From: Howard C. <hy...@sy...> - 2006-12-30 12:52:36
|
I'm getting no callgraph results on a 2.6.18.3 kernel, with AMD X2 3800+ processor. I had oprofile 0.8 on here originally, then downloaded and built 0.9.2, no luck. "opreport -c" complains about a lot of things that it doesn't complain about for a simple "opreport -l". : warning: /commoncap could not be found. warning: /ehci_hcd could not be found. warning: /home/hyc/OD/hobj/servers/slapd/back-hdb/.libs/back_hdb-2-devel.so.0.0. 0/CPU_CLK_UNHALTED.100000.0.all.all.all/{dep}/{root}/var/lib/oprofile/samples/cu rrent/{root}/home/hyc/OD/hobj/servers/slapd/back-hdb/.libs/back_hdb-2-devel.so.0 .0.0/{dep}/{root}/home/hyc/OD/hobj/servers/slapd/back-hdb/.libs/back_hdb-2-devel .so.0.0.0/CPU_CLK_UNHALTED.100000.0.all.all.all could not be found. warning: /home/hyc/OD/hobj/servers/slapd/slapd/CPU_CLK_UNHALTED.100000.0.all.all .all/{dep}/{root}/var/lib/oprofile/samples/current/{root}/home/hyc/OD/hobj/serve rs/slapd/slapd/{dep}/{root}/home/hyc/OD/hobj/servers/slapd/slapd/CPU_CLK_UNHALTE D.100000.0.all.all.all could not be found. warning: /home/local/BerkeleyDB.4.6/lib/libdb-4.6.so.nodebug/CPU_CLK_UNHALTED.10 0000.0.all.all.all/{dep}/{root}/var/lib/oprofile/samples/current/{root}/home/loc al/BerkeleyDB.4.6/lib/libdb-4.6.so.nodebug/{dep}/{root}/home/local/BerkeleyDB.4. 6/lib/libdb-4.6.so.nodebug/CPU_CLK_UNHALTED.100000.0.all.all.all could not be fo und. warning: /home/local/bin/oprofiled/CPU_CLK_UNHALTED.100000.0.all.all.all/{dep}/{ root}/var/lib/oprofile/samples/current/{root}/home/local/bin/oprofiled/{dep}/{ro ot}/home/local/bin/oprofiled/CPU_CLK_UNHALTED.100000.0.all.all.all could not be found. Does this ring any bells for anyone? -- -- Howard Chu Chief Architect, Symas Corp. http://www.symas.com Director, Highland Sun http://highlandsun.com/hyc OpenLDAP Core Team http://www.openldap.org/project/ |
From: Konstantin S. <kon...@gm...> - 2008-01-29 23:28:16
|
Hi, As far as I understand, oprofile's callgraph does not work on x86_64 now. Is there a chance that it will be supported in some near future? Thanks, --kcc |
From: John L. <le...@mo...> - 2008-01-30 01:39:38
|
On Tue, Jan 29, 2008 at 03:28:19PM -0800, Konstantin Serebryany wrote: > As far as I understand, oprofile's callgraph does not work on x86_64 now. This is not true as far as I'm aware, unless somebody's broken it (far from impossible, sadly) regards john |
From: Konstantin S. <kon...@gm...> - 2008-01-30 01:53:01
|
Hmm. I have a machine which works in mixed mode (2.6.18.5 mixed64-32). oprofile's flat profile works fine on it, but the callgraph does not. It shows me something like this: $ opreport -l -c ./bzip2 ...------------------------------------------------- 98372 51.8719 BZ2_blockSort 98372 100.000 BZ2_blockSort [self] ------------------------------------------------------------------------------- 71317 37.6057 fallbackSort 71317 100.000 fallbackSort [self] But if I boot this machine in pure 32-bit mode the callgraph works. I tried oprofile 0.9.2 and 0.9.3 --kcc On Jan 29, 2008 5:39 PM, John Levon <le...@mo...> wrote: > On Tue, Jan 29, 2008 at 03:28:19PM -0800, Konstantin Serebryany wrote: > > > As far as I understand, oprofile's callgraph does not work on x86_64 > now. > > This is not true as far as I'm aware, unless somebody's broken it (far > from impossible, sadly) > > regards > john > |
From: Christopher L. <cl...@an...> - 2008-01-30 02:28:38
|
On Tue, Jan 29, 2008 at 05:53:03PM -0800, Konstantin Serebryany wrote: > Hmm. > I have a machine which works in mixed mode (2.6.18.5 mixed64-32). > oprofile's flat profile works fine on it, but the callgraph does not. > It shows me something like this: > > $ opreport -l -c ./bzip2 > ...------------------------------------------------- > 98372 51.8719 BZ2_blockSort > 98372 100.000 BZ2_blockSort [self] > ------------------------------------------------------------------------------- > 71317 37.6057 fallbackSort > 71317 100.000 fallbackSort [self] For the record, I also see this behaviour. I can also remember seeing it atleast as far back as 4 months ago (from when I lasted tried using callgraph support). Unfortunately this isn't a multilib system so I can't try it under 32. # opreport -gl -c '*bz2*' CPU: AMD64 processors, speed 2400 MHz (estimated) Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 100000 warning: /var/lib/oprofile/samples/current/{root}/bin/bzip2/{dep}/{root}/lib64/libbz2.so.1.0.4/CPU_CLK_UNHALTED.100000.0.all.all.all is not in a usable binary format. samples % linenr info image name app name symbol name ------------------------------------------------------------------------------- 174693 54.0678 bzlib.c:808 libbz2.so.1.0.4 bzip2 BZ2_bzDecompress 174693 100.000 bzlib.c:808 libbz2.so.1.0.4 bzip2 BZ2_bzDecompress [self] ------------------------------------------------------------------------------- 148156 45.8545 decompress.c:107 libbz2.so.1.0.4 bzip2 BZ2_decompress 148156 100.000 decompress.c:107 libbz2.so.1.0.4 bzip2 BZ2_decompress [self] ------------------------------------------------------------------------------- 154 0.0477 bzlib.c:1165 libbz2.so.1.0.4 bzip2 BZ2_bzRead 154 100.000 bzlib.c:1165 libbz2.so.1.0.4 bzip2 BZ2_bzRead [self] ------------------------------------------------------------------------------- 95 0.0294 huffman.c:177 libbz2.so.1.0.4 bzip2 BZ2_hbCreateDecodeTables 95 100.000 huffman.c:177 libbz2.so.1.0.4 bzip2 BZ2_hbCreateDecodeTables [self] ------------------------------------------------------------------------------- 2 6.2e-04 (no location information) libbz2.so.1.0.4 bzip2 .plt 2 100.000 (no location information) libbz2.so.1.0.4 bzip2 .plt [self] ------------------------------------------------------------------------------- # uname -a Linux ns1 2.6.24 #1 SMP Sat Jan 26 09:32:00 GMT 2008 x86_64 Dual-Core AMD Opteron(tm) Processor 1216 AuthenticAMD GNU/Linux # opcontrol --version opcontrol: oprofile 0.9.3 compiled on Jan 27 2008 21:21:22 Oprofile is also compiled into the kernel (not a module). -cl |
From: Konstantin S. <kon...@gm...> - 2008-03-12 09:55:22
|
I just found out that compiling a program with -fno-omit-frame-pointer makes the problem go away. Might be nice to mention it in FAQ. Thanks, --kcc On Thu, Mar 6, 2008 at 11:12 AM, Konstantin Serebryany <kon...@gm...> wrote: > Hi again, > > Sorry for disturbing you all, but is there any hope to have callgraph > working on x86_64? > > Thanks, > > --kcc > > > > > > On Tue, Feb 12, 2008 at 2:32 PM, Konstantin Serebryany > <kon...@gm...> wrote: > > > I was also having callgraph trouble on my x86_64 machine. I was able to > > > get it work, however, by removing gcc optimization's from the 'libop' > > > and 'libdb' Makefiles. More specifically: > > > > Thanks Jerome, > > > > Unfortunately, it does not seem to help in my case. > > I still get profiles like > > 131305 47.3557 exe32 BZ2_blockSort > > 131305 100.000 exe32 BZ2_blockSort [self] > > > > I also (sometimes !) get 'opreport error: basic_string::erase'. > > > > --kcc > > > |
From: Jerome W. <whi...@em...> - 2008-02-06 17:13:05
|
Christopher Layne wrote: > On Tue, Jan 29, 2008 at 05:53:03PM -0800, Konstantin Serebryany wrote: >> Hmm. >> I have a machine which works in mixed mode (2.6.18.5 mixed64-32). >> oprofile's flat profile works fine on it, but the callgraph does not. >> It shows me something like this: >> >> $ opreport -l -c ./bzip2 >> ...------------------------------------------------- >> 98372 51.8719 BZ2_blockSort >> 98372 100.000 BZ2_blockSort [self] >> ------------------------------------------------------------------------------- >> 71317 37.6057 fallbackSort >> 71317 100.000 fallbackSort [self] > > For the record, I also see this behaviour. I can also remember seeing it atleast as far > back as 4 months ago (from when I lasted tried using callgraph support). Unfortunately > this isn't a multilib system so I can't try it under 32. > > # opreport -gl -c '*bz2*' > CPU: AMD64 processors, speed 2400 MHz (estimated) > Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 100000 > warning: /var/lib/oprofile/samples/current/{root}/bin/bzip2/{dep}/{root}/lib64/libbz2.so.1.0.4/CPU_CLK_UNHALTED.100000.0.all.all.all is not in a usable binary format. > samples % linenr info image name app name symbol name > ------------------------------------------------------------------------------- > 174693 54.0678 bzlib.c:808 libbz2.so.1.0.4 bzip2 BZ2_bzDecompress > 174693 100.000 bzlib.c:808 libbz2.so.1.0.4 bzip2 BZ2_bzDecompress [self] > ------------------------------------------------------------------------------- > 148156 45.8545 decompress.c:107 libbz2.so.1.0.4 bzip2 BZ2_decompress > 148156 100.000 decompress.c:107 libbz2.so.1.0.4 bzip2 BZ2_decompress [self] > ------------------------------------------------------------------------------- > 154 0.0477 bzlib.c:1165 libbz2.so.1.0.4 bzip2 BZ2_bzRead > 154 100.000 bzlib.c:1165 libbz2.so.1.0.4 bzip2 BZ2_bzRead [self] > ------------------------------------------------------------------------------- > 95 0.0294 huffman.c:177 libbz2.so.1.0.4 bzip2 BZ2_hbCreateDecodeTables > 95 100.000 huffman.c:177 libbz2.so.1.0.4 bzip2 BZ2_hbCreateDecodeTables [self] > ------------------------------------------------------------------------------- > 2 6.2e-04 (no location information) libbz2.so.1.0.4 bzip2 .plt > 2 100.000 (no location information) libbz2.so.1.0.4 bzip2 .plt [self] > ------------------------------------------------------------------------------- > # uname -a > Linux ns1 2.6.24 #1 SMP Sat Jan 26 09:32:00 GMT 2008 x86_64 Dual-Core AMD Opteron(tm) Processor 1216 AuthenticAMD GNU/Linux > # opcontrol --version > opcontrol: oprofile 0.9.3 compiled on Jan 27 2008 21:21:22 > > Oprofile is also compiled into the kernel (not a module). > > -cl > > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2008. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > oprofile-list mailing list > opr...@li... > https://lists.sourceforge.net/lists/listinfo/oprofile-list > I was also having callgraph trouble on my x86_64 machine. I was able to get it work, however, by removing gcc optimization's from the 'libop' and 'libdb' Makefiles. More specifically: /tmp/oprofile-0.9.3 $> diff ./libop/Makefile{~,} 101c101,102 < CFLAGS = -g -O2 --- > # CFLAGS = -g -O2 > CFLAGS = -g 107c108,109 < CXXFLAGS = -g -O2 --- > # CXXFLAGS = -g -O2 > CXXFLAGS = -g /tmp/oprofile-0.9.3 $> diff ./libdb/Makefile{~,} 99c99,100 < CFLAGS = -g -O2 --- > # CFLAGS = -g -O2 > CFLAGS = -g 105c106,107 < CXXFLAGS = -g -O2 --- > # CXXFLAGS = -g -O2 > CXXFLAGS = -g jerome |
From: Konstantin S. <kon...@gm...> - 2008-02-12 11:32:29
|
> I was also having callgraph trouble on my x86_64 machine. I was able to > get it work, however, by removing gcc optimization's from the 'libop' > and 'libdb' Makefiles. More specifically: Thanks Jerome, Unfortunately, it does not seem to help in my case. I still get profiles like 131305 47.3557 exe32 BZ2_blockSort 131305 100.000 exe32 BZ2_blockSort [self] I also (sometimes !) get 'opreport error: basic_string::erase'. --kcc |
From: Konstantin S. <kon...@gm...> - 2008-03-06 08:12:28
|
Hi again, Sorry for disturbing you all, but is there any hope to have callgraph working on x86_64? Thanks, --kcc On Tue, Feb 12, 2008 at 2:32 PM, Konstantin Serebryany <kon...@gm...> wrote: > > I was also having callgraph trouble on my x86_64 machine. I was able to > > get it work, however, by removing gcc optimization's from the 'libop' > > and 'libdb' Makefiles. More specifically: > > Thanks Jerome, > > Unfortunately, it does not seem to help in my case. > I still get profiles like > 131305 47.3557 exe32 BZ2_blockSort > 131305 100.000 exe32 BZ2_blockSort [self] > > I also (sometimes !) get 'opreport error: basic_string::erase'. > > --kcc > |
From: Anirban S. <AS...@ze...> - 2008-03-13 17:36:57
|
>-----Original Message----- >From: opr...@li... [mailto:oprofile-list- >bo...@li...] On Behalf Of Konstantin Serebryany >Sent: Wednesday, March 12, 2008 2:55 AM >To: opr...@li... >Subject: Re: callgraph on x86_64 > >I just found out that compiling a program with -fno-omit-frame-pointer >makes the problem go away. That flag turns off -fomit-frame-pointer which is enabled of you have optimization turned on: http://gcc.gnu.org/onlinedocs/gcc-4.1.1/gcc/Optimize-Options.html Are you compiling your code with -o in general? What happens if you simply turn off optimization (without doing anything with -fno-omit-frame-pointer)? Ani |
From: Konstantin S. <kon...@gm...> - 2008-03-14 07:00:42
|
> That flag turns off -fomit-frame-pointer which is enabled of you have > optimization turned on: > http://gcc.gnu.org/onlinedocs/gcc-4.1.1/gcc/Optimize-Options.html > > Are you compiling your code with -o in general? -O2 > What happens if you > simply turn off optimization (without doing anything with > -fno-omit-frame-pointer)? Indeed, compiling with -O0 -g (and no other flags) gives correct call graph. I never tried it since profiling unoptimized program is quite useless for me (it has very different profile due to not inlining tiny functions). Thanks, --kcc |