On 14/09/10 07:24 PM, Masoume Jabbarifar wrote:
There are two options to synchronise traces, accuracy and time that can be
chosen
by --sync-reduction in command line (something like --sync-reduction time)
This algorithm finds excess links (when we synchronise A, B and A, C, we can be
sure that we have synchronised B, C. But which link is excess? In this
algorithm,
the link that has less packet exchanges (less accuracy) has been removed so
synchronisation time has been improved.
In this patch, there are 55 trace files from 55 connected computers. This test
data can help better understanding of two synchronisation algorithms.
The sync code is organized in a chain of modules and this code bypasses
some of this modularization. I'm in the process of reviewing the patch
and suggesting changes that could make it cleaner and keep the whole
thing extensible. In the meantime, if you need a more pragmatic reason
not to merge it, it crashes lttv:
$ ./runlttv -m eval -H gdb
[...]
Program received signal SIGSEGV, Segmentation fault.
0x0806eb64 in initMatchingTCP (syncState=0x8087b68) at
sync/event_matching_tcp.c:118
118 if (syncState->reductionModule->preProcessReduction != NULL ||
syncState->stats) {
Signed-off-by: Masoume Jabbarifar<[email protected]>
---
lttv/lttv/Makefile.am | 2 +
lttv/lttv/sync/Makefile.am | 4 +-
lttv/lttv/sync/data_structures.c | 2 -
lttv/lttv/sync/event_matching_tcp.c | 87 +++-
lttv/lttv/sync/event_matching_tcp.h | 1 +
[mega snip]
Masoume, please read
http://lxr.linux.no/#linux+v2.6.35.4/Documentation/SubmittingPatches, in
particular section 8. That email was much too large. Besides, although
the test data is interesting, I doubt it should reside in the source
tree. You can instead provide a link to it in the "additional comments"
section of the patch (see section 15).
-Ben
_______________________________________________
ltt-dev mailing list
[email protected]
http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
