Hi,
I am using libint2 in my quantum chemistry program. Until now, I have developed it using gcc.
Recently, I have compiled everything using the intel compiler (version 19.1.1.217).
It can compute 1-body integrals correctly. However, when I try to compute 2 electron repulsion integrals, it only spits out zeros.
It reminded me of issue #118, but the intel bug seems to have been fixed. Disabling constexpr does not seem to help.
I am computing the integrals in an openmp parallel environment, similar to mpqc, creating a thread pool with intel tbb, and then accessing the local engine.
#pragma omp parallel
{
auto& loc_eng = engine->local();
const auto &results = loc_eng.results();
...
loc_eng.compute(sh1,sh2,sh3);
auto ints_shellsets = results[0];
if (ints_shellsets != nullptr) {
size_t idx = 0;
for (int i = 0; i != sh1.size(); ++i) {
for (int j = 0; j != sh2.size(); ++j) {
for (int k = 0; k != sh3.size(); ++k) {
blk(i + locblkoff1, j + locblkoff2, k + locblkoff3) // shellblock object
= ints_shellsets[idx++]; // <- only outputs zeroes.
}}}
}//endif shellsets
....
}
Hi, I am using libint2 in my quantum chemistry program. Until now, I have developed it using gcc. Recently, I have compiled everything using the intel compiler (version 19.1.1.217).
It can compute 1-body integrals correctly. However, when I try to compute 2 electron repulsion integrals, it only spits out zeros.
It reminded me of issue #118, but the intel bug seems to have been fixed. Disabling constexpr does not seem to help.
I am computing the integrals in an openmp parallel environment, similar to mpqc, creating a thread pool with intel tbb, and then accessing the local engine.
Any help would be appreciated.