Let me know if you guys have a better approach to the above problem.

Thanks!