------- Comment #2 from rguenth at gcc dot gnu dot org 2010-06-17 10:36 -------
The issue is not so much the number of functions but the number of calls
in main (and thus its size).
With GCC 4.3.4 I see
parser : 6.87 (12%) usr 0.79 (14%) sys 8.14 (13%) wall
266316 kB (18%) ggc
expand : 14.02 (24%) usr 1.38 (25%) sys 15.50 (24%) wall
619261 kB (41%) ggc
global alloc : 10.65 (18%) usr 0.27 ( 5%) sys 10.79 (17%) wall
81920 kB ( 5%) ggc
final : 6.16 (11%) usr 0.17 ( 3%) sys 6.21 (10%) wall
32704 kB ( 2%) ggc
TOTAL : 57.78 5.60 63.76
1509835 kB
peaking at 1.3GB virtual memory usage.
With GCC 4.5 I see
parser : 1.81 ( 3%) usr 0.66 ( 5%) sys 2.42 ( 3%) wall
196460 kB (13%) ggc
expand : 15.24 (25%) usr 3.45 (28%) sys 18.95 (26%) wall
506302 kB (34%) ggc
integrated RA : 11.75 (20%) usr 0.27 ( 2%) sys 12.09 (17%) wall
89600 kB ( 6%) ggc
reload : 6.07 (10%) usr 0.15 ( 1%) sys 5.98 ( 8%) wall
54527 kB ( 4%) ggc
thread pro- & epilogue: 2.97 ( 5%) usr 0.08 ( 1%) sys 3.22 ( 4%) wall
131586 kB ( 9%) ggc
final : 4.73 ( 8%) usr 0.15 ( 1%) sys 5.17 ( 7%) wall
36288 kB ( 2%) ggc
TOTAL : 59.83 12.45 72.64
1504261 kB
peaking at 1.7GB virtual memory usage.
Turning on basic optimization (-O1) inlines all your functions making
main() smaller, reducing peak memory usage to around 1GB but using
a lot of compile-time (I stopped it after 10min).
You basically have a gigantic basic-block in main() and GCC isn't
tuned to handle that case well.
Removing the main() function unfortunately doesn't make GCC behave much
saner here with regarding to memory usage. Removing all functions
but main reproduces the above figures with slightly reduced memory usage
and increased compile-time because the functions are not known to be const.
--
rguenth at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Ever Confirmed|0 |1
Keywords| |compile-time-hog, memory-hog
Last reconfirmed|0000-00-00 00:00:00 |2010-06-17 10:36:48
date| |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44563
