christoph/fail
1
Fork 0
Code Releases Activity

wmoo: experiments continue after trace window

Browse Source 
We don't know what data will be accessed later (trace is too short), so
do real experiments to be sure.  This could be done in a smarter way by
using a long enough trace.

git-svn-id: https://www4.informatik.uni-erlangen.de/i4svn/danceos/trunk/devel/fail@1153 8c4709b5-6ec9-48aa-a5cd-a96041d1645a
This commit is contained in:
hsc
2012-04-24 13:12:39 +00:00
parent c63155f072
commit 819cd757fe
1 changed files with 21 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
core/experiments/weathermonitor/campaign.cc
View File

@ -161,14 +161,32 @@ bool WeathermonitorCampaign::run()
// result comparable to the non-pruned campaign.
// XXX still true for weathermonitor?
current_ec.instr2 = instr - 1;
current_ec.instr2_absolute = 0; // won't be used
current_ec.instr2_absolute = 0; // unknown
current_ec.data_address = data_address;
// zero-sized? skip.
if (current_ec.instr1 > current_ec.instr2) {
continue;
}
// as the experiment ends, this byte is a "don't care":
ecs_no_effect.push_back(current_ec);
// the run continues after the FI window, so do this experiment
// XXX this creates at least one experiment for *every* bit!
// fix: full trace, limited FI window
//ecs_no_effect.push_back(current_ec);
ecs_need_experiment.push_back(current_ec);
// FIXME copy/paste, encapsulate this:
// instantly enqueue job: that way the job clients can already
// start working in parallel
WeathermonitorExperimentData *d = new WeathermonitorExperimentData;
// we pick the rightmost instruction in that interval
d->msg.set_instr_offset(current_ec.instr2);
//d->msg.set_instr_address(current_ec.instr2_absolute); // unknown!
d->msg.set_mem_addr(current_ec.data_address);
// store index into ecs_need_experiment
experiment_ecs[d] = ecs_need_experiment.size() - 1;
fi::campaignmanager.addParam(d);
++count;
}
fi::campaignmanager.noMoreParameters();