Message5861

Author jendrik
Recipients andrew.coles, erez, florian, jendrik, malte, silvia
Date 2016-12-12.11:56:48
Content
Sure, here we go:

m32:

base-vs-v1: pointers and ints have the same size, so memory and runtime
don't change much. Coverage drops by 1 in total.

v1-vs-v3: memory usage is significantly reduced. Runtime results are
mixed. There's a slight negative trend for bjolp, cegar and ipdb and a
stronger negative trend for blind. Overall, all configurations benefit
from the change coverage stays the same or increases in all domains for
all configurations. Together, the configurations solve 61 more tasks.

m64:

base-vs-v1: both memory and total_time scores go up leading to 15
additional solved tasks.

v1-vs-v3: The picture is very similar to the 32-bit v1-vs-v3
comparison. Memory usage goes down, runtime goes up. The change in
memory is much more significant than the one for runtime though. In
total, coverage increases by 122 tasks.


base: m32-vs-m64 - Coverage drops for all configurations. Total coverage diff = -142
v1:   m32-vs-m64 - Coverage drops for all configurations. Total coverage diff = -136
v1:   m32-vs-m64 - Coverage drops by 20 and 32 for bjolp and seq, respectively. It
                   dropy by at most five for divpot, ipdb, lmcut and mas. It remains
                   the same for blind and cegar. Total coverage diff = -65


base-m32 vs. v3-m64: Memory usage increases significantly for most
configurations. Results for runtime are mixed. Coverage decreases by 12
and 33 for bjolp and seq, respectively. It increases for all other
configurations (+40 in total). Total coverage diff = -5.
History
Date User Action Args
2016-12-12 11:56:48jendriksetmessageid: <1481540208.12.0.460356024847.issue213@unibas.ch>
2016-12-12 11:56:48jendriksetrecipients: + jendrik, malte, erez, andrew.coles, silvia, florian
2016-12-12 11:56:48jendriklinkissue213 messages
2016-12-12 11:56:48jendrikcreate