https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114989
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Version|14.0 |13.2.1
--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
baseml.i: 11.4: 2.04s, 12.3: 2.28s, 13.2: 2.42s 14.1: 2.69s
so at least there's a trend (more passes, more IL walks). A quick profile
shows
Samples: 9K of event 'cycles:u', Event count (approx.): 12946131497
Overhead Samples Command Shared Object Symbol
2.41% 238 cc1 cc1 [.] bitmap_set_bit
1.43% 141 cc1 cc1 [.] et_splay
1.13% 112 cc1 libc-2.31.so [.] _int_malloc
0.94% 93 cc1 cc1 [.] constrain_operands
with 14.1 it's instead
Samples: 11K of event 'cycles:u', Event count (approx.): 14418085320
Overhead Samples Command Shared Object Symbol
2.19% 242 cc1 cc1 [.]
wide_int_storage::opera
2.14% 236 cc1 cc1 [.] bitmap_set_bit
1.33% 147 cc1 cc1 [.]
irange_storage::get_ira
1.06% 117 cc1 cc1 [.] bitmap_bit_p
1.04% 115 cc1 cc1 [.] et_splay
0.99% 109 cc1 cc1 [.] bitmap_clear_bit
0.97% 107 cc1 cc1 [.] irange::set_varying
0.93% 103 cc1 libc-2.31.so [.] _int_malloc
so it might be all scattered ranger stuff.
