Collect with 32bit JVM failed

Question

Hi,
I am using collect tool from Sun Studio 12, i used it succeccfully on my system using 64-bit Sun JDK 1.6
but when i try to use it to collect Hardware counter data using 32-bit jvm it failes and shows error while i trying to analyze collect data using "Analyzer" tool
The error message it shows is *"test.1.er***Collector Error: Initializing Hardware counter profiling failed"*
OS version
$ uname -a
Linux m02 2.6.18-92.el5.src-PAPI #1 SMP Tue Jan 27 10:57:40 CET 2009 x86_64 x86_64 x86_64 GNU/Linux
Java version (32-bit version that is not working)
]$ java -version
java version "1.6.0_10"
Java(TM) SE Runtime Environment (build 1.6.0_10-b33)
Java HotSpot(TM) Server VM (build 11.0-b15, mixed mode)
the other Java version that is working fine (64-bit version)
]$ java -version
java version "1.6.0_13"
Java(TM) SE Runtime Environment (build 1.6.0_13-b03)
Java HotSpot(TM) 64-Bit Server VM (build 11.3-b02, mixed mode)
collect command output
]$ collect
NOTE: Linux-64-bit, 32 CPUs, CentOS_5.2 system "mi02" is supported by the Performance tools.
NOTE: The J2SE[tm] version 1.6.0_10 found at java (picked by PATH) is supported by the Performance tools.
usage:  collect <args> target <target-args>
        Sun Analyzer 7.7 Linux_i386 2009/06/03 [64-bit]
  -p <interval> specify clock-profiling
        clock profiling interval range on this system is from
        10.000 to 1000.000 millisec.; resolution is 10.000 millisec.
  -h <ctr_def>...[,<ctr_n_def>]
        specify HW counter profiling for up to 4 HW counters
        see below for more details
  -s <threshold>        specify synchronization wait tracing
  -r <option>   specify thread analyzer experiment; see man page
  -H {on|off}   specify heap tracing
  -M {version|off}      specify an MPI experiment
        Supported MPI versions: CT8.2, CT8.1, CT8, CT7, CT7.1, OPENMPI, MPICH2, MVAPICH2
  -m {on|off}   specify MPI tracing
  -j {on|off|path}      specify Java profiling
  -J <java-args>        specify arguments to Java for Java profiling
  -t <duration> specify time over which to record data
  -x    specify leaving the target waiting for a debugger attach
  -n    dry run -- don't run target or collect performance data
  -y <signal>[,r]       specify delayed initialization and pause/resume signal
        When set, the target starts in paused mode;
          if the optional r is provided, it starts in resumed mode
  -F {on|off|all|=<regex>}      specify following descendant processes
  -A {on|off|copy}      specify archiving of load-objects; default is on
  -S <interval> specify periodic sampling interval (secs.)
  -L <size>     specify experiment size limit (MB.)
  -l <signal>   specify signal for samples
  -o <expt>     specify experiment name
  -d <directory>        specify experiment directory
  -g <groupname>        specify experiment group
  -O <file>     redirect all of collect's output to file
  -v    print expanded log of processing
  -C <label>    specify comment label (up to 10 may appear)
  -R    show the README file and exit
  -V    print version number and exit
Default experiment:
        expt_name = test.3.er
        clock profiling enabled, 10.000 millisec.
        descendant processes will not be followed
        periodic sampling, 1 secs.
        experiment size limit 2000 MB.
        experiment archiving: on
        data descriptor: "p:10000;S:1;L:2000;A:1;"
                host: `m02', cpuver = 2501, ncpus = 32, clock frequency 1200 MHz.
                memory:  16518000 pages @ 4096 bytes = 64523 MB.
Specifying HW counters on `AMD Family 10h':
    <ctr_def> == [+]<ctr>[~<attr>=<val>]...[~<attrN>=<valN>][/<reg#>][,<interval>]
       <+>
          for memory-related counters, attempt to backtrack to find
          the triggering instruction and the virtual and physical
          addresses of the memory reference
       <ctr>
          counter name, must be selected from the available counters
          listed below.  On most systems, if a counter is not listed
          below, it may still be specified by its numeric value
       <attr>=<val>
          optional attribute where <val> can be in decimal or hex
          format, and <attr> can be one of:
             'umask'
             'os'
             'edge'
             'pc'
             'inv'
             'cmask'
       <reg#>
          forces use of a specific hardware register.  If not specified,
          collect will attempt to place the counter into the first
          available register and as a result, may be unable to place
          subsequent counters due to register conflicts.
       <interval> == {on|hi|lo|<value>}
          `on' selects the default rate, listed below
          `hi' specifies an interval ~10 times shorter than `on'
          `lo' specifies an interval ~10 times longer than `on'
Aliased HW counters available for profiling:
    cycles[/{0|1|2|3}],99999989 (`CPU Cycles', alias for BU_cpu_clk_unhalted; CPU-cycles)
    insts[/{0|1|2|3}],9999991 (`Instructions Executed', alias for FR_retired_x86_instr_w_excp_intr; events)
    ic[/{0|1|2|3}],100003 (`I$ Refs', alias for IC_fetch; events)
    icm[/{0|1|2|3}],100003 (`I$ Misses', alias for IC_miss; events)
    itlbh[/{0|1|2|3}],100003 (`ITLB Hits', alias for IC_itlb_L1_miss_L2_hit; events)
    itlbm[/{0|1|2|3}],100003 (`ITLB Misses', alias for IC_itlb_L1_miss_L2_miss; events)
    eci[/{0|1|2|3}],1000003 (`E$ Instr. Refs', alias for BU_internal_L2_req~umask=0x1; events)
    ecim[/{0|1|2|3}],10007 (`E$ Instr. Misses', alias for BU_fill_req_missed_L2~umask=0x1; events)
    dc[/{0|1|2|3}],1000003 (`D$ Refs', alias for DC_access; load events)
    dcm[/{0|1|2|3}],100003 (`D$ Misses', alias for DC_miss; load events)
    dtlbh[/{0|1|2|3}],100003 (`DTLB Hits', alias for DC_dtlb_L1_miss_L2_hit; load-store events)
    dtlbm[/{0|1|2|3}],100003 (`DTLB Misses', alias for DC_dtlb_L1_miss_L2_miss; load-store events)
    ecd[/{0|1|2|3}],1000003 (`E$ Data Refs', alias for BU_internal_L2_req~umask=0x2; load-store events)
    ecdm[/{0|1|2|3}],10007 (`E$ Data Misses', alias for BU_fill_req_missed_L2~umask=0x2; load-store events)
    fpadd[/{0|1|2|3}],1000003 (`FP Adds', alias for FP_dispatched_fpu_ops~umask=0x1; events)
    fpmul[/{0|1|2|3}],1000003 (`FP Muls', alias for FP_dispatched_fpu_ops~umask=0x2; events)
    fpustall[/{0|1|2|3}],1000003 (`FPU Stall Cycles', alias for FR_dispatch_stall_fpu_full; CPU-cycles)
    memstall[/{0|1|2|3}],1000003 (`Memory Unit Stall Cycles', alias for FR_dispatch_stall_ls_full; CPU-cycles)
    PAPI_l1_dcm[/{0|1|2|3}],100003 (`Level 1 data cache misses'; load-store events)
    PAPI_l1_icm[/{0|1|2|3}],100003 (`Level 1 instruction cache misses'; events)
    PAPI_l2_dcm[/{0|1|2|3}],100003 (`Level 2 data cache misses'; load-store events)
Raw HW counters available for profiling:
  *......  { I removed raw counter list to shorten this post}*
See section 3.15 of the "BIOS and Kernel Developer's Guide (BKDG) For AMD Family 10h Processors,"
AMD publication #31116
See the collect.1 man page for more information
analyzer tool version
$ analyzer -v
analyzer: Sun Analyzer 7.7 Linux_i386 2009/06/03
Any idea what is wrong with my configuration? Can we use collect for 32-bit and java apps?
Thanks a lot
Edited by: allo6 on Aug 20, 2009 3:25 AM

Answer 1

hi,
how to compile 32 bits libraries on a Centos5 64 bits running on AMD 64 bits (Family 10h)
running
make CC="gcc -m32" CFLAGS="-m32" LDFLAGS="-m32" ARCH=i386
or just
make CC="gcc -m32" CFLAGS="-m32" LDFLAGS="-m32" generate a 32 bit version
/usr/local/perfctr32/lib/libperfctr.so.5.2.6.35: ELF 32-bit LSB shared object, Intel 80386, version 1 (SYSV), not stripped
but the "collect"command (from sun studio for collecting HW values) returns HW counter not suported.
the above generated library is 32 bit but it is intel architecture, The architecture i have is AMD, what "ARCH" flag to spefify for this
if i do not specify any ARCH value there there is this error in the make command
ake[4]: Entering directory `perfctr-2.6.x/examples/perfex'
gcc -m32 -m32 perfex.o x86.o ../../usr.lib/libperfctr.a -o perfex
../../usr.lib/libperfctr.a(event_set_x86.o):(.rodata+0x24): undefined reference to `perfctr_p5_event_set'
../../usr.lib/libperfctr.a(event_set_x86.o):(.rodata+0x28): undefined reference to `perfctr_p5mmx_event_set'
../../usr.lib/libperfctr.a(event_set_x86.o):(.rodata+0x2c): undefined reference to `perfctr_ppro_event_set'
../../usr.lib/libperfctr.a(event_set_x86.o):(.rodata+0x30): undefined reference to `perfctr_p2_event_set'
../../usr.lib/libperfctr.a(event_set_x86.o):(.rodata+0x34): undefined reference to `perfctr_p3_event_set'
../../usr.lib/libperfctr.a(event_set_x86.o):(.rodata+0x38): undefined reference to `perfctr_mii_event_set'