[Bug tree-optimization/14741] graphite with loop blocking and interchanging doesn't optimize a matrix multiplication loop

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14741

--- Comment #30 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Since at least gcc 4.8.2, -floop-interchange is able to apply it to the test in
comment 25 (see also pr60997):

[Book15] f90/bug% gfortran-fsf-4.8 -Ofast pr14741.f90 -floop-interchange
[Book15] f90/bug% time a.out
  0.48877399999999999        10.239999999999826     
0.492u 0.007s 0:00.50 98.0%    0+0k 0+0io 0pf+0w
[Book15] f90/bug% gfortran-fsf-4.8 -Ofast pr14741.f90 
[Book15] f90/bug% time a.out
   3.9007499999999999        10.239999999999826     
3.897u 0.015s 0:03.91 99.7%    0+0k 0+0io 1pf+0w

However compiling it with -floop-nest-optimize --param loop-block-tile-size=xx,
gives the following error (see comment 18)

f951: warning: iteration 31 invokes undefined behavior
[-Waggressive-loop-optimizations]
f951: note: containing loop
f951: warning: iteration 31 invokes undefined behavior
[-Waggressive-loop-optimizations]
f951: note: containing loop
pr14741.f90:6:0: warning: iteration 31 invokes undefined behavior
[-Waggressive-loop-optimizations]
 B=0.1D0
 ^
f951: note: containing loop
pr14741.f90:5:0: warning: iteration 31 invokes undefined behavior
[-Waggressive-loop-optimizations]
 A=0.1D0
 ^
f951: note: containing loop

for any values of xx I have tested. If I add
-fno-aggressive-loop-optimizations, the code segfaults (Access to an undefined
portion of a memory object) at run time.