Intel Threading Building Blocks

Advanced example

An example of a more advanced associative operation is to find the index containing the smallest element of an array. A serial version might look like Example 3-12.

Example 3-12. Original minimization code

long SerialMinIndexFoo( const float a[], size_t n ) {
    float value_of_min = FLT_MAX;        // FLT_MAX from <climits>
    long index_of_min = -1;
    for( size_t i=0; i<n; ++i ) {
        float value = Foo(a[i]);
        if( value<value_of_min ) {
              value_of_min = value;
            index_of_min = i;
        }
    }
    return index_of_min;
}

The loop works by keeping track of the minimum value found so far, and the index of this value. This is the only information carried between loop iterations. To convert the loop to use parallel_reduce, the function object must keep track of the information carried, and must be able to merge this information when iterations are spread across multiple threads. Also, the function object must record a pointer to a to provide context.

Example 3-13 shows the complete function object.

Example 3-13. Function object for minimization

class MinIndexFoo { const float *const my_a; public: float value_of_min; long index_of_min; void operator()( const blocked_range<size_t>& r ) { const float *a = my_a; for( size_t i=r.begin(); i!=r.end(); ++i ) { float value = Foo(a[i]); if( value<value_of_min ) { value_of_min = value; index_of_min = i; } } } MinIndexFoo( MinIndexFoo& x, split ) : my_a(x.my_a), value_of_min(FLT_MAX), // FLT_MAX from <climits> index_of_min(-1) {} void join( const SumFoo& y ) { if( y.value_of_min<value_of_min ...