Hi, I think that NULL pointer checks should be added for all pointers must_raw, may_raw etc, not only for the *_no_source ones.
This will make the function more robust and easier to maintain. Indeed in the current code only the *_no_source pointers may be NULL, but this may change in the future so you don't want to base the correctness of the code on this assumption. Mircea ----- Original Message ----- > From: "Roman Gareev" <gareevro...@gmail.com> > To: gcc-patches@gcc.gnu.org > Cc: "Tobias Grosser" <tob...@grosser.es>, "mircea namolaru" > <mircea.namol...@inria.fr> > Sent: Monday, March 10, 2014 5:39:47 PM > Subject: Fix PR59586 > > This patch fixes PR59586. > The segfault is caused by NULL arguments passed to compute_deps by > loop_level_carries_dependences. > This causes an assignment of NULL values to the no_source parameters > of compute_deps. > They are passed to subtract_commutative_associative_deps and dereferenced. > > However, this NULL arguments are appropriate for the algorithm used > in loop_level_carries_dependences. It uses compute_deps > for finding RAW, WAR and WAW dependences of all basic blocks > in the body of the given loop. Subsequently, it tries to > determine presence of these dependences at the given level. > Therefore it maps the relation of the dependences to the relation > of the corresponding time-stamps and intersects the result with > the relation in which all the inputs before the DEPTH occur at the > same time as the output, and the input at the DEPTH occurs before output. > If the intersection is not empty, some dependences are carried > by the DEPTH we currently check and the loop is consequently not parallel. > > This patch tries to avoid the problem by adding NULL checking of the > no_source statements to > subtract_commutative_associative_deps. > > Tested x86_64-unknown-linux-gnu, applying to 4.8.3 and trunk. >
