peacock-data-public-datasets-idc-temp-code
/
cc-multilingual-main
/cc_net
/third_party
/kenlm
/util
/parallel_read.hh
/* Read pieces of a file in parallel. This has a very specific use case: | |
* reading files from Lustre is CPU bound so multiple threads actually | |
* increases throughput. Speed matters when an LM takes a terabyte. | |
*/ | |
namespace util { | |
void ParallelRead(int fd, void *to, std::size_t amount, uint64_t offset); | |
} // namespace util | |