diff --git a/gdb.py b/gdb.py index a00d4216e5a7e1f37261c4a564136d9c9651734f..ae430f65fc2f209a3e141d140dc50d8efb699c97 100644 --- a/gdb.py +++ b/gdb.py @@ -678,7 +678,7 @@ gdb.execute("continue") # Your implementation should be generic though # Looking up the WCETs from the `output_data`. # (It may be a good idea to make first pass and extract wcet per task) -## +# # The total utilisation bound allows us to discard task sets that are # obviously illegal (not the case here though) # @@ -752,6 +752,7 @@ gdb.execute("continue") # Verify that that the results are correct by hand computation (or make an Excel) # # Assignment 4. +# # Looking closer at 7.22 we see that its a recurrent equation. # Ri(0) indicating the initial value # Ri(0) = Ci + Bi @@ -795,3 +796,21 @@ gdb.execute("continue") # Assignment 2, 10 points # Assignment 3, 10 points # Assignment 4, 15 points +# +# To make sure the analysis works in the general case +# you can make further examles based on 'resource.rs' +# +# Notice, KLEE analysis does not work on hardware peripherals +# (this is not yet supported), so your new examples must NOT access +# any peripherals. +# +# HINTS +# You may start by cut and paste the output (table) to a file 'x.py' +# +# Implement the analysis in a seprate python file 'x.py' +# (reconstruct the 'outputdata' from the table) +# +# When you have your analysis working, +# integrate it in this script (operating on the real 'outputdata') +# +#