If you're only getting errors on cache statistics, then you're in
business... this is a bug with an uninitialized variable that we didn't
catch until recently. Note that both the reference value and the value
from the test run are completely bogus.
We'll be posting a 2.0b2 soon with this fix and a few other small ones.
Thanks for helping us work through the Python link issues.
Steve
Michael Van Biesbrouck wrote:
Setting LD_LIBRARY_PATH makes things work. I found one page online
suggesting that this linking problem for libpython.2.4.1.0.so only
occurs on 64-bit systems (it would be a python problem). I now have
an M5 without python shared library problems that builds with the
default scons file.
Unfortunately, it fails most of the regression tests. Here is a sample:
Largest 20 relative errors (> 0%):
system.cpu.dcache.ReadReq_avg_miss_latency 2918912699678311424
-241329268597313248 -3160241968275624448 -108.27%
system.cpu.dcache.ReadReq_miss_latency 160540198482307121152
-13273109772852228096 -173813308255159353344 -108.27%
system.cpu.dcache.WriteReq_avg_miss_latency 4476343852030456320
-298009189798403136 -4774353041828859904 -106.66%
Errors of around -100% are typical.
Michael
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
