It seems like some optimization passes of CLang which are enabled by
-O2 (at least in my version, 3.3) do not like nested scopes and long
functions too much. Thus, slightly change the generated source. Timing
on my (quite beefy) machine:
without patch:
make
rm -rf generated-source; time make
time make
[...]
real 10m31.110s
user 10m16.264s
sys 0m13.672s
with patch:
make
rm -rf generated-source; time make
time make
[...]
real 0m47.011s
user 0m44.670s
sys 0m1.968s
the memory used by the compiler goes from 28.8GB to about 330 MB. (I
suppose not everybody has 32 Gigs of memory yet. ;)
these includes are required by the headers. If the affected files
would have been included without the headers included before, a
compiler error would have been produced.
for this, an Section::isDeckValid() method is introduced which checks
that a given deck is valid when it comes to the sections
(i.e. presence of mandatory sections and section ordering)
We now do not require the sections to be correctly ordered and the
presence of the mandatory sections since even the unit tests did not
always specify all mandatory sections, which lead to a section
containing the rest of the deck if one of the expected next sections
was not specified.
also it seems like the DeckKeyword::getDeckIndex() does not correctly
work in some situations which lead
assert(deck->getKeyword(i)->name() == startKeyword);
to fail in the Section::populateKeywords() method. now the deck is
always sequentially traversed to find the position of a section's
start keyword. (This is necessary anyway if one wants to make sure
that the deck does not specify the same section more than once, a
feature which this patch also adds.)
The new constructor allows us to create EclipseGrid objects from decks
that do not have a strict sectioned structure.
Also modify some implementation details. Many methods are now templates,
and can take Deck or Section objects (they have the same
hasKeyword/getKeyword interface).
this comes with a minor API change as well, as the FaultCollection
class did not use the Cartesian size of the grid at all, so I decided
to remove the attributes and the arguments to the constructor...
it returns a const reference to a vector which does not allow
manipulation of its members. the getData() can thus be safely called
on constant GridProperty objects.
i.e. MULT[XYZ]- was trimmed to MULT[XYZ]. Also, the RawKeyword now
uses ParserKeyword::isValidDeckName() instead of a regular expression
which makes it automatically consistent and also should make it
slightly faster...
