bfs verification fails on kronecker graphs scale >= 16 about gapbs HOT 5 CLOSED

Thanks for the quick response. No, I unzipped and hit "make" - no changes. Besides, PrintStep is doing exactly what it should in the "Source" call - even though you do a cast to int64 in the benchmark code, PrintStep is called with two arguments and promptly changes the second argument back to a double. No worries there. I modified the Makefile just now to do no optimization and include debugging info. I added some more info to the reachability error and got ./bfs -g16 -n1 -v Generate Time: 0.99920 Build Time: 0.06588 Graph has 65536 nodes and 909645 undirected edges for degree: 13 Source 13098.00000 i 0.00040 td 23 0.00006 td 7636 0.00015 e 0.00005 bu 36959 0.00101 bu 2062 0.00036 c 0.00022 td 7 0.00004 td 0 0.00000 Trial Time: 0.00261 Reachability mismatch (u = 50186, depth[u] -1 != parent[u] -2) Verification: FAIL Verification Time: 0.02534 Average Time: 0.00261 If you point me in the right direction, I might be able to help you debug...again, thanks for all your work on this. Kirk

Woo hoo! Great job! You want to close the issue or shall I? FYI - I'm planning to use your system to create test data, baseline performance metrics, and create a correctness oracle for graph processing hardware accelerators and compilers developed under the DARPA HIVE program (for which we are the test contractor). I will most likely add some streaming capabilities to the generators and develop bridges into TinkerPop/Gremlin for the larger graphs (although I've got some 384G RAM machines queued up, so we'll see). Your license file and references will be prominently displayed - thanks! Kirk Dunkelberger, PhD Purdue 86