christoph/fail
1
Fork 0
Code Releases Activity

ecos: fault-space cutoff specific for this campaign

Browse Source 
For the eCos kernel test campaign we define a relatively "special" metric
to compare FI results from different applications:  Instead of aligning
the fault-space dimensions (e.g., by artificially adding "don't care" space
to the right for shorter application variants), we only keep the data
address axis constant.  The rationale behind this is that -- despite the
benchmark applications' run-to-completion behavior -- an operating system's
scheduler usually runs ad infinitum, and that we therefore can extrapolate
from a short application run to any longer period.

In essence, this means we use a failure/experiment-length quotient (ignoring
the Y axis for simplicity, as it is constant in size) to compare
application variants, instead of an absolute failure count.  One important
side effect is that we do not punish any application slowdown with this
metric.  (And simply prolongating non-susceptible program sections with,
e.g., NOPs, seemingly "improves" fault tolerance.)

git-svn-id: https://www4.informatik.uni-erlangen.de/i4svn/danceos/trunk/devel/fail@1922 8c4709b5-6ec9-48aa-a5cd-a96041d1645a
This commit is contained in:
hsc
2012-11-12 13:55:45 +00:00
parent 6f98d64613
commit a0e6ddd519
1 changed files with 9 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
src/experiments/ecos_kernel_test/campaign.cc
View File

@ -201,6 +201,8 @@ bool EcosKernelTestCampaign::run()
// instruction counter within trace
unsigned instr = 0;
// "rightmost" instr where we did a FI experiment
unsigned instr_rightmost = 0;
// fill open_ecs with one EC for every address
for (MemoryMap::iterator it = mm.begin(); it != mm.end(); ++it) {
@ -247,6 +249,7 @@ bool EcosKernelTestCampaign::run()
// a sequence ending with READ: we need to do one experiment to
// cover it completely
add_experiment_ec(variant, benchmark, data_address, instr1, instr2, instr_absolute);
instr_rightmost = instr2;
} else if (ev.accesstype() == ev.WRITE) {
// a sequence ending with WRITE: an injection anywhere here
// would have no effect.
@ -267,12 +270,17 @@ bool EcosKernelTestCampaign::run()
address_t data_address = lastuse_it->first;
int instr1 = lastuse_it->second;
#if 0
// Why -1? In most cases it does not make sense to inject before the
// very last instruction, as we won't execute it anymore. This *only*
// makes sense if we also inject into parts of the result vector. This
// is not the case in this experiment, and with -1 we'll get a result
// comparable to the non-pruned campaign.
int instr2 = instr - 1;
#else
// EcosKernelTestCampaign only variant: fault space ends with the last FI experiment
int instr2 = instr_rightmost;
#endif
int instr_absolute = 0; // unknown
// zero-sized? skip.
@ -299,6 +307,7 @@ bool EcosKernelTestCampaign::run()
m_log << variant << "/" << benchmark
<< " exp " << (count_exp - local_count_exp) << " (" << (count_exp_jobs - local_count_exp_jobs) << " jobs)"
<< " known " << (count_known - local_count_known) << " (" << (count_known_jobs - local_count_known_jobs) << " jobs)"
<< " faultspace cutoff @ " << instr_rightmost << " out of " << instr
<< endl;
}
}