// vector from the input, j is the second, and k is the result index. This
// parallelizes over the range of k and infers what i and j are from the
// value of k.
- parallel_for(0, combs, internal::GRAIN_SIZE / (16 * m), [p, self_start, self_end, n, m, res_start, combs](int64_t k, int64_t end) {
+ parallel_for(0, combs, internal::GRAIN_SIZE / (16 * m), [p, self_start, self_end, n, m, res_start](int64_t k, int64_t end) {
const Vec pvec(p);
double n2 = n - .5;
// The -1 accounts for floating point truncation issues