This is the mail archive of the gcc-regression@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

A recent patch increased GCC's memory consumption!


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:
  Amount of memory still referenced at the end of compilation increased from 6711k to 6726k, overall 0.22%
    Overall memory needed: 24816k -> 24812k
    Peak memory use before GGC: 9651k
    Peak memory use after GGC: 8963k -> 8964k
    Maximum of released memory in single GGC run: 2791k
    Garbage: 42422k -> 42454k
    Leak: 6711k -> 6726k
    Overhead: 5914k -> 5935k
    GGC runs: 330

comparing combine.c compilation at -O1 level:
  Amount of produced GGC garbage increased from 63452k to 63520k, overall 0.11%
    Overall memory needed: 27644k -> 27636k
    Peak memory use before GGC: 9177k -> 9178k
    Peak memory use after GGC: 8747k -> 8748k
    Maximum of released memory in single GGC run: 2204k
    Garbage: 63452k -> 63520k
    Leak: 7092k -> 7091k
    Overhead: 7690k -> 7711k
    GGC runs: 525

comparing combine.c compilation at -O2 level:
  Peak amount of GGC memory allocated before garbage collecting increased from 18366k to 18412k, overall 0.25%
  Peak amount of GGC memory still allocated after garbage collectin increased from 18179k to 18224k, overall 0.25%
    Overall memory needed: 24852k -> 24936k
    Peak memory use before GGC: 18366k -> 18412k
    Peak memory use after GGC: 18179k -> 18224k
    Maximum of released memory in single GGC run: 2522k
    Garbage: 88771k -> 88852k
    Leak: 7042k -> 7040k
    Overhead: 11024k -> 11049k
    GGC runs: 484

comparing combine.c compilation at -O3 level:
  Peak amount of GGC memory allocated before garbage collecting increased from 18563k to 18623k, overall 0.32%
  Peak amount of GGC memory still allocated after garbage collectin increased from 18179k to 18224k, overall 0.25%
    Overall memory needed: 25640k -> 25704k
    Peak memory use before GGC: 18563k -> 18623k
    Peak memory use after GGC: 18179k -> 18224k
    Maximum of released memory in single GGC run: 3389k -> 3404k
    Garbage: 123557k -> 123668k
    Leak: 7152k -> 7135k
    Overhead: 15239k -> 15277k
    GGC runs: 545 -> 544

comparing insn-attrtab.c compilation at -O0 level:
  Amount of produced GGC garbage increased from 154203k to 154701k, overall 0.32%
    Overall memory needed: 85596k
    Peak memory use before GGC: 73852k -> 73853k
    Peak memory use after GGC: 45356k -> 45357k
    Maximum of released memory in single GGC run: 37612k
    Garbage: 154203k -> 154701k
    Leak: 11538k -> 11538k
    Overhead: 19905k -> 20071k
    GGC runs: 268

comparing insn-attrtab.c compilation at -O1 level:
  Amount of produced GGC garbage increased from 306690k to 307192k, overall 0.16%
    Overall memory needed: 101264k -> 101424k
    Peak memory use before GGC: 77258k -> 77259k
    Peak memory use after GGC: 66356k -> 66358k
    Maximum of released memory in single GGC run: 37101k -> 37102k
    Garbage: 306690k -> 307192k
    Leak: 11599k -> 11600k
    Overhead: 37513k -> 37679k
    GGC runs: 385

comparing insn-attrtab.c compilation at -O2 level:
  Peak amount of GGC memory allocated before garbage collecting increased from 120391k to 120533k, overall 0.12%
  Peak amount of GGC memory still allocated after garbage collectin increased from 94016k to 94158k, overall 0.15%
  Amount of produced GGC garbage increased from 410312k to 410818k, overall 0.12%
    Overall memory needed: 155080k -> 153688k
    Peak memory use before GGC: 120391k -> 120533k
    Peak memory use after GGC: 94016k -> 94158k
    Maximum of released memory in single GGC run: 32813k -> 32951k
    Garbage: 410312k -> 410818k
    Leak: 11438k -> 11438k
    Overhead: 50912k -> 51078k
    GGC runs: 307

comparing insn-attrtab.c compilation at -O3 level:
  Peak amount of GGC memory allocated before garbage collecting increased from 120393k to 120535k, overall 0.12%
  Peak amount of GGC memory still allocated after garbage collectin increased from 94018k to 94160k, overall 0.15%
  Amount of produced GGC garbage increased from 411105k to 411609k, overall 0.12%
    Overall memory needed: 155092k -> 153648k
    Peak memory use before GGC: 120393k -> 120535k
    Peak memory use after GGC: 94018k -> 94160k
    Maximum of released memory in single GGC run: 32813k -> 32951k
    Garbage: 411105k -> 411609k
    Leak: 11457k -> 11456k
    Overhead: 51036k -> 51202k
    GGC runs: 312

comparing Gerald's testcase PR8361 compilation at -O0 level:
  Peak amount of GGC memory allocated before garbage collecting increased from 102907k to 103134k, overall 0.22%
  Peak amount of GGC memory still allocated after garbage collectin increased from 101886k to 102097k, overall 0.21%
  Amount of produced GGC garbage increased from 247847k to 248307k, overall 0.19%
    Overall memory needed: 126912k -> 127372k
    Peak memory use before GGC: 102907k -> 103134k
    Peak memory use after GGC: 101886k -> 102097k
    Maximum of released memory in single GGC run: 21112k -> 20902k
    Garbage: 247847k -> 248307k
    Leak: 53828k -> 53806k
    Overhead: 43099k -> 43210k
    GGC runs: 347

comparing Gerald's testcase PR8361 compilation at -O1 level:
  Overall memory allocated via mmap and sbrk increased from 120296k to 122780k, overall 2.06%
  Peak amount of GGC memory allocated before garbage collecting increased from 111955k to 113480k, overall 1.36%
  Peak amount of GGC memory still allocated after garbage collectin increased from 100814k to 100997k, overall 0.18%
  Amount of produced GGC garbage increased from 678695k to 682542k, overall 0.57%
    Overall memory needed: 120296k -> 122780k
    Peak memory use before GGC: 111955k -> 113480k
    Peak memory use after GGC: 100814k -> 100997k
    Maximum of released memory in single GGC run: 20025k -> 20054k
    Garbage: 678695k -> 682542k
    Leak: 57767k -> 57764k
    Overhead: 86872k -> 87919k
    GGC runs: 521 -> 519

comparing Gerald's testcase PR8361 compilation at -O2 level:
  Overall memory allocated via mmap and sbrk increased from 120304k to 122780k, overall 2.06%
  Peak amount of GGC memory allocated before garbage collecting increased from 111955k to 113480k, overall 1.36%
  Peak amount of GGC memory still allocated after garbage collectin increased from 100814k to 100997k, overall 0.18%
  Amount of produced GGC garbage increased from 791193k to 795041k, overall 0.49%
    Overall memory needed: 120304k -> 122780k
    Peak memory use before GGC: 111955k -> 113480k
    Peak memory use after GGC: 100814k -> 100997k
    Maximum of released memory in single GGC run: 20025k -> 20054k
    Garbage: 791193k -> 795041k
    Leak: 58792k -> 58790k
    Overhead: 106895k -> 107945k
    GGC runs: 619 -> 617

comparing Gerald's testcase PR8361 compilation at -O3 level:
  Peak amount of GGC memory allocated before garbage collecting increased from 113953k to 115563k, overall 1.41%
  Amount of produced GGC garbage increased from 813836k to 817758k, overall 0.48%
    Overall memory needed: 123184k -> 124496k
    Peak memory use before GGC: 113953k -> 115563k
    Peak memory use after GGC: 102646k -> 102656k
    Maximum of released memory in single GGC run: 21329k -> 21391k
    Garbage: 813836k -> 817758k
    Leak: 59486k -> 59468k
    Overhead: 110022k -> 111107k
    GGC runs: 624 -> 617

Head of the ChangeLog is:

--- /usr/src/SpecTests/sandbox-britten-memory/x86_64/mem-result/ChangeLog	2005-05-19 06:52:30.000000000 +0000
+++ /usr/src/SpecTests/sandbox-britten-memory/gcc/gcc/ChangeLog	2005-05-19 19:39:42.000000000 +0000
@@ -1,3 +1,61 @@
+2005-05-19  Richard Guenther  <rguenth@gcc.gnu.org>
+
+	* tree-ssa-loop-ivopts.c (determine_base_object): Use
+	build_fold_addr_expr to generate trees of correct type.
+	Use fold_buildN if possible.
+
+2005-05-19  Jan Hubicka  <jh@suse.cz>
+
+	* basic-block.h (REG_BR_PROB_BASE): Define.
+	* cgraph.c (cgraph_create_edge): Initialize loop_nest and count.
+	(dump_cgraph_node): Dump count.
+	(cgraph_clone_edge): Rescale counts.
+	(cgraph_clone_node): Likewise.
+	* cgraph.h: Include basic-block.h
+	(cgraph_node): Add count.
+	(cgraph_edge): Add count and loop_nest.
+	(cgraph_node, cgraph_edge, cgraph_clone_edge, cgraph_clone_node):
+	Update prototypes.
+	* cgraphunit.c: Kill now redundant inlining comment.
+	(cgraph_create_edges): Make static, maintain current basic block;
+	fix pasto.
+	(record_call_1): Fill in new fields.
+	* ipa-inline.c (cgraph_clone_inlined_nodes): Update call of
+	cgraph_clone_node.
+	(cgraph_decide_recursive_inlining): Likewise.
+	* rtl.h (REG_BR_PROB_BASE): Kill.
+	* tree-inline.c (copy_body_r): Update call of cgraph_clone_edge.
+	(expand_call_inline): Update call of cgraph_create_edge.
+	* tree-optimize.c (tree_rest_of_compilation): Likewise.
+
+2005-05-19  Nick Clifton  <nickc@redhat.com>
+
+	* config/rs6000/eabispe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Use the
+	new rs6000_explicit_options structure.
+
+2005-05-19  Richard Henderson  <rth@redhat.com>
+
+	PR target/21390
+	* config/alpha/alpha.c (alpha_emit_conditional_branch): Move 
+	TFmode code swapping ...
+	(alpha_emit_xfloating_compare): ... here.
+	(alpha_emit_setcc): Update to match.
+	(alpha_lookup_xfloating_lib_func): Assert TARGET_HAS_XFLOATING_LIBS.
+
+2005-05-19  Richard Henderson  <rth@redhat.com>
+
+	PR target/21657
+	* config/ia64/ia64.c (ia64_expand_tls_address): Remove extra
+	memory indirection in local-dynamic case.
+
+2005-05-18  Jie Zhang  <jie.zhang@analog.com>
+	    Bernd Schmidt  <bernd.schmidt@analog.com>
+
+	* config/bfin/bfin.c (branch_dest): Add comment why it's
+	necessary.
+	* config/bfin/bfin.md (attr "length" default): Change the offset of
+	forward conditional branch of length 4 from 4096 to 4092.
+
 2005-05-19  Jan Beulich  <jbeulich@novell.com>
 
 	* unwind-compat.c: Include tconfig.h and tsystem.h.
@@ -44,12 +102,12 @@
 	
 2005-05-18  Richard Henderson  <rth@redhat.com>
 
-        * tree-ssa-forwprop.c (cfg_changed): New.
-        (tidy_after_forward_propagate_addr): New.
-        (forward_propagate_addr_expr): Use it, and fold_stmt_inplace.
-        (forward_propagate_addr_into_variable_array_index): Likewise.
-        (tree_ssa_forward_propagate_single_use_vars): Cleanup cfg if
-        edges were removed.
+	* tree-ssa-forwprop.c (cfg_changed): New.
+	(tidy_after_forward_propagate_addr): New.
+	(forward_propagate_addr_expr): Use it, and fold_stmt_inplace.
+	(forward_propagate_addr_into_variable_array_index): Likewise.
+	(tree_ssa_forward_propagate_single_use_vars): Cleanup cfg if
+	edges were removed.
 
 2005-05-18  Geoffrey Keating  <geoffk@apple.com>
 
--- /usr/src/SpecTests/sandbox-britten-memory/x86_64/mem-result/ChangeLog.cp	2005-05-19 06:52:33.000000000 +0000
+++ /usr/src/SpecTests/sandbox-britten-memory/gcc/gcc/cp/ChangeLog	2005-05-19 19:39:46.000000000 +0000
@@ -1,3 +1,9 @@
+2005-05-19  Jakub Jelinek  <jakub@redhat.com>
+
+	PR c++/21495
+	* decl.c (grokdeclarator): Fix "storage class specified for"
+	error reporting.
+
 2005-05-19  Kazu Hirata  <kazu@cs.umass.edu>
 
 	* parser.c: Fix comment typos.


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.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]