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]

GCC memory consumption increased by recent patch!


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


comparing combine.c compilation at -O0 level:
    Overall memory needed: 25261k -> 25253k
    Peak memory use before GGC: 9336k -> 9316k
    Peak memory use after GGC: 8647k -> 8627k
    Maximum of released memory in single GGC run: 2930k -> 2914k
    Garbage: 42989k -> 42954k
    Leak: 6089k -> 6081k
    Overhead: 5653k -> 5652k
    GGC runs: 364

comparing combine.c compilation at -O1 level:
    Overall memory needed: 26725k -> 26645k
    Peak memory use before GGC: 9479k -> 9439k
    Peak memory use after GGC: 8871k -> 8831k
    Maximum of released memory in single GGC run: 2063k -> 2054k
    Garbage: 73204k -> 73211k
    Leak: 6677k -> 6668k
    Overhead: 11471k -> 11521k
    GGC runs: 577

comparing combine.c compilation at -O2 level:
    Overall memory needed: 29953k -> 29929k
    Peak memory use before GGC: 12772k -> 12732k
    Peak memory use after GGC: 12612k -> 12571k
    Maximum of released memory in single GGC run: 2579k
    Garbage: 88370k -> 88249k
    Leak: 6431k -> 6430k
    Overhead: 16012k -> 15983k
    GGC runs: 578 -> 576

comparing combine.c compilation at -O3 level:
    Overall memory needed: 21468k -> 21344k
    Peak memory use before GGC: 13445k -> 13412k
    Peak memory use after GGC: 12721k -> 12688k
    Maximum of released memory in single GGC run: 3440k -> 3433k
    Garbage: 118795k -> 118717k
    Leak: 7032k -> 7014k
    Overhead: 21196k -> 21184k
    GGC runs: 644 -> 642

comparing insn-attrtab.c compilation at -O0 level:
    Overall memory needed: 118640k
    Peak memory use before GGC: 79535k -> 79536k
    Peak memory use after GGC: 46180k -> 46181k
    Maximum of released memory in single GGC run: 43441k
    Garbage: 163172k -> 163201k
    Leak: 10643k -> 10648k
    Overhead: 20458k -> 20461k
    GGC runs: 307

comparing insn-attrtab.c compilation at -O1 level:
    Overall memory needed: 130584k -> 130628k
    Peak memory use before GGC: 90827k
    Peak memory use after GGC: 70897k -> 70898k
    Maximum of released memory in single GGC run: 42004k
    Garbage: 462759k -> 462780k
    Leak: 11074k -> 11079k
    Overhead: 74732k -> 74729k
    GGC runs: 463

comparing insn-attrtab.c compilation at -O2 level:
  Amount of memory still referenced at the end of compilation increased from 10956k to 10968k, overall 0.11%
    Overall memory needed: 209824k -> 209852k
    Peak memory use before GGC: 105788k -> 105792k
    Peak memory use after GGC: 86457k -> 86461k
    Maximum of released memory in single GGC run: 35798k -> 35799k
    Garbage: 512367k -> 512332k
    Leak: 10956k -> 10968k
    Overhead: 82807k -> 82796k
    GGC runs: 385 -> 384

comparing insn-attrtab.c compilation at -O3 level:
    Overall memory needed: 209832k -> 209860k
    Peak memory use before GGC: 105796k -> 105793k
    Peak memory use after GGC: 86466k -> 86462k
    Maximum of released memory in single GGC run: 35798k -> 35799k
    Garbage: 513748k -> 513743k
    Leak: 11011k -> 11006k
    Overhead: 82995k -> 82984k
    GGC runs: 394

comparing Gerald's testcase PR8361 compilation at -O0 level:
    Overall memory needed: 112008k
    Peak memory use before GGC: 88771k
    Peak memory use after GGC: 87648k
    Maximum of released memory in single GGC run: 19623k
    Garbage: 254347k -> 254347k
    Leak: 58643k
    Overhead: 46746k -> 46746k
    GGC runs: 367

comparing Gerald's testcase PR8361 compilation at -O1 level:
    Overall memory needed: 100816k
    Peak memory use before GGC: 87755k
    Peak memory use after GGC: 86603k
    Maximum of released memory in single GGC run: 19232k
    Garbage: 693432k -> 693432k
    Leak: 60742k
    Overhead: 130760k -> 130760k
    GGC runs: 625

comparing Gerald's testcase PR8361 compilation at -O2 level:
    Overall memory needed: 101892k
    Peak memory use before GGC: 87755k
    Peak memory use after GGC: 86603k
    Maximum of released memory in single GGC run: 19231k
    Garbage: 744918k -> 744918k
    Leak: 61288k
    Overhead: 154237k -> 154237k
    GGC runs: 656

comparing Gerald's testcase PR8361 compilation at -O3 level:
    Overall memory needed: 118548k
    Peak memory use before GGC: 100244k
    Peak memory use after GGC: 89103k
    Maximum of released memory in single GGC run: 19979k
    Garbage: 849559k -> 849559k
    Leak: 61699k
    Overhead: 166038k -> 166038k
    GGC runs: 640

Head of changelog is:

--- /usr/src/SpecTests/sandbox-britten-memory/x86_64/mem-result/ChangeLog	2004-10-17 15:02:10.000000000 +0000
+++ /usr/src/SpecTests/sandbox-britten-memory/gcc/gcc/ChangeLog	2004-10-17 20:31:06.000000000 +0000
@@ -1,3 +1,50 @@
+2004-10-17  Andrew Pinski  <pinskia@physics.uc.edu>
+
+	PR middle-end/17925
+	* cfgexpand.c (expand_gimple_cond_expr): Emit line notes for next basic
+	block if there is a goto with a locus.
+
+2004-10-17  Kazu Hirata  <kazu@cs.umass.edu>
+
+	* config/elfos.h, config/gofast.h, config/interix.h,
+	config/netbsd.h, config/svr3.h, config/vxworks.h,
+	config/alpha/alpha-modes.def, config/alpha/alpha-protos.h,
+	config/alpha/ev4.md, config/alpha/ev5.md, config/alpha/ev6.md,
+	config/alpha/netbsd.h, config/arm/arm-modes.def,
+	config/arm/unknown-elf.h, config/c4x/c4x-modes.def,
+	config/cris/aout.h, config/frv/frv-modes.def,
+	config/i386/beos-elf.h, config/i386/gas.h,
+	config/i386/i386-coff.h, config/i386/i386-modes.def,
+	config/i386/linux.h, config/i386/linux64.h,
+	config/i386/mingw32.h, config/i386/netbsd-elf.h,
+	config/i386/netbsd64.h, config/i386/sco5.h,
+	config/i386/sol2.h, config/i386/uwin.h, config/i860/i860.md,
+	config/ia64/ia64-modes.def, config/ia64/itanium1.md,
+	config/ia64/itanium2.md, config/m68k/m68k-modes.def,
+	config/mips/mips-modes.def, config/mips/sdb.h,
+	config/mips/vr.h, config/mips/vxworks.h,
+	config/pa/pa-hpux10.h, config/pa/pa-hpux11.h,
+	config/pa/pa32-linux.h, config/rs6000/40x.md,
+	config/rs6000/7450.md, config/rs6000/8540.md,
+	config/rs6000/linuxspe.h, config/rs6000/power4.md,
+	config/rs6000/rs6000-modes.def, config/rs6000/rtems.h,
+	config/rs6000/spe.h, config/s390/2064.md, config/s390/linux.h,
+	config/s390/s390-modes.def, config/sh/elf.h,
+	config/sh/sh-modes.def, config/sh/sh64.h,
+	config/sparc/ultra1_2.md, config/sparc/ultra3.md,
+	config/stormy16/stormy16.md, config/v850/v850-protos.h,
+	config/vax/vax.md: Update copyright.
+
+2004-10-17  Kazu Hirata  <kazu@cs.umass.edu>
+
+	* expr.c (expand_expr_real_1): Remove an obsolete comment.
+
+2004-10-17  Andrew Pinski  <pinskia@physics.uc.edu>
+
+	* c-typeck.c (default_function_array_conversion): Always create
+	&a[0] for array types.
+	(build_unary_op): Do not fold &a[x] into a + x.
+
 2004-10-17  Jakub Jelinek  <jakub@redhat.com>
 
 	* pointer-set.c (hash1): Use integer part of 2^64 / phi

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

The results can be reproduced by building 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.

Yours testing script.


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