[DataConverter] If SSE2 is not available, mul_const is incorrect

[btgoodwin]

In mathOptimizations.c, the ifdef __SSE2__ is within the if statement but its corresponding #endif is outside the closing brace of that if statement. If __SSE2__ is not defined, this will result in the following error:

| mathOptimizations.c: In function 'mul_const':
| mathOptimizations.c:2421:1: error: expected declaration or statement at end of input
|  }
|  ^

[btgoodwin]

Judging by the surrounding logic, this method may actually have a performance bug as well. There is a loop after the endif that seems to re-multiply and re-store the samples. It's likely the code should be:

  void mul_const(float* c, const float *a, const float b, int n){
#ifdef HAVE_IPP
    //(void)ippsRealToCplx_32f((const Ipp32f*) r,(const Ipp32f*) c, (Ipp32fc*)rc, elements);
#else
    int k=0;
    __m128 b1 = _mm_load1_ps(&b);
#ifdef __SSE2__
    if ( ALIGNED2(c,a) && stack_simd_aligned() ) {
        for( ; k+3<n;k+=4){
            __m128 a1 = _mm_load_ps((const float*)(a+k));
              a1 = _mm_mul_ps(a1,b1);
            _mm_store_ps((float*)(c+k),a1);
        }
    }
#else
    if (k<(n-1)){
        for(;k+1<n;k+=2){
            c[k] = a[k] * b;
        }
    }
#endif
#endif
}

Note the movement of the #ifdef __SSE2__ to ahead of the alignment check (which would be required for SSE), and the change of the corresponding #endif to #else followed by another #endif to close out the __SSE2__ check.

[RedhawkDeployer]

We have confirmed the issue and have put the issue on our backlog for scheduling.