Difference between revisions of "Graph500"
(→Primitive Results) |
(→Primitive Results) |
||
(3 intermediate revisions by one user not shown) | |||
Line 4: | Line 4: | ||
Graph500 ranked computers on their ability to perform complex, data-intensive analytics involved in medical research and social networks today. | Graph500 ranked computers on their ability to perform complex, data-intensive analytics involved in medical research and social networks today. | ||
+ | The intent of this benchmark problem ("Search") is to develop a compact application that has multiple analysis techniques (multiple kernels) accessing a single data structure representing a weighted, undirected graph. | ||
+ | Unfortunately this data-intensive benchmark only testing performance moving search data in the memory, not I/O. | ||
+ | |||
+ | Evidence: | ||
+ | Out of memory trying to allocate 17179869184 byte(s) | ||
+ | Aborted | ||
Line 9: | Line 15: | ||
Here post some primitive results that still needs further verifications. These results are tested on a dual core Atom D510 with 2 disks RAID0 single node machine. | Here post some primitive results that still needs further verifications. These results are tested on a dual core Atom D510 with 2 disks RAID0 single node machine. | ||
This is a very small graph size with the parameter of scale 14 and edgefacter 16 by default. | This is a very small graph size with the parameter of scale 14 and edgefacter 16 by default. | ||
+ | Testing data size is calculated as | ||
+ | power(2,scale) * edgefacter * 16 | ||
There are four binary generated: make-edgelist and other three in subfolders, omp-csr, seq-csr and seq-list. | There are four binary generated: make-edgelist and other three in subfolders, omp-csr, seq-csr and seq-list. | ||
Line 79: | Line 87: | ||
harmonic_mean_TEPS: 2.24216087879419364e+07 | harmonic_mean_TEPS: 2.24216087879419364e+07 | ||
harmonic_stddev_TEPS: 4.62228554873762114e+04 | harmonic_stddev_TEPS: 4.62228554873762114e+04 | ||
+ | |||
+ | ssdraid@ssdraid:~/graph500-2.1.4/seq-list$ ./seq-list | ||
+ | SCALE: 14 | ||
+ | nvtx: 16384 | ||
+ | edgefactor: 16 | ||
+ | terasize: 4.19430399999999992e-06 | ||
+ | A: 5.69999999999999951e-01 | ||
+ | B: 1.90000000000000002e-01 | ||
+ | C: 1.90000000000000002e-01 | ||
+ | D: 5.00000000000000444e-02 | ||
+ | generation_time: 1.59625916000000001e+00 | ||
+ | construction_time: 5.31880390000000064e-02 | ||
+ | nbfs: 64 | ||
+ | min_time: 1.60696909999999998e-01 | ||
+ | firstquartile_time: 1.60945583000000003e-01 | ||
+ | median_time: 1.61183791999999992e-01 | ||
+ | thirdquartile_time: 1.61940605250000008e-01 | ||
+ | max_time: 1.63005658999999969e-01 | ||
+ | mean_time: 1.61398998015624989e-01 | ||
+ | stddev_time: 5.82602302734929671e-04 | ||
+ | min_nedge: 2.62141000000000000e+05 | ||
+ | firstquartile_nedge: 2.62141000000000000e+05 | ||
+ | median_nedge: 2.62141000000000000e+05 | ||
+ | thirdquartile_nedge: 2.62141000000000000e+05 | ||
+ | max_nedge: 2.62141000000000000e+05 | ||
+ | mean_nedge: 2.62141000000000000e+05 | ||
+ | stddev_nedge: 0.00000000000000000e+00 | ||
+ | min_TEPS: 1.60817116171408538e+06 | ||
+ | firstquartile_TEPS: 1.61934634283171245e+06 | ||
+ | median_TEPS: 1.62650713840318588e+06 | ||
+ | thirdquartile_TEPS: 1.62919558264090167e+06 | ||
+ | max_TEPS: 1.63127592185811163e+06 | ||
+ | harmonic_mean_TEPS: 1.62417984760117414e+06 | ||
+ | harmonic_stddev_TEPS: 7.38644052798570328e+02 |
Latest revision as of 14:46, 18 October 2011
Graph500 ranked computers on their ability to perform complex, data-intensive analytics involved in medical research and social networks today. The intent of this benchmark problem ("Search") is to develop a compact application that has multiple analysis techniques (multiple kernels) accessing a single data structure representing a weighted, undirected graph. Unfortunately this data-intensive benchmark only testing performance moving search data in the memory, not I/O.
Evidence:
Out of memory trying to allocate 17179869184 byte(s) Aborted
Primitive Results
Here post some primitive results that still needs further verifications. These results are tested on a dual core Atom D510 with 2 disks RAID0 single node machine. This is a very small graph size with the parameter of scale 14 and edgefacter 16 by default. Testing data size is calculated as
power(2,scale) * edgefacter * 16
There are four binary generated: make-edgelist and other three in subfolders, omp-csr, seq-csr and seq-list.
ssdraid@ssdraid:~/graph500-2.1.4/omp-csr$ ./omp-csr SCALE: 14 nvtx: 16384 edgefactor: 16 terasize: 4.19430399999999992e-06 A: 5.69999999999999951e-01 B: 1.90000000000000002e-01 C: 1.90000000000000002e-01 D: 5.00000000000000444e-02 generation_time: 8.30255961999999959e-01 construction_time: 2.21812309000000013e-01 nbfs: 64 min_time: 8.55234900000000076e-03 firstquartile_time: 9.19533025000000143e-03 median_time: 9.86104100000000094e-03 thirdquartile_time: 1.05384902499999992e-02 max_time: 1.13258580000000011e-02 mean_time: 9.81586751562499998e-03 stddev_time: 7.43772331933209075e-04 min_nedge: 2.62141000000000000e+05 firstquartile_nedge: 2.62141000000000000e+05 median_nedge: 2.62141000000000000e+05 thirdquartile_nedge: 2.62141000000000000e+05 max_nedge: 2.62141000000000000e+05 mean_nedge: 2.62141000000000000e+05 stddev_nedge: 0.00000000000000000e+00 min_TEPS: 2.31453546389156543e+07 firstquartile_TEPS: 2.52003883657021038e+07 median_TEPS: 2.67909385396952778e+07 thirdquartile_TEPS: 2.87849155494778901e+07 max_TEPS: 3.06513450281320326e+07 harmonic_mean_TEPS: 2.67058412904128172e+07 harmonic_stddev_TEPS: 2.54945481740110496e+05
ssdraid@ssdraid:~/graph500-2.1.4/seq-csr$ ./seq-csr SCALE: 14 nvtx: 16384 edgefactor: 16 terasize: 4.19430399999999992e-06 A: 5.69999999999999951e-01 B: 1.90000000000000002e-01 C: 1.90000000000000002e-01 D: 5.00000000000000444e-02 generation_time: 1.59608408599999985e+00 construction_time: 3.38959786000000041e-01 nbfs: 64 min_time: 1.11526760000000003e-02 firstquartile_time: 1.16040705000000008e-02 median_time: 1.17280435000000002e-02 thirdquartile_time: 1.18378765000000004e-02 max_time: 1.20137560000000004e-02 mean_time: 1.16914447343750005e-02 stddev_time: 1.91305940566682196e-04 min_nedge: 2.62141000000000000e+05 firstquartile_nedge: 2.62141000000000000e+05 median_nedge: 2.62141000000000000e+05 thirdquartile_nedge: 2.62141000000000000e+05 max_nedge: 2.62141000000000000e+05 mean_nedge: 2.62141000000000000e+05 stddev_nedge: 0.00000000000000000e+00 min_TEPS: 2.18200702594592385e+07 firstquartile_TEPS: 2.22079344693168029e+07 median_TEPS: 2.23580749660745859e+07 thirdquartile_TEPS: 2.26550392355549298e+07 max_TEPS: 2.35047624444572777e+07 harmonic_mean_TEPS: 2.24216087879419364e+07 harmonic_stddev_TEPS: 4.62228554873762114e+04
ssdraid@ssdraid:~/graph500-2.1.4/seq-list$ ./seq-list SCALE: 14 nvtx: 16384 edgefactor: 16 terasize: 4.19430399999999992e-06 A: 5.69999999999999951e-01 B: 1.90000000000000002e-01 C: 1.90000000000000002e-01 D: 5.00000000000000444e-02 generation_time: 1.59625916000000001e+00 construction_time: 5.31880390000000064e-02 nbfs: 64 min_time: 1.60696909999999998e-01 firstquartile_time: 1.60945583000000003e-01 median_time: 1.61183791999999992e-01 thirdquartile_time: 1.61940605250000008e-01 max_time: 1.63005658999999969e-01 mean_time: 1.61398998015624989e-01 stddev_time: 5.82602302734929671e-04 min_nedge: 2.62141000000000000e+05 firstquartile_nedge: 2.62141000000000000e+05 median_nedge: 2.62141000000000000e+05 thirdquartile_nedge: 2.62141000000000000e+05 max_nedge: 2.62141000000000000e+05 mean_nedge: 2.62141000000000000e+05 stddev_nedge: 0.00000000000000000e+00 min_TEPS: 1.60817116171408538e+06 firstquartile_TEPS: 1.61934634283171245e+06 median_TEPS: 1.62650713840318588e+06 thirdquartile_TEPS: 1.62919558264090167e+06 max_TEPS: 1.63127592185811163e+06 harmonic_mean_TEPS: 1.62417984760117414e+06 harmonic_stddev_TEPS: 7.38644052798570328e+02