------- Additional Comments From pinskia at gcc dot gnu dot org 2005-09-18
23:30 -------
This is at least a 4.1 compile time regression.
3.3.2 gave:
[zhivago:~/src/localgccPRs] pinskia% time ~/gcc-3.3/bin/gcc pr21430.c -S -O1
10.870u 0.270s 0:13.11 84.9% 0+0k 2+4io 0pf+0w
While 4.1 gives (with --disable-checking):
[zhivago:~/src/localgccPRs] pinskia% time ~/local3/bin/gcc pr21430.c -S -O1
98.010u 0.650s 1:44.03 94.8% 0+0k 0+4io 0pf+0w
This is a bad regression, even though this is a testcase which just was made up
to show the issues.
Almost all of the time (90%) is spent in finalize_ssa_vuses.
tree operand scan : 90.11 (91%) usr 1.23 (37%) sys 101.08 (89%) wall
750 kB ( 4%) ggc
This is only a 4.1 regression as 4.0 gave:
12.590u 0.240s 0:13.80 92.9% 0+0k 0+4io 0pf+0w
--
What |Removed |Added
----------------------------------------------------------------------------
Component|middle-end |tree-optimization
Summary|Quadratic behavior with |[4.1 Regression] Quadratic
|constant initializers |behavior with constant
| |initializers
Target Milestone|--- |4.1.0
Version|unknown |4.1.0
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21430
