https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111624
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |dmalcolm at gcc dot gnu.org
--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
With -g0 it's (trunk)
191.36user 2.29system 3:13.74elapsed 99%CPU (0avgtext+0avgdata
5827388maxresident)k
776inputs+143592outputs (1major+1815547minor)pagefaults 0swaps
pre-IPA we use 256MB for ad-hoc location stuff with 5.4 million entries.
Also:
Total Overhead page size 67108864: 8
Total Allocated page size 67108864: 128M
Total Overhead page size 134217728: 8
Total Allocated page size 134217728: 256M
Total Overhead page size 268435456: 8
Total Allocated page size 268435456: 512M
decls 880k 118M
refs 1699k 60M
random kinds 1837k 71M
exprs 7271k 278M
----------------------------------------
Total 13M 617M
convert_expr 463k
call_expr 974k
non_lvalue_expr 1009k
view_convert_expr 1150k
addr_expr 1368k
nop_expr 1402k
tree_list 1677k
GIMPLE is just 120M
ash-table.h:858 (expand) 25M: 2.9% 66M:
9.9% 129M: 12.7% 38k: 0.1% 700
dwarf2out.cc:5827 (new_die_raw) 26M: 3.0% 142k:
0.0% 0 : 0.0% 0 : 0.0% 336k
cgraph.cc:891 (create_edge) 26M: 3.0% 0 :
0.0% 0 : 0.0% 0 : 0.0% 265k
gimple.cc:231 (gimple_build_call_1) 41M: 4.7% 10M:
1.6% 0 : 0.0% 1622k: 4.1% 341k
tree-ssanames.cc:148 (init_ssanames) 52M: 5.9% 0 :
0.0% 61k: 0.0% 17M: 44.9% 34k
dwarf2out.cc:4506 (add_dwarf_attr) 121M: 13.7% 300k:
0.0% 44M: 4.3% 8603k: 21.7% 588k
toplev.cc:755 (realloc_for_line_map) 258M: 29.2% 0 :
0.0% 256M: 25.2% 40 : 0.0% 22
--------------------------------------------------------------------------------------------------------------------------------------------
GGC memory Leak Garbage
Freed Overhead Times
--------------------------------------------------------------------------------------------------------------------------------------------
Total 884M:100.0%
668M:100.0% 1017M:100.0% 38M:100.0% 20M
final memory use is
Total Allocated page size 1073741824: 1024M
Total Overhead page size 2147483648: 0
Total Allocated page size 2147483648: 2048M
and 2GB of it is the ad-hoc table:
Ad-hoc table size: 2048M
Ad-hoc table entries used: 43M
optimized_ranges: 3206k
unoptimized_ranges: 5767k
I think we never get rid of any of that but we possibly create a lot of
garbage due to inlining? Or the new suppressed option stuff?
We now have
GIMPLE statements
Kind Stmts Bytes
---------------------------------------
assignments 14M 1260M
phi nodes 62k 725M
conditionals 11M 921M
everything else 15M 949M
---------------------------------------
Total 42M 3856M
so that's 4GB worth of GIMPLE!
rtvec 510k 8546k
-----------------------------------------------
Total 51M 1578M
but only 1.5GB worth of RTL?!
We really seem to go wild inlining here.
