A recent patch increased GCC's memory consumption!

[gcctest@suse.de]

Hi,

I am a friendly script caring about memory consumption in GCC.  Please
contact jh@suse.cz if something is going wrong.

Comparing memory consumption on compilation of combine.i, insn-attrtab.i,
and generate-3.4.ii I got:


comparing combine.c compilation at -O0 level:
    Overall memory needed: 25320k -> 25324k
    Peak memory use before GGC: 9566k
    Peak memory use after GGC: 8913k
    Maximum of released memory in single GGC run: 2649k
    Garbage: 40060k
    Leak: 6740k
    Overhead: 5738k
    GGC runs: 313

comparing combine.c compilation at -O1 level:
    Overall memory needed: 26892k
    Peak memory use before GGC: 17434k
    Peak memory use after GGC: 17255k
    Maximum of released memory in single GGC run: 2309k
    Garbage: 62589k
    Leak: 6889k
    Overhead: 7488k
    GGC runs: 393

comparing combine.c compilation at -O2 level:
  Amount of produced GGC garbage increased from 80093k to 83409k, overall 4.14%
  Amount of memory still referenced at the end of compilation increased from 6962k to 6977k, overall 0.21%
    Overall memory needed: 26892k
    Peak memory use before GGC: 17436k
    Peak memory use after GGC: 17255k
    Maximum of released memory in single GGC run: 2388k -> 2410k
    Garbage: 80093k -> 83409k
    Leak: 6962k -> 6977k
    Overhead: 10032k -> 10316k
    GGC runs: 450 -> 451

comparing combine.c compilation at -O3 level:
  Peak amount of GGC memory allocated before garbage collecting increased from 18273k to 18414k, overall 0.77%
  Peak amount of GGC memory still allocated after garbage collectin increased from 17952k to 17970k, overall 0.10%
  Amount of produced GGC garbage increased from 111389k to 115499k, overall 3.69%
    Overall memory needed: 26892k
    Peak memory use before GGC: 18273k -> 18414k
    Peak memory use after GGC: 17952k -> 17970k
    Maximum of released memory in single GGC run: 3378k -> 3385k
    Garbage: 111389k -> 115499k
    Leak: 7038k -> 7040k
    Overhead: 13723k -> 14088k
    GGC runs: 502

comparing insn-attrtab.c compilation at -O0 level:
    Overall memory needed: 80948k
    Peak memory use before GGC: 69508k
    Peak memory use after GGC: 45044k
    Maximum of released memory in single GGC run: 36220k
    Garbage: 146427k
    Leak: 10131k
    Overhead: 19750k
    GGC runs: 247

comparing insn-attrtab.c compilation at -O1 level:
    Overall memory needed: 109164k -> 109560k
    Peak memory use before GGC: 91973k
    Peak memory use after GGC: 81071k
    Maximum of released memory in single GGC run: 32143k
    Garbage: 296462k
    Leak: 10054k
    Overhead: 36102k
    GGC runs: 245

comparing insn-attrtab.c compilation at -O2 level:
    Overall memory needed: 122560k
    Peak memory use before GGC: 109356k
    Peak memory use after GGC: 80974k
    Maximum of released memory in single GGC run: 31578k
    Garbage: 381018k -> 381029k
    Leak: 10156k
    Overhead: 47871k -> 47872k
    GGC runs: 274

comparing insn-attrtab.c compilation at -O3 level:
    Overall memory needed: 122616k
    Peak memory use before GGC: 109382k
    Peak memory use after GGC: 80999k
    Maximum of released memory in single GGC run: 31908k
    Garbage: 381619k -> 381630k
    Leak: 10160k
    Overhead: 48068k -> 48069k
    GGC runs: 276

comparing Gerald's testcase PR8361 compilation at -O0 level:
    Overall memory needed: 118424k
    Peak memory use before GGC: 95198k
    Peak memory use after GGC: 94250k
    Maximum of released memory in single GGC run: 20356k
    Garbage: 223714k
    Leak: 49472k
    Overhead: 37113k
    GGC runs: 369

comparing Gerald's testcase PR8361 compilation at -O1 level:
    Overall memory needed: 108528k
    Peak memory use before GGC: 95331k
    Peak memory use after GGC: 93251k
    Maximum of released memory in single GGC run: 20233k
    Garbage: 565076k
    Leak: 52291k
    Overhead: 63655k
    GGC runs: 531

comparing Gerald's testcase PR8361 compilation at -O2 level:
  Amount of produced GGC garbage increased from 655144k to 690223k, overall 5.35%
  Amount of memory still referenced at the end of compilation increased from 53114k to 53426k, overall 0.59%
    Overall memory needed: 108532k
    Peak memory use before GGC: 95331k
    Peak memory use after GGC: 93252k
    Maximum of released memory in single GGC run: 20232k
    Garbage: 655144k -> 690223k
    Leak: 53114k -> 53426k
    Overhead: 74999k -> 77124k
    GGC runs: 594 -> 617

comparing Gerald's testcase PR8361 compilation at -O3 level:
  Amount of produced GGC garbage increased from 715025k to 753221k, overall 5.34%
  Amount of memory still referenced at the end of compilation increased from 53992k to 54359k, overall 0.68%
    Overall memory needed: 114216k -> 114280k
    Peak memory use before GGC: 96797k
    Peak memory use after GGC: 94327k
    Maximum of released memory in single GGC run: 20729k
    Garbage: 715025k -> 753221k
    Leak: 53992k -> 54359k
    Overhead: 79548k -> 81858k
    GGC runs: 606 -> 626

Head of the ChangeLog is:

--- /usr/src/SpecTests/sandbox-britten-memory/x86_64/mem-result/ChangeLog	2006-02-06 05:01:16.000000000 +0000
+++ /usr/src/SpecTests/sandbox-britten-memory/gcc/gcc/ChangeLog	2006-02-06 15:23:10.000000000 +0000
@@ -1,4 +1,34 @@
-2006-02-04  Ben Elliston  <bje@au.ibm.com>
+2006-02-06  Daniel Berlin  <dberlin@dberlin.org>
+
+	* tree-ssa-pre.c (bb_value_sets_t): Add antic_safe_loads.
+	(ANTIC_SAFE_LOADS): New macro.
+	(find_or_generate_expression): Add prototype.
+	(set_contains_value): Allow null set for sake of not always having
+	to allocate ANTIC_SAFE_LOADS.
+	(phi_translate): Move placement of AGGREGATE_TYPE_P check.
+	Allow COMPONENT_REF too.
+	(valid_in_set): Allow COMPONENT_REF.
+	Check ANTIC_SAFE_LOADS too.
+	(compute_antic_aux): Print out ANTIC_SAFE_LOADS.
+	(compute_rvuse_and_antic_safe): Add ANTIC_SAFE computation, and
+	rename.
+	(can_PRE_operation): Add COMPONENT_REF.
+	(create_component_ref_by_pieces): New function.
+	(create_expression_by_pieces): Use create_component_ref_by_pieces.
+	(insert_aux): Move AGGREGATE_TYPE_P check here.
+	(compute_avail): Set bb local stmt uids.
+	(pass_pre): Use TODO_update_ssa_only_virtuals.
+	
+2006-02-06  Ben Elliston  <bje@au.ibm.com>
+
+	* configure.ac: Unconditionally disable decimal float by default.
+	* configure: Regenerate.
+
+2006-02-06  Ben Elliston  <bje@au.ibm.com>
+
+	* read-rtl.c (read_name): Terminate reading on EOF.
+
+2006-02-06  Ben Elliston  <bje@au.ibm.com>
 
 	* dfp.c (decimal_real_arithmetic): Change second argument type
 	from int to enum tree_code. Do not always return true, but now


The results can be reproduced by building a compiler with

--enable-gather-detailed-mem-stats targetting x86-64

and compiling preprocessed combine.c or testcase from PR8632 with:

-fmem-report --param=ggc-min-heapsize=1024 --param=ggc-min-expand=1 -Ox -Q

The memory consumption summary appears in the dump after detailed listing
of the places they are allocated in.  Peak memory consumption is actually
computed by looking for maximal value in {GC XXXX -> YYYY} report.

Your testing script.