Hi
Are there any issues with running GEMS with a Linux/Sparc target
machine? I am simulating a 16-processor sparc machine in simics (using
their bagle script) and was trying to run the transactional
microbenchmark (tm-dequeue) that comes with GEMS. I loaded ruby module
but not opal since I'm only concerned with the memory access patterns. I
used the "MESI_SMP_LogTM_directory" cache protocol. I am giving a
screenshot of the error I got:
simics> run-command-file file = setup_ruby.sh
Turning D-STC off and flushing old data
successful installation of the ruby timing model.
Ruby Timing Mode
Warning: optimizations not enabled.
Creating event queue...
Creating event queue done
Creating system...
Processors: 16
Creating system done
Ruby initialization complete
simics> c
<****The following of course came after starting the tm-deque program****>
[cpu17] v:0x0000000000010bd8 p:0x0000000004f2ebd8 magic (sethi 0x40000,
%g0)
Setting new inspection cpu: cpu17
simics> c
Exception in hap handler: Non-existent Register
Exception in hap handler
[cpu17] v:0x000000000001098c p:0x0000000004f2e98c magic (sethi
0x140000, %g0)
simics>
The first signs of problems is ofcourse cpu17. No idea where that came
from with 16 processors! Before I investigate the problem any further, I
wanted to figure out if there are any known issues with running
transactional benchmarks using magic instructions with Linux/Sparc
setup. I know it has been made to work with Solaris/Sparc setup. Does it
not run on Linux/Sparc machines? Has anyone tried using magic
instructions with Linux/Sparc setup? Any feedback will be really useful.
Thanks
|