https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46590
Thomas Koenig <tkoenig at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Priority|P2 |P3
--- Comment #56 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
I thought I'd dust off this PR and run the original test case again.
This is on cfarm120, with a compiler with checking, with the original
test case. This may have regressed:
Timings with gfortran -ftime-report -O0 gener.f90
Time variable wall GGC
phase setup : 0.01 ( 0%) 187k ( 0%)
phase parsing : 1.89 ( 7%) 72M ( 10%)
phase lang. deferred : 0.01 ( 0%) 0 ( 0%)
phase opt and generate : 24.72 ( 93%) 647M ( 90%)
[...]
verify RTL sharing : 2.11 ( 8%) 0 ( 0%)
TOTAL : 26.64 720M
gfortran -ftime-report -O1 gener.f90 :
Time variable wall GGC
phase setup : 0.01 ( 0%) 187k ( 0%)
phase parsing : 1.94 ( 1%) 72M ( 7%)
phase lang. deferred : 0.01 ( 0%) 0 ( 0%)
phase opt and generate : 308.84 ( 99%) 993M ( 93%)
garbage collection : 1.00 ( 0%) 0 ( 0%)
callgraph construction : 0.27 ( 0%) 61M ( 6%)
callgraph optimization : 0.28 ( 0%) 0 ( 0%)
callgraph functions expansion : 276.56 ( 89%) 713M ( 67%)
callgraph ipa passes : 30.83 ( 10%) 114M ( 11%)
[...]
TOTAL : 310.81 1066M
Extra diagnostic checks enabled; compiler may run slowly.
Configure with --enable-checking=release to disable checks.
gfortran -ftime-report -O2 -o gener gener.f90
Time variable wall GGC
phase setup : 0.01 ( 0%) 187k ( 0%)
phase parsing : 2.03 ( 0%) 72M ( 4%)
phase lang. deferred : 0.01 ( 0%) 0 ( 0%)
phase opt and generate :2462.68 (100%) 1568M ( 96%)
garbage collection : 1.55 ( 0%) 0 ( 0%)
callgraph construction : 0.31 ( 0%) 61M ( 4%)
callgraph optimization : 0.27 ( 0%) 0 ( 0%)
callgraph functions expansion :2418.29 ( 98%) 1267M ( 77%)
[...]
repair loop structures : 0.18 ( 0%) 0 ( 0%)
TOTAL :2464.74 1641M
