test-results.txt

The following results were collected on a dual processor
Intel Xeon 5690 3.5 GHz system with 48G of memory.
There are 12 total cores and 24 Hyper Threads.

For hybrid runs, the following rules for deriving the number of
OMP threads (NT) and MPI or UPC processes (P) were used.

    If NP <= 8 Then P=2, NT=NP/2
    If NP >  8 Then P=4, NT=NP/4

This system is shared memory only, thus the MPI and UPC runs
are single node only.

Each test run sorted 100 million (100,000,000) random 32-bit integers.

The results are sorted on the NP=16 column.
                                               NP
                             1     2     4     8    12    16    20    24
omp_mergesort            11.83  6.54  3.76  2.46  2.44  1.92  2.00  1.98
upc_hybrid_mergesort     11.59  6.30  4.00  2.61  2.71  2.15  2.14  2.17
upc_mergesort            11.60  6.29  3.82  2.71  2.49  2.50  2.72  2.80
hybrid_mergesort         11.73 12.00  7.00  5.66  5.60  4.07  4.07  3.88
upc_no_copy_mergesort    11.59 14.90  8.32  5.06  4.01  4.24  4.33  4.16
mpi_mergesort            11.47 11.77  9.15  6.52  4.69  5.20  4.42  4.23
serial_mergesort         11.85 11.84 11.84 11.84 11.85 11.82 12.17 11.83