Searched defs:numelPerWarp (Results 1 – 1 of 1) sorted by relevance
470 const auto numelPerWarp = numelPerThread * kWarpSize; in getLaunchConfig() local520 const size_t numelPerWarp = in oneShotAllReduce() local584 size_t numelPerWarp = kBytesPerThread / input.element_size() * kWarpSize; in twoShotAllReduce() local639 size_t numelPerWarp = kBytesPerThread / input.element_size() * kWarpSize; in hybridCubeMeshAllReduce() local