Yeah, probably an uninitialized value or a bad pointer. This will be
annoying to figure out since parser is such a long running benchmark.
What I would suggest is that you find out where one of the offending
stats is incremented and add a DPRINTF, and then run it a few times
until you have a good run and a bad run. You can use diff on the two
undoubtedly large trace files to see what tick the extra or missing
increment happened or didn't happen. You'll want to create a new trace
flag (debug flag?) that will only trigger your new DPRINTF. You'll
want to use DPRINTF to get curTick in your output. You could also use
warn() or similar and add curTick manually.
Gabe
Quoting Andreas Hansson <[email protected]>:
Dear all,
I noticed during the weekend that the following regression
fails...again, despite change set 8880 that I pushed to adjust for
8868, and it seems that it is not failing consistently.
zizzer says:
...
system.cpu.l2cache.demand_hits::cpu.data 974876 974878
2 +0.00%
system.cpu.l2cache.overall_hits::cpu.data 974876 974878
2 +0.00%
system.cpu.l2cache.demand_hits::total 989052 989054
2 +0.00%
system.cpu.l2cache.overall_hits::total 989052 989054
2 +0.00%
system.cpu.dcache.writebacks::total 1073316 1073318
2 +0.00%
system.cpu.dcache.writebacks::writebacks 1073316 1073318
2 +0.00%
system.cpu.l2cache.Writeback_accesses::total 1073316 1073318
2 +0.00%
system.cpu.l2cache.Writeback_accesses::writebacks 1073316
1073318 2 +0.00%
system.cpu.l2cache.Writeback_hits::total 1073316 1073318
2 +0.00%
[... showing top 20 errors only, additional errors omitted ...]
***** build/ARM/tests/opt/long/se/20.parser/arm/linux/o3-timing FAILED!
On a different host I get "passed" for five runs in a row, and the
sixth one fails with similar changes.
Any clues? Uninitialized values perhaps? Any help is appreciated.
Andreas
-- IMPORTANT NOTICE: The contents of this email and any attachments
are confidential and may also be privileged. If you are not the
intended recipient, please notify the sender immediately and do not
disclose the contents to any other person, use it for any purpose,
or store or copy the information in any medium. Thank you.
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev
