https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83017
--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
If I "fix" GCC to consider the loop you annotate parallel:
do concurrent (i = 1:nsplit)
pi(i) = sum(compute( low(i), high(i) ))
end do
then we arrive at computing 4 iterations of that loop and with 2 threads
and MIN_PER_THREAD 100 (arbitrary define) we run into
if (!flag_loop_parallelize_all
&& !oacc_kernels_p
&& ((estimated != -1
&& estimated <= (HOST_WIDE_INT) n_threads * MIN_PER_THREAD)
/* Do not bother with loops in cold areas. */
|| optimize_loop_nest_for_size_p (loop)))
continue;
(estimated is 4). With -floop-parallelize-all I then get:
> ./f951 -quiet t.f90 -Ofast -ftree-parallelize-loops=2
> -fdump-tree-parloops-details -floop-parallelize-all -fopt-info-loop
t.f90:28:0: note: loop with 5 iterations completely unrolled (header execution
count 375)
t.f90:26:0: note: loop with 5 iterations completely unrolled (header execution
count 1500)
t.f90:38:0: note: loop with 5 iterations completely unrolled (header execution
count 1500)
t.f90:18:0: note: loop with 4 iterations completely unrolled (header execution
count 375)
t.f90:15:0: note: loop with 5 iterations completely unrolled (header execution
count 375)
t.f90:26:0: note: parallelizing outer loop 3
t.f90:24:0: note: basic block vectorized
t.f90:41:0: note: basic block vectorized
t.f90:41:0: note: basic block vectorized
yay.
