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:
  Peak amount of GGC memory allocated before garbage collecting increased from 9553k to 9566k, overall 0.14%
  Peak amount of GGC memory still allocated after garbage collectin increased from 8901k to 8913k, overall 0.13%
  Amount of memory still referenced at the end of compilation increased from 6728k to 6740k, overall 0.18%
    Overall memory needed: 25305k -> 25324k
    Peak memory use before GGC: 9553k -> 9566k
    Peak memory use after GGC: 8901k -> 8913k
    Maximum of released memory in single GGC run: 2649k
    Garbage: 40059k -> 40060k
    Leak: 6728k -> 6740k
    Overhead: 5736k -> 5738k
    GGC runs: 313

comparing combine.c compilation at -O1 level:
    Overall memory needed: 26884k -> 26892k
    Peak memory use before GGC: 17423k -> 17434k
    Peak memory use after GGC: 17243k -> 17255k
    Maximum of released memory in single GGC run: 2310k -> 2309k
    Garbage: 62581k -> 62568k
    Leak: 6877k -> 6881k
    Overhead: 7484k -> 7485k
    GGC runs: 393

comparing combine.c compilation at -O2 level:
  Amount of memory still referenced at the end of compilation increased from 6948k to 6968k, overall 0.29%
    Overall memory needed: 26884k -> 26892k
    Peak memory use before GGC: 17426k -> 17436k
    Peak memory use after GGC: 17244k -> 17255k
    Maximum of released memory in single GGC run: 2389k -> 2388k
    Garbage: 79995k -> 79972k
    Leak: 6948k -> 6968k
    Overhead: 10012k -> 10012k
    GGC runs: 448

comparing combine.c compilation at -O3 level:
  Amount of memory still referenced at the end of compilation increased from 7026k to 7038k, overall 0.17%
    Overall memory needed: 26884k -> 26892k
    Peak memory use before GGC: 18263k -> 18273k
    Peak memory use after GGC: 17941k -> 17952k
    Maximum of released memory in single GGC run: 3378k
    Garbage: 111250k -> 111230k
    Leak: 7026k -> 7038k
    Overhead: 13694k -> 13694k
    GGC runs: 502

comparing insn-attrtab.c compilation at -O0 level:
  Amount of memory still referenced at the end of compilation increased from 10119k to 10131k, overall 0.12%
    Overall memory needed: 80932k -> 80948k
    Peak memory use before GGC: 69495k -> 69508k
    Peak memory use after GGC: 45032k -> 45044k
    Maximum of released memory in single GGC run: 36220k
    Garbage: 146441k -> 146427k
    Leak: 10119k -> 10131k
    Overhead: 19749k -> 19750k
    GGC runs: 247

comparing insn-attrtab.c compilation at -O1 level:
  Amount of memory still referenced at the end of compilation increased from 10042k to 10054k, overall 0.12%
    Overall memory needed: 109152k -> 109164k
    Peak memory use before GGC: 91964k -> 91973k
    Peak memory use after GGC: 81062k -> 81071k
    Maximum of released memory in single GGC run: 32143k
    Garbage: 296324k -> 296337k
    Leak: 10042k -> 10054k
    Overhead: 36075k -> 36077k
    GGC runs: 245

comparing insn-attrtab.c compilation at -O2 level:
  Amount of memory still referenced at the end of compilation increased from 10144k to 10156k, overall 0.12%
    Overall memory needed: 122544k -> 122560k
    Peak memory use before GGC: 109346k -> 109356k
    Peak memory use after GGC: 80963k -> 80974k
    Maximum of released memory in single GGC run: 31578k
    Garbage: 380310k -> 380308k
    Leak: 10144k -> 10156k
    Overhead: 47727k -> 47729k
    GGC runs: 274

comparing insn-attrtab.c compilation at -O3 level:
  Amount of memory still referenced at the end of compilation increased from 10148k to 10160k, overall 0.12%
    Overall memory needed: 122596k -> 122608k
    Peak memory use before GGC: 109373k -> 109384k
    Peak memory use after GGC: 80991k -> 81001k
    Maximum of released memory in single GGC run: 31906k -> 31908k
    Garbage: 380923k -> 380919k
    Leak: 10148k -> 10160k
    Overhead: 47928k -> 47930k
    GGC runs: 276

comparing Gerald's testcase PR8361 compilation at -O0 level:
    Overall memory needed: 118236k -> 118252k
    Peak memory use before GGC: 95015k -> 95027k
    Peak memory use after GGC: 94067k -> 94080k
    Maximum of released memory in single GGC run: 20299k
    Garbage: 223498k -> 223500k
    Leak: 49456k -> 49470k
    Overhead: 37086k -> 37088k
    GGC runs: 369

comparing Gerald's testcase PR8361 compilation at -O1 level:
    Overall memory needed: 108388k -> 108428k
    Peak memory use before GGC: 95130k -> 95143k
    Peak memory use after GGC: 93139k -> 93151k
    Maximum of released memory in single GGC run: 20158k
    Garbage: 564020k -> 564040k
    Leak: 52234k -> 52248k
    Overhead: 63604k -> 63608k
    GGC runs: 532

comparing Gerald's testcase PR8361 compilation at -O2 level:
    Overall memory needed: 108524k -> 108536k
    Peak memory use before GGC: 95131k -> 95143k
    Peak memory use after GGC: 93139k -> 93152k
    Maximum of released memory in single GGC run: 20158k
    Garbage: 657613k -> 655293k
    Leak: 53052k -> 53066k
    Overhead: 75239k -> 75038k
    GGC runs: 597 -> 598

comparing Gerald's testcase PR8361 compilation at -O3 level:
    Overall memory needed: 110764k -> 110784k
    Peak memory use before GGC: 96525k -> 96537k
    Peak memory use after GGC: 94567k -> 94579k
    Maximum of released memory in single GGC run: 20583k -> 20582k
    Garbage: 713924k -> 713909k
    Leak: 53941k -> 53955k
    Overhead: 79320k -> 79299k
    GGC runs: 611 -> 609

Head of the ChangeLog is:

--- /usr/src/SpecTests/sandbox-britten-memory/x86_64/mem-result/ChangeLog	2006-01-31 10:19:20.000000000 +0000
+++ /usr/src/SpecTests/sandbox-britten-memory/gcc/gcc/ChangeLog	2006-01-31 20:39:24.000000000 +0000
@@ -1,3 +1,42 @@
+2006-01-31  Nicolas Pitre  <nico@cam.org>
+
+	* arm.md (smaxsi3): Make aware of smax_m1.
+	(smax_m1): New.
+
+2006-01-31  Nathan Sidwell  <nathan@codesourcery.com>
+
+	* gcc/config/m68k/lb1sf68.asm: Use moveq to load small constants.
+
+2006-01-31  Richard Guenther  <rguenther@suse.de>
+
+	* doc/invoke.texi (-msselibm): Document.
+	* target.h (expand_library_builtin): New target hook.
+	* builtins.c (expand_builtin): Use expand_library_builtin.
+	(default_expand_library_builtin): New function.
+	* gcc.c (LINK_GCC_MATH_SPEC): Define.
+	(LINK_COMMAND_SPEC): Handle %(link_gcc_math).
+	(link_gcc_math_spec): Declare.
+	(static_specs): Add link_gcc_math_spec.
+	* expr.h (default_expand_library_builtin): Declare.
+	* target-def.h (TARGET_EXPAND_LIBRARY_BUILTIN): Define.
+	(TARGET_INITIALIZER): Add TARGET_EXPAND_LIBRARY_BUILTIN.
+	* config/i386/i386.h: Provide LINK_GCC_MATH_SPEC.
+	* config/i386/i386.opt (msselibm): New target option.
+	* config/i386/i386.c (ix86_builtin_function_variants): New array.
+	(ix86_init_sse_abi_builtins): New function.
+	(ix86_expand_library_builtin): Likewise.
+	(TARGET_EXPAND_LIBRARY_BUILTIN): Define.
+	(override_options): Handle error conditions wrt -msselibm.
+	(ix86_builtins): Add function codes for SSE2 ABI builtins.
+	(ix86_init_builtins): Call ix86_init_sse_abi_builtins.
+	* doc/extend.texi (__builtin_sse2_*): Document new target specific
+	builtins.
+
+2006-01-31  Richard Guenther  <rguenther@suse.de>
+	Paolo Bonzini  <bonzini@gnu.org>
+
+	* doc/install.texi (--disable-libgcc-math): Document.
+
 2006-01-30  Marcin Dalecki  <martin@dalecki.de>
 
 	* expr.h (expand_normal): new inline function.


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]