gnu(7104)=13094

[markuspf]

GAP e403849e47a7bdf008c9406a15912559fe2e9f8c
GrpConst 2.5
Runtime of ConstructAllGroups: 12430708 ms
Returned a list of length 13091 containing 13088 groups and 3 sublists
Length of sublists : [ 2, 2, 2 ]
Runtime to check non-isomorphism of groups in pairs: 151512400 ms

[olexandr-konovalov]

Thank you for submitting new gnu(n). We will recalculate it and add to the database if we get the same result.

Checklist for the reviewer:

• checked and assigned labels "new (n<50000)" or "recomputation", if needed
• set "replication needed" label (unless it starts immediately)
• changed it to "replication in progress" label when the replication has been started
• changed it to "replicated" after successful replication review
• reproducible value of gnu(n)
• no essential runtime differences
• sufficient level of technical details

Checklist for the maintainer:

• correct arguments of commands to store data
• correct location of new line in the data file
• mention reviewer in the commit message

[olexandr-konovalov]

gnu(7104)=13094

GAP 4.8.3
GrpConst 2.5
Runtime of ConstructAllGroups: 9092465 ms (2.53 hrs)
Returned a list of length 13091 containing 13088 groups and 3 sublists
Length of sublists : [ 2, 2, 2 ]
Runtime to check non-isomorphism of groups in pairs: 1457885 ms (24.28 min)

Replicated by @alex-konovalov on Ubuntu 16.04 on Azure cloud standard DS13 v2 instance (8 cores, 56 GB memory)

[olexandr-konovalov]

With GAP 4.8.3, non-isomorphism checked in less than half an hour, and with master branch - over more than two days???

[olexandr-konovalov]

@markuspf can you try also this example with the new state of the master branch, please?

[olexandr-konovalov]

@markuspf: I've rerun this in the current master branch: now the runtime of the non-isomorphism check is 5.52 hours - not 2+ days as observed in your computation, but also not half an hour, as it was in mine:

gnu(7104)=13094

GAP 8552b72a9f49dffb8b9e50d0b68df388d50e37f2
GrpConst 2.5
Runtime of ConstructAllGroups: 8438075 ms (2.34 hrs)
Returned a list of length 13091 containing 13088 groups and 3 sublists
Length of sublists : [ 2, 2, 2 ]
Runtime to check non-isomorphism of groups in pairs: 19867633 ms (5.52 hrs)

[olexandr-konovalov]

My plan is to keep this open and first isolate the problem of checking the non-isomorphism from groups construction and the compare the performance. Since we have seen by now that even the pattern of sublist lengths can be different, we can not be sure that we actually the groups in these three pairs are actually the same. I will calculate them again, save into file(s), check isomorphism across outputs produced with different versions of GAP, and then compare the performance of the master branch with GAP 4.8.

[olexandr-konovalov]

Another replication with a different version of GAP:

gnu(7104)=13094

GAP 4.8.4
GrpConst 2.5
Runtime of ConstructAllGroups: 14798774 ms (4.11 hrs)
Returned a list of length 13091 containing 13088 groups and 3 sublists
Length of sublists : [ 2, 2, 2 ]
Runtime to check non-isomorphism of groups in pairs: 4053532 ms (1.13 hrs)

Replicated by @alex-konovalov on Linux server (AMD Opteron 6376 1400 MHz processor, 512 GB RAM)