http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46590
--- Comment #41 from Richard Biener <rguenth at gcc dot gnu.org> --- After I will have committed the patch the original testcase will use about 1GB of memory (regardless of optimization level) on x86_64, 122s at -O1 and 161s at -Ofast with the following main contributors: > ./f951 -quiet -Ofast t2.f90 -ftime-report Execution times (seconds) phase opt and generate : 159.14 (98%) usr 116.92 (100%) sys 276.41 (99%) wall 821437 kB (92%) ggc df reaching defs : 36.37 (22%) usr 91.52 (78%) sys 128.33 (46%) wall 0 kB ( 0%) ggc df live regs : 2.60 ( 2%) usr 0.19 ( 0%) sys 2.84 ( 1%) wall 0 kB ( 0%) ggc alias stmt walking : 36.03 (22%) usr 0.46 ( 0%) sys 36.53 (13%) wall 1451 kB ( 0%) ggc parser (global) : 2.65 ( 2%) usr 0.08 ( 0%) sys 2.74 ( 1%) wall 70763 kB ( 8%) ggc tree CFG cleanup : 8.23 ( 5%) usr 0.03 ( 0%) sys 8.19 ( 3%) wall 4982 kB ( 1%) ggc tree PTA : 22.95 (14%) usr 0.17 ( 0%) sys 23.12 ( 8%) wall 7882 kB ( 1%) ggc complete unrolling : 4.88 ( 3%) usr 0.14 ( 0%) sys 5.08 ( 2%) wall 80683 kB ( 9%) ggc loop init : 5.93 ( 4%) usr 0.01 ( 0%) sys 5.92 ( 2%) wall 25175 kB ( 3%) ggc integrated RA : 2.97 ( 2%) usr 0.03 ( 0%) sys 3.02 ( 1%) wall 64778 kB ( 7%) ggc TOTAL : 161.83 117.01 279.22 892406 kB (everything > 1% usr listed)