http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56049
Bug #: 56049 Summary: [4.8 Regression] Simplification to constants not done Classification: Unclassified Product: gcc Version: 4.8.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: tree-optimization AssignedTo: unassig...@gcc.gnu.org ReportedBy: tkoe...@gcc.gnu.org >From http://gcc.gnu.org/ml/fortran/2013-01/msg00158.html : program inline integer i integer a(8,8), b(8,8) a = 0 do i = 1, 10000000 call add(b, a, 1) a = b end do print *, a contains subroutine add(b, a, o) integer, intent(inout) :: b(8,8) integer, intent(in) :: a(8,8), o b = a + o end subroutine add end program inline is simplified all the way to the final constant with 4.6 and 4.7 (example for 4.6.2): ;; Function inline (MAIN__) (executed once) inline () { struct array2_integer(kind=4) parm.3; struct __st_parameter_dt dt_parm.2; integer(kind=4) a[64]; <bb 2>: a = {}; MEM[(integer(kind=4)[64] *)&a] = 10000000; MEM[(integer(kind=4)[64] *)&a + 4B] = 10000000; MEM[(integer(kind=4)[64] *)&a + 8B] = 10000000; MEM[(integer(kind=4)[64] *)&a + 12B] = 10000000; MEM[(integer(kind=4)[64] *)&a + 16B] = 10000000; MEM[(integer(kind=4)[64] *)&a + 20B] = 10000000; MEM[(integer(kind=4)[64] *)&a + 24B] = 10000000; MEM[(integer(kind=4)[64] *)&a + 28B] = 10000000; MEM[(integer(kind=4)[64] *)&a + 32B] = 10000000; ... and so on. Current trunk converts this to ;; Function inline (MAIN__, funcdef_no=0, decl_uid=1874, cgraph_uid=1) (executed once) inline () { vector(4) integer(kind=4) vect_var_.16; vector(4) integer(kind=4) vect_var_.15; vector(4) integer(kind=4) vect_var_.14; struct array2_integer(kind=4) parm.3; struct __st_parameter_dt dt_parm.2; integer(kind=4) b[64]; integer(kind=4) a[64]; unsigned int ivtmp_153; unsigned int ivtmp_154; <bb 2>: a = {}; <bb 3>: # ivtmp_154 = PHI <10000000(2), ivtmp_153(4)> vect_var_.14_1 = MEM[(integer(kind=4)[64] *)&a]; vect_var_.15_42 = MEM[(integer(kind=4)[64] *)&a + 16B]; vect_var_.16_43 = vect_var_.14_1 + { 1, 1, 1, 1 }; vect_var_.16_44 = vect_var_.15_42 + { 1, 1, 1, 1 }; MEM[(integer(kind=4)[64] *)&b] = vect_var_.16_43; MEM[(integer(kind=4)[64] *)&b + 16B] = vect_var_.16_44; vect_var_.14_71 = MEM[(integer(kind=4)[64] *)&a + 32B]; vect_var_.15_77 = MEM[(integer(kind=4)[64] *)&a + 48B]; vect_var_.16_78 = vect_var_.14_71 + { 1, 1, 1, 1 }; vect_var_.16_79 = vect_var_.15_77 + { 1, 1, 1, 1 };