OProfile

Hi,
I was wondering if anyone could help me. My problem is that, Oprofile
is not detecting my Xeon 3.2Ghz processor and hence I am unable to use
the hardware performance counters.
I am working with oprofile support compiled on a standard 2.6.8.1 kernel. 
my /proc/cpuinfo shows :
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 15
model           : 4
model name      : Intel(R) Xeon(TM) CPU 3.20GHz
stepping        : 1
cpu MHz         : 3202.307
cache size      : 1024 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 3
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
pni monitor ds_cpl cid
bogomips        : 6324.22

but    /dev/oprofile/cpu_type is showing "timer" 

Is the processor not supported or have I made some mistake somewhere?
I will be highly grateful if someone could help me.

thank you
Shourya

Re: oprofile not detecting P4 Xeon 800Mhz FSB processor From: William Cohen <wcohen@nc...> - 2004-12-28 14:32 Shourya P. Bhattacharya wrote: > Hi, > I was wondering if anyone could help me. My problem is that, Oprofile > is not detecting my Xeon 3.2Ghz processor and hence I am unable to use > the hardware performance counters. > I am working with oprofile support compiled on a standard 2.6.8.1 kernel. > my /proc/cpuinfo shows : > processor : 0 > vendor_id : GenuineIntel > cpu family : 15 > model : 4 > model name : Intel(R) Xeon(TM) CPU 3.20GHz > stepping : 1 > cpu MHz : 3202.307 > cache size : 1024 KB > fdiv_bug : no > hlt_bug : no > f00f_bug : no > coma_bug : no > fpu : yes > fpu_exception : yes > cpuid level : 3 > wp : yes > flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge > mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm > pni monitor ds_cpl cid > bogomips : 6324.22 > > but /dev/oprofile/cpu_type is showing "timer" > > Is the processor not supported or have I made some mistake somewhere? > I will be highly grateful if someone could help me. OProfile uses the local APIC to generate the interrupts for the performance monitoring hardware. If you have a uniprocessor kernel and the local APIC is not enabled for UP, then OProfile is going to fall back on the timer mechanism. If you are building a UP kernel check that you have CONFIG_X86_UP_APIC set in .config. On SMP kernels the APIC hardware is enabled, so the SMP kernel should use the performance monitoring hardware without any special settings. Most UP machines can run the SMP kernels. So if you have a SMP kernel around you could use that. -Will

Re: oprofile not detecting P4 Xeon 800Mhz FSB processor From: Shourya P. Bhattacharya <sbhattacharya@gm...> - 2004-12-28 15:20 Attachments: 2681.config      Hi, > > OProfile uses the local APIC to generate the interrupts for the > performance monitoring hardware. If you have a uniprocessor kernel and > the local APIC is not enabled for UP, then OProfile is going to fall > back on the timer mechanism. If you are building a UP kernel check that > you have CONFIG_X86_UP_APIC set in .config. > > On SMP kernels the APIC hardware is enabled, so the SMP kernel should > use the performance monitoring hardware without any special settings. > Most UP machines can run the SMP kernels. So if you have a SMP kernel > around you could use that. > First of all thanks for your response. I actually tried with both UNI and SMP kernels. I did enable the local APIC option when compiling as UNI, and with SMP I enabled the RTC. I will tell you the steps which I took, First took a clean 2.6.8.1 source tree, compiled a monolithic kernel with the options present in file I am attaching alongwith (for UNI). Then rebooted and tried the following commands. $opcontrol --init $op_help which showed "Using timer interrupt". The same kernel on a P4 (non-HT) 1.5 Ghz detected the cpu correctly and showed the list of valid events. The current system is a dual Xeon P4(HT) 3200Mhz with 800Mhz FSB. Firstly, Is there any chance that the processor does not have performance counters? I tried searching on the net, though I have not fully sure but it looks like all P4 Xeon's do have performance counters. Secondly, is there any chance that some BIOS setting may be causing oprofile not to detect the cpu properly? and in that case what should I be searching for? I have already tried enabling /disabling HT. Since /proc/cpuinfo is showing the right info, I think even BIOS should not be a problem. It will be of great help if you could suggest some solution. Thank you once again for your time. -Shourya > -Will > >

Re: oprofile not detecting P4 Xeon 800Mhz FSB processor From: William Cohen <wcohen@nc...> - 2004-12-28 15:53 Shourya P. Bhattacharya wrote: > Hi, > >>OProfile uses the local APIC to generate the interrupts for the >>performance monitoring hardware. If you have a uniprocessor kernel and >>the local APIC is not enabled for UP, then OProfile is going to fall >>back on the timer mechanism. If you are building a UP kernel check that >>you have CONFIG_X86_UP_APIC set in .config. >> >>On SMP kernels the APIC hardware is enabled, so the SMP kernel should >>use the performance monitoring hardware without any special settings. >>Most UP machines can run the SMP kernels. So if you have a SMP kernel >>around you could use that. >> > > First of all thanks for your response. I actually tried with both UNI > and SMP kernels. I did enable the local APIC option when compiling > as UNI, and with SMP I enabled the RTC. I will tell you the steps > which I took, > First took a clean 2.6.8.1 source tree, compiled a monolithic kernel > with the options present in file I am attaching alongwith (for > UNI). Then rebooted and tried the following commands. > $opcontrol --init > $op_help > which showed "Using timer interrupt". The same kernel on a P4 (non-HT) > 1.5 Ghz detected the cpu correctly and showed the list of valid > events. The current system is a dual Xeon P4(HT) 3200Mhz with 800Mhz > FSB. Firstly, Is there any chance that the processor does not have > performance counters? I tried searching on the net, though I have not > fully sure but it looks like all P4 Xeon's do have performance > counters. Secondly, is there any chance that some BIOS setting may be > causing oprofile not to detect the cpu properly? and in that case what > should I be searching for? I have already tried enabling /disabling > HT. Since /proc/cpuinfo is showing the right info, I think even BIOS > should not be a problem. > It will be of great help if you could suggest some solution. > Thank you once again for your time. > > -Shourya I don't know of any P4 processors that omit the performance monitoring hardware. The method used by oprofile doesn't actually look for the performance monitoring hardware; it looks at the processor model number and APIC hardware. These are the key things for using performance monitoring hardware. Without them OProfile has to fall back on the timer mechanism. I see the problem now. Your /proc/cpuinfo has: model : 4 The oprofile linux kernel code for the check in arch/i386/oprofile/nmi_int.c:p4_init() has the following: if (cpu_model > 3) return 0; So processors with model 3 or 4 are going to use timer interrupt mechanisms rather than the performance monitoring hardware. Have to look at Intels updated cpu id documentation to find the correct range for this. Looking at http://www.intel.com/design/xeon/applnots/241618.htm, it only lists up to model 3. It looks like there documentation is lagging some. There isn't a model 4 listed on it. -Will

Re: oprofile not detecting P4 Xeon 800Mhz FSB processor From: Shourya P. Bhattacharya <sbhattacharya@gm...> - 2004-12-29 05:29 Hi, > I don't know of any P4 processors that omit the performance monitoring > hardware. The method used by oprofile doesn't actually look for the > performance monitoring hardware; it looks at the processor model number > and APIC hardware. These are the key things for using performance > monitoring hardware. Without them OProfile has to fall back on the timer > mechanism. > > I see the problem now. Your /proc/cpuinfo has: > > model : 4 > > The oprofile linux kernel code for the check in > arch/i386/oprofile/nmi_int.c:p4_init() has the following: > > if (cpu_model > 3) > return 0; > > So processors with model 3 or 4 are going to use timer interrupt > mechanisms rather than the performance monitoring hardware. Have to look > at Intels updated cpu id documentation to find the correct range for > this. Looking at http://www.intel.com/design/xeon/applnots/241618.htm, > it only lists up to model 3. It looks like there documentation is > lagging some. There isn't a model 4 listed on it. As a temporary workaround I simply changed the "if" condition to "if(cpu_model > 4)" and tried my luck. So far things are working fine. I will report in case of any abnormal behaviour. -Shourya

Re: oprofile not detecting P4 Xeon 800Mhz FSB processor From: Philippe Elie <phil.el@wa...> - 2004-12-29 13:03 On Tue, 28 Dec 2004 at 10:53 +0000, William Cohen wrote: > I don't know of any P4 processors that omit the performance monitoring > hardware. The method used by oprofile doesn't actually look for the > performance monitoring hardware; it looks at the processor model number > and APIC hardware. These are the key things for using performance > monitoring hardware. Without them OProfile has to fall back on the timer > mechanism. > > I see the problem now. Your /proc/cpuinfo has: > > model : 4 > > The oprofile linux kernel code for the check in > arch/i386/oprofile/nmi_int.c:p4_init() has the following: > > if (cpu_model > 3) > return 0; > > So processors with model 3 or 4 are going to use timer interrupt > mechanisms rather than the performance monitoring hardware. Have to look > at Intels updated cpu id documentation to find the correct range for > this. Looking at http://www.intel.com/design/xeon/applnots/241618.htm, > it only lists up to model 3. It looks like there documentation is > lagging some. There isn't a model 4 listed on it. > see http://www.intel.com/design/pentium4/specupdt/302352.htm, doc update show a bunch of P4 with cpu id = 0F41h, look like all core stepping E0 are model 4. The same doc says than P4 90 nm can be identified by family 0F model 3 but the next table disagree with that. Philippe Elie

Hello!

Trying to setup oprofile on this kernel.  It works with 2.6.10-rc3. This 
is under Debian unstable.

warren:~# opcontrol -v
opcontrol: oprofile 0.8.1 compiled on Oct 18 2004 16:55:36

Any operation, such as

warren:~# opcontrol --no-vmlinux --setup

result into this:

--------------------
  <1>Unable to handle kernel NULL pointer dereference at virtual address 
00000000
  printing eip:
d8a5fd65
*pde = 00000000
Oops: 0000 [#14]
PREEMPT
Modules linked in: oprofile i2c_viapro i2c_core psmouse floppy pcspkr 
evdev videodev uhci_hcd usbcore parport_pc parport dm_mod rtc eth1394 unix
CPU:    0
EIP:    0060:[<d8a5fd65>]    Not tainted VLI
EFLAGS: 00010292   (2.6.10-rc3-mm1-20041214)
EIP is at oprofilefs_str_to_user+0x15/0x50 [oprofile]
eax: 00000000   ebx: 00000000   ecx: ffffffff   edx: 00000000
esi: cf55c658   edi: 00000000   ebp: cf523f64   esp: cf523f4c
ds: 007b   es: 007b   ss: 0068
Process op_help (pid: 2960, threadinfo=cf522000 task=d7261870)
Stack: 00000000 00000000 000b7fe8 00000000 cf55c658 cf523fa8 cf523f90 
c017fe8f
        00000000 b7fe8000 00001000 cf523fa8 d72ad068 d72ad094 cf55c658 
fffffff7
        00000000 cf523fbc c01801cb cf55c658 b7fe8000 00001000 cf523fa8 
00000000
Call Trace:
  [<c0104c4f>] show_stack+0x7f/0xa0
  [<c0104df6>] show_registers+0x156/0x1c0
  [<c0105076>] die+0x156/0x2d0
  [<c011a6d6>] do_page_fault+0x3b6/0x714
  [<c010476f>] error_code+0x2b/0x30
  [<c017fe8f>] vfs_read+0xcf/0x170
  [<c01801cb>] sys_read+0x4b/0x80
  [<c0103cc7>] syscall_call+0x7/0xb
Code: 89 53 50 89 43 44 89 53 48 89 d8 8b 5d fc 89 ec 5d c3 8d 74 26 00 
55 31 c0 89 e5 83 ec 18 b9 ff ff ff ff 8b 55 08 89 7d fc 89 d7 <f2> ae 
f7 d1 49 89 4c 24 10 8b 45 14 89 54 24 0c 89 44 24 08 8b
--------------------

bye,

Ricky Ng-Adam

I'm not sure what is triggering this issue, but it's come up
several times now.  After collecting data and using opreport
for some time - and getting good reports - I start seeing the
following error:

# opreport
opreport error: /var/lib/oprofile/samples/current/{kern}/af_packet/{dep}/{kern}/af_packet/CPU_CYCLES.100000.0.all.all.all: Invalid argument
#

Anyone seen this before?  Is there a fix?

I'm running the 0.8.1 bundle on an IA64 box with a CPU metric kernel
patch + a perfmond patch to allow CPU metrics to be collected.  The
kernel is a SLES-9-SP-1-Beta3 setup - in case that matters.

Any thoughts would be appreciated - fixes even more!  :)

thanks,

brent

Re: opreport error message From: John Levon <levon@mo...> - 2004-12-07 13:57 On Mon, Dec 06, 2004 at 01:14:08PM -0600, Brent Henderson wrote: > opreport error: > /var/lib/oprofile/samples/current/{kern}/af_packet/{dep}/{kern}/af_packet/CPU_CYCLES.100000.0.all.all.all: Invalid argument > # > > Anyone seen this before? Is there a fix? truss opreport to see where EINVAL is coming from > I'm running the 0.8.1 bundle on an IA64 box with a CPU metric kernel > patch + a perfmond patch to allow CPU metrics to be collected. The > kernel is a SLES-9-SP-1-Beta3 setup - in case that matters. You really shouldn't be reporting oprofile bugs without using a standard Linus kernel, but it probably doesn't matter here... What exactly are your "CPU metric" patches?? john

Re: opreport error message From: Andy Fleming <afleming@fr...> - 2004-12-07 18:07 I've seen that before. It looks like the old opreport files are corrupt, or ...something. Anyway, I found that opcontrol --reset fixes the problem. On Dec 6, 2004, at 13:14, Brent Henderson wrote: > I'm not sure what is triggering this issue, but it's come up > several times now. After collecting data and using opreport > for some time - and getting good reports - I start seeing the > following error: > > # opreport > opreport error: > /var/lib/oprofile/samples/current/{kern}/af_packet/{dep}/{kern}/ > af_packet/CPU_CYCLES.100000.0.all.all.all: Invalid argument > # > > Anyone seen this before? Is there a fix? > > I'm running the 0.8.1 bundle on an IA64 box with a CPU metric kernel > patch + a perfmond patch to allow CPU metrics to be collected. The > kernel is a SLES-9-SP-1-Beta3 setup - in case that matters. > > Any thoughts would be appreciated - fixes even more! :) > > thanks, > > brent > > > ------------------------------------------------------- > SF email is sponsored by - The IT Product Guide > Read honest & candid reviews on hundreds of IT Products from real > users. > Discover which products truly live up to the hype. Start reading now. > http://productguide.itmanagersjournal.com/ > _______________________________________________ > oprofile-list mailing list > oprofile-list@... > https://lists.sourceforge.net/lists/listinfo/oprofile-list > Andy Fleming

Re: opreport error message From: <alext@fc...> - 2004-12-07 18:17 On Tue, Dec 07, 2004 at 12:03:58PM -0600, Andy Fleming wrote: > I've seen that before. It looks like the old opreport files are > corrupt, or ...something. Anyway, I found that opcontrol --reset fixes > the problem. Yes, but 'opcontrol --reset' just erases your old profile data files. In this case, the error seems to be repeating with new files gathered. Thanks, Alex

Re: opreport error message From: Brent Henderson <brent@rs...> - 2004-12-08 17:23 John Levon wrote: > On Mon, Dec 06, 2004 at 01:14:08PM -0600, Brent Henderson wrote: > > >>opreport error: >>/var/lib/oprofile/samples/current/{kern}/af_packet/{dep}/{kern}/af_packet/CPU_CYCLES.100000.0.all.all.all: Invalid argument >># >> >>Anyone seen this before? Is there a fix? > > > truss opreport to see where EINVAL is coming from > > The whole log is huge, hopefully this will be enough info - if I read this right, close(4) is returning 0, then the 'opreport error' came out before any other calls were logged. . . . open("/usr/local/share/oprofile//ia64/itanium2/unit_masks", O_RDONLY) = 4 fstat(4, {st_mode=S_IFREG|0644, st_size=9144, ...}) = 0 mmap(NULL, 65536, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x2000000000700000 read(4, "# IA-64 Itanium 2 possible unit "..., 16384) = 9144 read(4, "", 16384) = 0 close(4) = 0 munmap(0x2000000000700000, 65536) = 0 open("/usr/local/share/oprofile//ia64/itanium2/events", O_RDONLY) = 4 fstat(4, {st_mode=S_IFREG|0644, st_size=24197, ...}) = 0 mmap(NULL, 65536, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x2000000000700000 read(4, "# IA-64 Itanium 2 events\n\n# IA64"..., 16384) = 16384 read(4, "ne is available\nevent:0xb8 count"..., 16384) = 7813 read(4, "", 16384) = 0 close(4) = 0 munmap(0x2000000000700000, 65536) = 0 open("/var/lib/oprofile/samples/current/{kern}/no-vmlinux/{dep}/{kern}/no-vmlinux/CPU_CYCLES.10000.0.all.all.all", O_RDONLY) = 4 fstat(4, {st_mode=S_IFREG|0644, st_size=82024, ...}) = 0 mmap(NULL, 82024, PROT_READ, MAP_SHARED, 4, 0) = 0x2000000000700000 munmap(0x2000000000700000, 82024) = 0 close(4) = 0 open("/var/lib/oprofile/samples/current/{kern}/af_packet/{dep}/{kern}/af_packet/CPU_CYCLES.100000.0.all.all.all", O_RDONLY) = 4 fstat(4, {st_mode=S_IFREG|0644, st_size=4096, ...}) = 0 mmap(NULL, 4084, PROT_READ, MAP_SHARED, 4, 0) = 0x2000000000700000 munmap(0x2000000000700000, 4084) = 0 close(4) = 0 write(2, "o", 1o) = 1 write(2, "p", 1p) = 1 write(2, "r", 1r) = 1 write(2, "e", 1e) = 1 write(2, "p", 1p) = 1 write(2, "o", 1o) = 1 write(2, "r", 1r) = 1 write(2, "t", 1t) = 1 write(2, " ", 1 ) = 1 write(2, "e", 1e) = 1 write(2, "r", 1r) = 1 write(2, "r", 1r) = 1 write(2, "o", 1o) = 1 write(2, "r", 1r) = 1 write(2, ":", 1:) = 1 . . . I set a breakpoint at exit of opreport, but it claimed the stack was hosed, so it is probably not useful, but here it is: Breakpoint 2, 0x2000000000441090 in exit () from /lib/tls/libc.so.6.1 (gdb) bt #0 0x2000000000441090 in exit () from /lib/tls/libc.so.6.1 #1 0x2000000000415830 in __libc_start_main () from /lib/tls/libc.so.6.1 #2 0x0000000000000000 in ?? () Previous frame identical to this frame (corrupt stack?) (gdb) Seems like the code should be way past __libc_start_main at this point. >>I'm running the 0.8.1 bundle on an IA64 box with a CPU metric kernel >>patch + a perfmond patch to allow CPU metrics to be collected. The >>kernel is a SLES-9-SP-1-Beta3 setup - in case that matters. > > > You really shouldn't be reporting oprofile bugs without using a standard > Linus kernel, but it probably doesn't matter here... > I've rebuild a new system with the SLES9 bits, and the issue still shows up. (I just wanted to rule out the Beta OS I was using.) > What exactly are your "CPU metric" patches?? > > john > The kernel patch was for enabling metrics on ia64 processors. Here's a link: http://sourceforge.net/mailarchive/message.php?msg_id=9286819 The oprofiled update was for a problem with sched_setaffinity - I'm using a very recent copy of opd_perfmon.c that addresses this issue. All the other code is stock. thanks, brent

Re: opreport error message From: John Levon <levon@mo...> - 2004-12-08 19:19 On Wed, Dec 08, 2004 at 11:21:36AM -0600, Brent Henderson wrote: > The whole log is huge, hopefully this will be enough info - if I read > this right, close(4) is returning 0, then the 'opreport error' came out > before any other calls were logged. The EINVAL is being set internally. If you want to investigate, grep oprofile sources for EINVAL and add printf()s. You should then be able to see what path is causing the EINVAL and can proceed from there. My suspect would be libdb/ john

Re: opreport error message From: Brent Henderson <brent@rs...> - 2004-12-13 19:09 Hmm, more digging shows that is some memory corruption - possibly caused by popt or stl code. Hopefully I'm not 'off by one' in my debugging sequence. This might also explain why I'm still in __libc_start_main - static initializers - right? (The report was generated using an old copy of libdbmalloc I had lying around.) SLES9 has libpopt v 1.7 which appears to be the latest version although there appears to be little activity since 20-Sep-2002. Maybe I have the wrong source link though, I'm looking at: http://freshmeat.net/projects/popt/ brent shepton:/home/brent/oprofile-0.8.1/pp # ./opreport MALLOC Warning from free(): Data has overrun beyond requested number of bytes This error is *probably* associated with the following allocation: A call to malloc for 192 bytes in an unknown file. This was the 11th call to malloc. MALLOC Warning from free(): Data has overrun beyond requested number of bytes This error is *probably* associated with the following allocation: A call to malloc for 384 bytes in an unknown file. This was the 14th call to malloc. MALLOC Warning from free(): Data has overrun beyond requested number of bytes This error is *probably* associated with the following allocation: A call to malloc for 768 bytes in an unknown file. This was the 19th call to malloc. Segmentation fault shepton:/home/brent/oprofile-0.8.1/pp # gdb ./opreport GNU gdb 6.1 Copyright 2004 Free Software Foundation, Inc. GDB is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. Type "show copying" to see the conditions. There is absolutely no warranty for GDB. Type "show warranty" for details. This GDB was configured as "ia64-suse-linux"...Using host libthread_db library "/lib/tls/libthread_db.so.1". (gdb) b malloc Breakpoint 1 at 0x4000000000032911: file malloc.c, line 99. (gdb) c The program is not being run. (gdb) R Starting program: /home/brent/oprofile-0.8.1/pp/opreport Breakpoint 1, malloc (size=24) at malloc.c:99 99 return( debug_malloc(NULL,-1,size) ); Current language: auto; currently c (gdb) b debug_malloc Breakpoint 2 at 0x4000000000032970: file malloc.c, line 126. (gdb) d 1 (gdb) c Continuing. Breakpoint 2, debug_malloc (file=0x0, line=-1, size=24) at malloc.c:126 126 call_counter++; (gdb) d 2 (gdb) l 121 static IDTYPE call_counter; 122 123 /* 124 * increment the counter for the number of calls to this func. 125 */ 126 call_counter++; 127 128 return( DBFmalloc("malloc",M_T_MALLOC,call_counter,file,line,size) ); 129 130 } /* debug_malloc(... */ (gdb) b 128 Breakpoint 3 at 0x4000000000032982: file malloc.c, line 128. (gdb) condition Display all 13102 possibilities? (y or n) (gdb) help condition Specify breakpoint number N to break only if COND is true. Usage is `condition N COND', where N is an integer and COND is an expression to be evaluated whenever breakpoint N is reached. (gdb) condition 3 call_counter==11 (gdb) c Continuing. Breakpoint 3, debug_malloc (file=0x0, line=-1, size=192) at malloc.c:128 128 return( DBFmalloc("malloc",M_T_MALLOC,call_counter,file,line,size) ); (gdb) print call_counter $1 = 11 (gdb) bt #0 debug_malloc (file=0x0, line=-1, size=192) at malloc.c:128 #1 0x4000000000032930 in malloc (size=192) at malloc.c:99 #2 0x20000000002b6bf0 in operator new () from /usr/lib/libstdc++.so.5 #3 0x2000000000290650 in std::__default_alloc_template<true, 0>::allocate () from /usr/lib/libstdc++.so.5 #4 0x40000000000e17c0 in std::vector<poptOption, std::allocator<poptOption> >::_M_insert_aux (this=0x60000000000057e0, __position= {<std::iterator<std::random_access_iterator_tag, poptOption, long, poptOption*, poptOption&>> = {<No data fields>}, _M_current = 0x6000000000011150}, __x=@0x60000fffffffafd0) at vector.tcc:223 #5 0x40000000000dfba0 in option_base (this=0x6000000000012190, name=0x4000000000125038 "threshold", short_name=116 't', help=0x4000000000125048 "minimum percentage needed to produce output", arg_help=0x4000000000125078 "percent", data=0x60000000000121a0, popt_flags=1) at popt_options.cpp:246 #6 0x40000000000dff20 in option_imp (this=0x6000000000012190, val=@0x60000000000051b8, name=0x4000000000125038 "threshold", short_name=116 't', help=0x4000000000125048 "minimum percentage needed to produce output", arg_help=0x4000000000125078 "percent") at popt_options.cpp:288 #7 0x40000000000dffd0 in option<std::string> (this=0x6000000000005660, value=@0x60000000000051b8, name=0x4000000000125038 "threshold", short_name=116 't', ---Type <return> to continue, or q <return> to quit--- help=0x4000000000125048 "minimum percentage needed to produce output", arg_help=0x4000000000125078 "percent") at popt_options.cpp:229 #8 0x400000000002d780 in global constructors keyed to _ZN34_GLOBAL__N_common_option.cppgk6poM9thresholdE () at common_option.cpp:44 #9 0x4000000000124670 in __do_global_ctors_aux () #10 0x40000000001245e0 in __libc_csu_init () at elf-init.c:67 #11 0x2000000000415710 in __libc_start_main () from /lib/tls/libc.so.6.1 #12 0x0000000000000000 in ?? () Previous frame identical to this frame (corrupt stack?) (gdb) condition 3 call_counter==14 (gdb) c Continuing. Breakpoint 3, debug_malloc (file=0x0, line=-1, size=384) at malloc.c:128 128 return( DBFmalloc("malloc",M_T_MALLOC,call_counter,file,line,size) ); (gdb) bt #0 debug_malloc (file=0x0, line=-1, size=384) at malloc.c:128 #1 0x4000000000032930 in malloc (size=384) at malloc.c:99 #2 0x20000000002b6bf0 in operator new () from /usr/lib/libstdc++.so.5 #3 0x2000000000290650 in std::__default_alloc_template<true, 0>::allocate () from /usr/lib/libstdc++.so.5 #4 0x40000000000e17c0 in std::vector<poptOption, std::allocator<poptOption> >::_M_insert_aux (this=0x60000000000057e0, __position= {<std::iterator<std::random_access_iterator_tag, poptOption, long, poptOption*, poptOption&>> = {<No data fields>}, _M_current = 0x6000000000012300}, __x=@0x60000fffffffafc0) at vector.tcc:223 #5 0x40000000000dfba0 in option_base (this=0x6000000000012440, name=0x40000000001247b0 "output-file", short_name=111 'o', help=0x40000000001247c0 "output to the given filename", arg_help=0x40000000001247e0 "file", data=0x6000000000012450, popt_flags=1) at popt_options.cpp:246 #6 0x40000000000dff20 in option_imp (this=0x6000000000012440, val=@0x6000000000005188, name=0x40000000001247b0 "output-file", short_name=111 'o', help=0x40000000001247c0 "output to the given filename", arg_help=0x40000000001247e0 "file") at popt_options.cpp:288 #7 0x40000000000dffd0 in option<std::string> (this=0x6000000000005300, value=@0x6000000000005188, name=0x40000000001247b0 "output-file", short_name=111 'o', ---Type <return> to continue, or q <return> to quit--- help=0x40000000001247c0 "output to the given filename", arg_help=0x40000000001247e0 "file") at popt_options.cpp:229 #8 0x4000000000020f30 in global constructors keyed to _ZN37_GLOBAL__N_opreport_options.cppugaq4D7outfileE () at opreport_options.cpp:95 #9 0x4000000000124670 in __do_global_ctors_aux () #10 0x40000000001245e0 in __libc_csu_init () at elf-init.c:67 #11 0x2000000000415710 in __libc_start_main () from /lib/tls/libc.so.6.1 #12 0x0000000000000000 in ?? () Previous frame identical to this frame (corrupt stack?) (gdb) condition 3 call_counter==19 (gdb) c Continuing. MALLOC Warning from free(): Data has overrun beyond requested number of bytes This error is *probably* associated with the following allocation: A call to malloc for 192 bytes in an unknown file. This was the 11th call to malloc. Breakpoint 3, debug_malloc (file=0x0, line=-1, size=768) at malloc.c:128 128 return( DBFmalloc("malloc",M_T_MALLOC,call_counter,file,line,size) ); (gdb) bt #0 debug_malloc (file=0x0, line=-1, size=768) at malloc.c:128 #1 0x4000000000032930 in malloc (size=768) at malloc.c:99 #2 0x20000000002b6bf0 in operator new () from /usr/lib/libstdc++.so.5 #3 0x2000000000290650 in std::__default_alloc_template<true, 0>::allocate () from /usr/lib/libstdc++.so.5 #4 0x40000000000e17c0 in std::vector<poptOption, std::allocator<poptOption> >::_M_insert_aux (this=0x60000000000057e0, __position= {<std::iterator<std::random_access_iterator_tag, poptOption, long, poptOption*, poptOption&>> = {<No data fields>}, _M_current = 0x6000000000012670}, __x=@0x60000fffffffafc0) at vector.tcc:223 #5 0x40000000000dfba0 in option_base (this=0x6000000000012860, name=0x4000000000124870 "exclude-dependent", short_name=120 'x', help=0x4000000000124888 "exclude libs, kernel, and module samples for applications", arg_help=0x0, data=0x6000000000012878, popt_flags=0) at popt_options.cpp:246 #6 0x40000000000e0420 in option_imp (this=0x6000000000012860, val=@0x6000000000005199, name=0x4000000000124870 "exclude-dependent", short_name=120 'x', help=0x4000000000124888 "exclude libs, kernel, and module samples for applications") at popt_options.cpp:261 #7 0x40000000000e04d0 in option (this=0x6000000000005320, value=@0x6000000000005199, name=0x4000000000124870 "exclude-dependent", short_name=120 'x', ---Type <return> to continue, or q <return> to quit--- help=0x4000000000124888 "exclude libs, kernel, and module samples for applications") at popt_options.cpp:209 #8 0x4000000000021040 in global constructors keyed to _ZN37_GLOBAL__N_opreport_options.cppugaq4D7outfileE () at opreport_options.cpp:95 #9 0x4000000000124670 in __do_global_ctors_aux () #10 0x40000000001245e0 in __libc_csu_init () at elf-init.c:67 #11 0x2000000000415710 in __libc_start_main () from /lib/tls/libc.so.6.1 #12 0x0000000000000000 in ?? () Previous frame identical to this frame (corrupt stack?) (gdb)

--- oprofile-0.8.1.orig/utils/op_help.c	Sat May 29 19:03:03 2004
+++ oprofile-0.8.1/utils/op_help.c	Mon Dec 13 00:27:23 2004
@@ -403,6 +403,30 @@
 		printf("See Intel XScale Core Developer's Manual\n"
 		       "Chapter 8 Performance Monitoring\n");
 		break;
+
+	case CPU_MIPS_R10000:
+	case CPU_MIPS_R12000:
+		printf("See NEC R10000 / R12000 User's Manual\n"
+		       "http://www.necelam.com/docs/files/U10278EJ3V0UM00.pdf\n");
+		break;
+	case CPU_MIPS_RM7000:
+		printf("See RM7000 Family User Manual "
+		       "available from http://www.pmc-sierra.com\n");
+		break;
+	case CPU_MIPS_RM9000:
+		printf("See RM9000x2 Family User Manual "
+		       "available from http://www.pmc-sierra.com\n");
+		break;
+	case CPU_MIPS_SB1:
+	case CPU_MIPS_VR5432:
+		printf("See NEC VR5443 User's Manual, Volume 1\n"
+		       "http://www.necelam.com/docs/files/1375_V1.pdf\n");
+		break;
+	case CPU_MIPS_VR5500:
+		printf("See NEC R10000 / R12000 User's Manual\n"
+		     "http://www.necel.com/nesdis/image/U16677EJ3V0UM00.pdf\n");
+		break;
+
 	case CPU_RTC:
 		break;

Below are patches that add MIPS support for the performance counters in
the MIPS R10000, R12000, QED RM7000, PMC-Sierra RM9000, Sibyte SB1,
NEC VR5432, VR5500, VR5532 and VR7701 to the userspace part of oprofile.

Please merge, thanks.

  Ralf

 ChangeLog                     |   27 +++++
 README                        |    2 
 daemon/opd_cookie.c           |   11 ++
 events/Makefile.am            |   11 +-
 events/Makefile.in            |    9 +
 events/mips/r10000/events     |   36 +++++++
 events/mips/r10000/unit_masks |    5 +
 events/mips/r12000/events     |   35 +++++++
 events/mips/r12000/unit_masks |    7 +
 events/mips/rm7000/events     |   34 +++++++
 events/mips/rm7000/unit_masks |    5 +
 events/mips/rm9000/events     |   32 +++++++
 events/mips/rm9000/unit_masks |    5 +
 events/mips/sb1/events        |   73 +++++++++++++++
 events/mips/sb1/unit_masks    |    5 +
 events/mips/vr5432/events     |   14 +++
 events/mips/vr5432/unit_masks |    5 +
 events/mips/vr5500/events     |   16 +++
 events/mips/vr5500/unit_masks |    5 +
 libop/op_cpu_type.c           |    7 +
 libop/op_cpu_type.h           |    7 +
 libop/op_events.c             |   19 ++++
 23 files changed, 367 insertions(+), 195 deletions(-)

diff -purN oprofile-0.8.1.orig/ChangeLog oprofile-0.8.1/ChangeLog
--- oprofile-0.8.1.orig/ChangeLog	Sat Sep 11 20:33:26 2004
+++ oprofile-0.8.1/ChangeLog	Sun Dec 12 16:58:49 2004
@@ -1,3 +1,30 @@
+2004-12-12  Ralf Baechle <ralf@...>
+
+	* daemon/opd_cookie.c: Define __NR_lookup_dcookie for all three ABIs
+	supported ABIs.
+	* events/mips/r10000/events: New file
+	* events/mips/r10000/unit_masks: New file
+	* events/mips/r12000/events: New file
+	* events/mips/r12000/unit_masks: New file
+	* events/mips/r5432/events: New file
+	* events/mips/r5432/unit_masks: New file
+	* events/mips/r5500/events: New file
+	* events/mips/r5500/unit_masks: New file
+	* events/mips/rm7000/events: New file
+	* events/mips/rm7000/unit_masks: New file
+	* events/mips/rm9000/events: New file
+	* events/mips/rm9000/unit_masks: New file
+	* events/mips/sb1/events: New file
+	* events/mips/sb1/unit_masks: New file
+	* events/Makefile.am: List new files for MIPS in event_files.
+	* events/Makefile.in: Rebuild.
+	* events/Makefile: Likewise.
+	* libop/op_cpu_type.c: Add entries for the MIPS R10000 and R12000,
+	QED RM7000 and PMC-Sierra RM9000, NEC VR5432 and VR5500 and Sibyte SB1
+	processors.
+	* libop/op_cpu_type.h: Add entries for the same processors to op_cpu.
+	* README: List myself as the caretaker of the MIPS bits.
+
 2004-09-11  John Levon  <levon@...>
 
 	* configure.in: bump to 0.8.1
diff -purN oprofile-0.8.1.orig/README oprofile-0.8.1/README
--- oprofile-0.8.1.orig/README	Mon Sep 15 03:20:30 2003
+++ oprofile-0.8.1/README	Sun Dec 12 16:58:21 2004
@@ -19,6 +19,8 @@ oprofile release, and contributed bug fi
 
 Graydon Hoare <graydon@...> provided P4 port, bug fixes and cleanups.
 
+Ralf Baechle <ralf@...> provided the MIPS port.
+
 Other contributors are listed in the ChangeLog.
 
 Building
diff -purN oprofile-0.8.1.orig/daemon/opd_cookie.c oprofile-0.8.1/daemon/opd_cookie.c
--- oprofile-0.8.1.orig/daemon/opd_cookie.c	Sat May 29 19:02:59 2004
+++ oprofile-0.8.1/daemon/opd_cookie.c	Fri Dec 10 14:48:08 2004
@@ -40,6 +40,17 @@
 #define __NR_lookup_dcookie 110
 #elif defined(__arm__)
 #define __NR_lookup_dcookie (__NR_SYSCALL_BASE+249)
+#elif defined(__mips__)
+#include <sgidefs.h>
+#if _MIPS_SIM == _MIPS_SIM_ABI32
+#define __NR_lookup_dcookie 4247		/* O32 */
+#elif _MIPS_SIM == _MIPS_SIM_ABI64
+#define __NR_lookup_dcookie 5206		/* N64 */
+#elif _MIPS_SIM == _MIPS_SIM_NABI32
+#define __NR_lookup_dcookie 6206		/* N32 */
+#else
+#error Unknown MIPS ABI: Dunno __NR_lookup_dcookie
+#endif
 #else
 #error Please define __NR_lookup_dcookie for your architecture
 #endif
diff -purN oprofile-0.8.1.orig/events/Makefile.am oprofile-0.8.1/events/Makefile.am
--- oprofile-0.8.1.orig/events/Makefile.am	Wed Apr  7 17:01:03 2004
+++ oprofile-0.8.1/events/Makefile.am	Sun Dec 12 22:21:07 2004
@@ -1,4 +1,4 @@
-event_files = \
+Event_files = \
 	alpha/ev4/events alpha/ev4/unit_masks \
 	alpha/ev5/events alpha/ev5/unit_masks \
 	alpha/ev67/events alpha/ev67/unit_masks \
@@ -17,7 +17,14 @@ event_files = \
 	rtc/events rtc/unit_masks \
 	x86-64/hammer/events x86-64/hammer/unit_masks \
 	arm/xscale1/events arm/xscale1/unit_masks \
-	arm/xscale2/events arm/xscale2/unit_masks
+	arm/xscale2/events arm/xscale2/unit_masks \
+	mips/rm7000/events mips/rm7000/unit_masks \
+	mips/rm9000/events mips/rm9000/unit_masks \
+	mips/sb1/events mips/sb1/unit_masks \
+	mips/r10000/events mips/r10000/unit_masks \
+	mips/r12000/events mips/r12000/unit_masks \
+	mips/vr5432/events mips/vr5432/unit_masks \
+	mips/vr5500/events mips/vr5500/unit_masks
 
 install-data-local:
 	for i in ${event_files} ; do \
diff -purN oprofile-0.8.1.orig/events/Makefile.in oprofile-0.8.1/events/Makefile.in
--- oprofile-0.8.1.orig/events/Makefile.in	Sat Sep 11 19:45:43 2004
+++ oprofile-0.8.1/events/Makefile.in	Sun Dec 12 22:21:17 2004
@@ -131,7 +131,14 @@ event_files = \
 	rtc/events rtc/unit_masks \
 	x86-64/hammer/events x86-64/hammer/unit_masks \
 	arm/xscale1/events arm/xscale1/unit_masks \
-	arm/xscale2/events arm/xscale2/unit_masks
+	arm/xscale2/events arm/xscale2/unit_masks \
+	mips/rm7000/events mips/rm7000/unit_masks \
+	mips/rm9000/events mips/rm9000/unit_masks \
+	mips/sb1/events mips/sb1/unit_masks \
+	mips/r10000/events mips/r10000/unit_masks \
+	mips/r12000/events mips/r12000/unit_masks \
+	mips/vr5432/events mips/vr5432/unit_masks \
+	mips/vr5500/events mips/vr5500/unit_masks
 
 
 EXTRA_DIST = $(event_files)
diff -purN oprofile-0.8.1.orig/events/mips/r10000/events oprofile-0.8.1/events/mips/r10000/events
--- oprofile-0.8.1.orig/events/mips/r10000/events	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/r10000/events	Sun Dec 12 14:46:30 2004
@@ -0,0 +1,36 @@
+#
+# R10000 events
+#
+# The same event numbers mean different things on the two counters
+#
+event:0x00 counters:0,1 um:zero minimum:500 name:CYCLES : Cycles
+event:0x01 counters:0 um:zero minimum:500 name:INSTRUCTIONS_ISSUED : Instructions issued
+event:0x01 counters:1 um:zero minimum:500 name:INSTRUCTIONS_GRADUATED : Instructions graduated
+event:0x02 counters:0 um:zero minimum:500 name:LOAD_PREFETC_SYNC_CACHEOP_ISSUED : Load / prefetch / sync / CacheOp issued
+event:0x02 counters:1 um:zero minimum:500 name:LOAD_PREFETC_SYNC_CACHEOP_GRADUATED : Load / prefetch / sync / CacheOp graduated
+event:0x03 counters:0 um:zero minimum:500 name:STORES_ISSUED : Stores issued
+event:0x03 counters:1 um:zero minimum:500 name:STORES_GRADUATED : Stores graduated
+event:0x04 counters:0 um:zero minimum:500 name:STORE_COND_ISSUED : Store conditional issued
+event:0x04 counters:1 um:zero minimum:500 name:STORE_COND_GRADUATED : Store conditional graduated
+event:0x05 counters:0 um:zero minimum:500 name:FAILED_STORE_CONDITIONAL : Failed store conditional
+event:0x05 counters:1 um:zero minimum:500 name:FP_INSTRUCTON_GRADUATED : Floating-point instructions graduated
+event:0x06 counters:0 um:zero minimum:500 name:BRANCHES_RESOLVED : Branches resolved
+event:0x06 counters:1 um:zero minimum:500 name:QUADWORDS_WB_FROM_PRIMARY_DCACHE : Quadwords written back from primary data cache
+event:0x07 counters:0 um:zero minimum:500 name:QUADWORDS_WB_FROM_SCACHE : Quadwords written back from secondary cache
+event:0x07 counters:1 um:zero minimum:500 name:TLB_REFILL_EXCEPTIONS : TLB refill exceptions
+event:0x08 counters:0 um:zero minimum:500 name:CORRECTABLE_ECC_ERRORS_SCACHE : Correctable ECC errors on secondary cache data
+event:0x08 counters:1 um:zero minimum:500 name:BRANCHES_MISPREDICTED : Branches mispredicted
+event:0x09 counters:0 um:zero minimum:500 name:INSTRUCTION_CACHE_MISSES : Instruction cache misses
+event:0x09 counters:1 um:zero minimum:500 name:SCACHE_LOAD_STORE_CACHEOP_OPERATIONS : Secondary cache load / store and cache-ops operations
+event:0x0a counters:0 um:zero minimum:500 name:SCACHE_MISSES_INSTRUCTION : Secondary cache misses (instruction)
+event:0x0a counters:1 um:zero minimum:500 name:SCACHE_MISSES_DATA : Secondary cache misses (data)
+event:0x0b counters:0 um:zero minimum:500 name:SCACHE_WAY_MISPREDICTED_INSN : Secondary cache way mispredicted (instruction)
+event:0x0b counters:1 um:zero minimum:500 name:SCACHE_WAY_MISPREDICTED_DATA : Secondary cache way mispredicted (data)
+event:0x0c counters:0 um:zero minimum:500 name:EXTERNAL_INTERVENTION_RQ : External intervention requests
+event:0x0c counters:1 um:zero minimum:500 name:EXTERNAL_INTERVENTION_RQ_HITS_SCACHE : External intervention request is determined to have hit in secondary cache
+event:0x0d counters:0 um:zero minimum:500 name:EXTERNAL_INVALIDATE_RQ : External invalidate requests
+event:0x0d counters:1 um:zero minimum:500 name:EXTERNAL_INVALIDATE_RQ_HITS_SCACHE : External invalidate request is determined to have hit in secondary cache
+event:0x0e counters:0 um:zero minimum:500 name:FUNCTIONAL_UNIT_COMPLETION_CYCLES : Functional unit completion cycles
+event:0x0e counters:1 um:zero minimum:500 name:STORES_OR_STORE_PREF_TO_CLEANEXCLUSIVE_SCACHE_BLOCKS : Stores or prefetches with store hint to CleanExclusive secondary cache blocks
+event:0x0f counters:0 um:zero minimum:500 name:INSTRUCTION_GRADUATED : Instructions graduated
+event:0x0f counters:1 um:zero minimum:500 name:STORES_OR_STORE_PREF_TO_SHD_SCACHE_BLOCKS : Stores or prefetches with store hint to Shared secondary cache blocks
diff -purN oprofile-0.8.1.orig/events/mips/r10000/unit_masks oprofile-0.8.1/events/mips/r10000/unit_masks
--- oprofile-0.8.1.orig/events/mips/r10000/unit_masks	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/r10000/unit_masks	Sun Dec 12 22:16:40 2004
@@ -0,0 +1,5 @@
+#
+# MIPS R10000 possible unit masks
+#
+name:zero type:mandatory default:0x0
+	0x0 No unit mask
diff -purN oprofile-0.8.1.orig/events/mips/r12000/events oprofile-0.8.1/events/mips/r12000/events
--- oprofile-0.8.1.orig/events/mips/r12000/events	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/r12000/events	Sun Dec 12 22:15:47 2004
@@ -0,0 +1,35 @@
+#
+# R12000 events
+#
+event:0 counters:0,1,2,3 um:zero minimum:500 name:CYCLES : Cycles
+event:1 counters:0,1,2,3 um:zero minimum:500 name:DECODED_INSTRUCTIONS : Decoded instructions
+event:2 counters:0,1,2,3 um:zero minimum:500 name:DECODED_LOADS : Decoded loads
+event:3 counters:0,1,2,3 um:zero minimum:500 name:DECODED_STORES : Decoded stores
+event:4 counters:0,1,2,3 um:zero minimum:500 name:MISS_TABLE_OCCUPANCY : Miss Handling Table Occupancy
+event:5 counters:0,1,2,3 um:zero minimum:500 name:FAILED_STORE_CONDITIONAL : Failed store conditional
+event:6 counters:0,1,2,3 um:zero minimum:500 name:RESOLVED_BRANCH_CONDITIONAL : Resolved conditional branches
+event:7 counters:0,1,2,3 um:zero minimum:500 name:QUADWORRDS_WRITEBACK_FROM_SC : Quadwords written back from secondary cache
+event:8 counters:0,1,2,3 um:zero minimum:500 name:CORRECTABLE_ECC_ERRORS : Correctable ECC errors on secondary cache data
+event:9 counters:0,1,2,3 um:zero minimum:500 name:ICACHE_MISSES : Instruction cache misses
+event:10 counters:0,1,2,3 um:zero minimum:500 nameINSTRUCTION_SECONDARY_CACHE_MISSES : Secondary cache misses (instruction)
+event:11 counters:0,1,2,3 um:zero minimum:500 name:SECONDARY_CACHE_WAY_MISSPREDICTED : Secondary cache way mispredicted (instruction)
+event:12 counters:0,1,2,3 um:zero minimum:500 name:INTERVENTION_REQUESTS : External intervention requests
+event:13 counters:0,1,2,3 um:zero minimum:500 name:EXTERNAL_REQUESTS : External invalidate requests
+
+event:15 counters:0,1,2,3 um:zero minimum:500 name:INSTRUCTIONS_GRADUATED : Instructions graduated
+event:16 counters:0,1,2,3 um:zero minimum:500 name:PREFETCH_INSTRUCTIONS_EXECUTED : Executed prefetch instructions
+event:17 counters:0,1,2,3 um:zero minimum:500 name:PREFETCH_MISSES_IN_DCACHE : Primary data cache misses by prefetch instructions
+event:18 counters:0,1,2,3 um:zero minimum:500 name:GRADUATED_LOADS : Graduated loads
+event:19 counters:0,1,2,3 um:zero minimum:500 name:GRADUATED_STORES : Graduated stores
+event:20 counters:0,1,2,3 um:zero minimum:500 name:GRADUATED_STORE_CONDITIONALS : Graduated store conditionals
+event:21 counters:0,1,2,3 um:zero minimum:500 name:GRADUATED_FP_INSTRUCTIONS : Graduated floating point instructions
+event:22 counters:0,1,2,3 um:zero minimum:500 name:QUADWORDS : Quadwords written back from primary data cache
+event:23 counters:0,1,2,3 um:zero minimum:500 name:TLB_MISSES : TLB misses
+event:24 counters:0,1,2,3 um:zero minimum:500 name:MISPREDICTED_BRANCHES : Mispredicted branches
+event:25 counters:0,1,2,3 um:zero minimum:500 name:DCACHE_MISSES : Primary data cache misses
+event:26 counters:0,1,2,3 um:zero minimum:500 name:SCACHE_MISSES : Secondary cache misses (data)
+event:27 counters:0,1,2,3 um:zero minimum:500 name:SCACHE_WAY_MISPREDICTION : Misprediction from scache way prediction table (data)
+event:28 counters:0,1,2,3 um:zero minimum:500 name:STATE_OF_SCACHE_INTERVENTION_HIT : State of external intervention hit in secondary cache
+event:29 counters:0,1,2,3 um:zero minimum:500 name:STATE_OF_EXTERNAL_INVALIDATION_HIT : State of external invalidation hits in secondary cache
+event:30 counters:0,1,2,3 um:zero minimum:500 name:STORE_PREFETCH_EXCLUSIVE_TO_CLEAN_SC_BLOCK : Store/prefetch exclusive to clean block in secondary cache
+event:31 counters:0,1,2,3 um:zero minimum:500 name:STORE_PREFETCH_EXCLUSIVE_SHARED_SC_BLOCK : Store/prefetch exclusive to shared block in secondary
diff -purN oprofile-0.8.1.orig/events/mips/r12000/unit_masks oprofile-0.8.1/events/mips/r12000/unit_masks
--- oprofile-0.8.1.orig/events/mips/r12000/unit_masks	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/r12000/unit_masks	Sun Dec 12 22:16:43 2004
@@ -0,0 +1,7 @@
+#
+# MIPS R12000 possible unit masks
+#
+# We don't support the R12000 conditional count feature yet.
+#
+name:zero type:mandatory default:0x0
+	0x0 No unit mask
diff -purN oprofile-0.8.1.orig/events/mips/rm7000/events oprofile-0.8.1/events/mips/rm7000/events
--- oprofile-0.8.1.orig/events/mips/rm7000/events	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/rm7000/events	Sun Dec 12 14:21:33 2004
@@ -0,0 +1,34 @@
+#
+# RM7000 events
+#
+event:0x00 counters:0,1 um:zero minimum:500 name:CYCLES : Clock cycles
+event:0x01 counters:0,1 um:zero minimum:500 name:INSTRUCTIONS_ISSUED : Total instructions issued
+event:0x02 counters:0,1 um:zero minimum:500 name:FP_INSTRUCTIONS_ISSUED : Floating-point instructions issued
+event:0x03 counters:0,1 um:zero minimum:500 name:INTEGER_INSTRUCTIONS_ISSUED : Integer instructions issued
+event:0x04 counters:0,1 um:zero minimum:500 name:LOAD_INSTRUCTIONS_ISSUED : Load instructions issued
+event:0x05 counters:0,1 um:zero minimum:500 name:STORE_INSTRUCTIONS_ISSUED : Store instructions issued
+event:0x06 counters:0,1 um:zero minimum:500 name:DUAL_ISSUED_PAIRS : Dual issued pairs
+event:0x07 counters:0,1 um:zero minimum:500 name:BRANCH_PREFETCHES : Branch prefetches
+event:0x08 counters:0,1 um:zero minimum:500 name:EXTERNAL_CACHE_MISSES : External Cache Misses
+event:0x09 counters:0,1 um:zero minimum:500 name:STALL_CYCLES : Stall cycles
+event:0x0a counters:0,1 um:zero minimum:500 name:SCACHE_MISSES : Secondary cache misses
+event:0x0b counters:0,1 um:zero minimum:500 name:ICACHE_MISSES : Instruction cache misses
+event:0x0c counters:0,1 um:zero minimum:500 name:DCACHE_MISSES : Data cache misses
+event:0x0d counters:0,1 um:zero minimum:500 name:DTLB_MISSES : Data TLB misses
+event:0x0e counters:0,1 um:zero minimum:500 name:ITLB_MISSES : Instruction TLB misses
+event:0x0f counters:0,1 um:zero minimum:500 name:JTLB_INSTRUCTION_MISSES : Joint TLB instruction misses
+event:0x10 counters:0,1 um:zero minimum:500 name:JTLB_DATA_MISSES : Joint TLB data misses
+event:0x11 counters:0,1 um:zero minimum:500 name:BRANCHES_TAKEN : Branches taken
+event:0x12 counters:0,1 um:zero minimum:500 name:BRANCHES_ISSUED : Branches issued
+event:0x13 counters:0,1 um:zero minimum:500 name:SCACHE_WRITEBACKS : Secondary cache writebacks
+event:0x14 counters:0,1 um:zero minimum:500 name:PCACHE_WRITEBACKS : Primary cache writebacks
+event:0x15 counters:0,1 um:zero minimum:500 name:DCACHE_MISS_STALL_CYCLES : Dcache miss stall cycles (cycles where both cache miss tokens taken and a third try is requested)
+event:0x16 counters:0,1 um:zero minimum:500 name:CACHE_MISSES : Cache misses
+event:0x17 counters:0,1 um:zero minimum:500 name:FP_EXCEPTION_STALL_CYCLES : FP possible exception cycles
+event:0x18 counters:0,1 um:zero minimum:500 name:SLIP_CYCLES_DUE_MULTIPLIER_BUSY : Slip Cycles due to multiplier busy
+event:0x19 counters:0,1 um:zero minimum:500 name:COP0_SLIP_CYCLES : Coprocessor 0 slip cycles
+event:0x1a counters:0,1 um:zero minimum:500 name:SLIP_CYCLES_PENDING_NON_BLKING_LOAD : Slip cycles due to pending non-blocking loads
+event:0x1c counters:0,1 um:zero minimum:500 name:WRITE_BUFFER_FULL_STALL_CYCLES : Write buffer full stall cycles
+event:0x1d counters:0,1 um:zero minimum:500 name:CACHE_INSTRUCTION_STALL_CYCLES : Cache instruction stall cycles
+event:0x1e counters:0,1 um:zero minimum:500 name:MULTIPLIER_STALL_CYCLES : Multiplier stall cycles
+event:0x1f counters:0,1 um:zero minimum:500 name:STALL_CYCLES_PENDING_NON_BLKING_LOAD : Stall cycles due to pending non-blocking loads - stall start of exception
diff -purN oprofile-0.8.1.orig/events/mips/rm7000/unit_masks oprofile-0.8.1/events/mips/rm7000/unit_masks
--- oprofile-0.8.1.orig/events/mips/rm7000/unit_masks	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/rm7000/unit_masks	Sun Dec 12 22:17:03 2004
@@ -0,0 +1,5 @@
+#
+# MIPS RM7000 possible unit masks
+#
+name:zero type:mandatory default:0x0
+	0x0 No unit mask
diff -purN oprofile-0.8.1.orig/events/mips/rm9000/events oprofile-0.8.1/events/mips/rm9000/events
--- oprofile-0.8.1.orig/events/mips/rm9000/events	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/rm9000/events	Sun Dec 12 22:02:58 2004
@@ -0,0 +1,32 @@
+#
+# RM9000 events
+#
+event:0x00 counters:0,1 um:zero minimum:500 name:CYCLES : Processor clock cycles
+event:0x01 counters:0,1 um:zero minimum:500 name:INSTRUCTIONS_ISSUED : Instructions issued
+event:0x02 counters:0,1 um:zero minimum:500 name:FP_INSTRUCTIONS_ISSUED : Floating-point instructions issued
+event:0x03 counters:0,1 um:zero minimum:500 name:INT_INSTRUCTIONS_ISSUED : Integer instructions issued
+event:0x04 counters:0,1 um:zero minimum:500 name:LOAD_INSTRUCTIONS_ISSUED : Load instructions issued
+event:0x05 counters:0,1 um:zero minimum:500 name:STORE_INSTRUCTIONS_ISSUED : Store instructions issued
+event:0x06 counters:0,1 um:zero minimum:500 name:INSTRUCTIONS_DUAL_ISSUED : Dual-issued instruction pairs
+event:0x07 counters:0,1 um:zero minimum:500 name:BRANCH_MISSPREDICTS : Branch mispredictions
+event:0x09 counters:0,1 um:zero minimum:500 name:STALL_CYCLES : Stall cycles
+event:0x0a counters:0,1 um:zero minimum:500 name:L2_CACHE_MISSES : L2 cache misses
+event:0x0b counters:0,1 um:zero minimum:500 name:ICACHE_MISSES : Icache misses
+event:0x0c counters:0,1 um:zero minimum:500 name:DCACHE_MISSES : Dcache misses
+event:0x0d counters:0,1 um:zero minimum:500 name:DTLB_MISSES : Data TLB misses
+event:0x0e counters:0,1 um:zero minimum:500 name:ITLB_MISSES : Instruction TLB misses
+event:0x0f counters:0,1 um:zero minimum:500 name:JTLB_INSTRUCTION_MISSES : Joint TLB instruction misses
+event:0x10 counters:0,1 um:zero minimum:500 name:JTLB_DATA_MISSES : Joint TLB data misses
+event:0x11 counters:0,1 um:zero minimum:500 name:BRANCHES_TAKEN : Branches taken
+event:0x12 counters:0,1 um:zero minimum:500 name:BRANCHES_ISSUED : Branch instructions issued
+event:0x13 counters:0,1 um:zero minimum:500 name:L2_WRITEBACKS : L2 cache writebacks
+event:0x14 counters:0,1 um:zero minimum:500 name:DCACHE_WRITEBACKS : Dcache writebacks
+event:0x15 counters:0,1 um:zero minimum:500 name:DCACHE_MISS_STALL_CYCLES : Dcache-miss stall cycles
+event:0x16 counters:0,1 um:zero minimum:500 name:CACHE_REMISSES : Cache remisses
+event:0x17 counters:0,1 um:zero minimum:500 name:FP_POSSIBLE_EXCEPTION_CYCLES : Floating-point possible exception cycles
+event:0x18 counters:0,1 um:zero minimum:500 name:MULTIPLIER_BUSY_SLIP_CYCLES : Slip cycles due to busy multiplier
+event:0x19 counters:0,1 um:zero minimum:500 name:COP0_SLIP_CYCLES : Co-processor 0 slip cycles
+event:0x1a counters:0,1 um:zero minimum:500 name:NONBLOCKING_LOAD_SLIP_CYCLES : Slip cycles due to pending non-blocking loads
+event:0x1b counters:0,1 um:zero minimum:500 name:WRITE_BUFFER_FULL_STALL_CYCLES : Stall cycles due to a full write buffer
+event:0x1c counters:0,1 um:zero minimum:500 name:CACHE_INSN_STALL_CYCLES : Stall cycles due to cache instructions
+event:0x1e counters:0,1 um:zero minimum:500 name:NONBLOCKING_LOAD_PENDING_EXCEPTION_STALL_CYCLES : Stall cycles due to pending non-blocking loads - stall start of exception
diff -purN oprofile-0.8.1.orig/events/mips/rm9000/unit_masks oprofile-0.8.1/events/mips/rm9000/unit_masks
--- oprofile-0.8.1.orig/events/mips/rm9000/unit_masks	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/rm9000/unit_masks	Fri Dec 10 01:50:51 2004
@@ -0,0 +1,5 @@
+#
+# MIPS RM9000 possible unit masks
+#
+name:zero type:mandatory default:0x0
+	0x0 No unit mask
diff -purN oprofile-0.8.1.orig/events/mips/sb1/events oprofile-0.8.1/events/mips/sb1/events
--- oprofile-0.8.1.orig/events/mips/sb1/events	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/sb1/events	Sun Dec 12 22:18:28 2004
@@ -0,0 +1,73 @@
+#
+# Sibyte SB1 events
+#
+
+event:16 counters:0,1,2,3 um:zero minimum:500 name:CYCLES :Elapsed cycles
+
+# Execution Counts and Instruction Slotting
+event:40 counters:1,2,3 um:zero minimum:500 name:ISSUE_L0 :Issue to L0
+event:41 counters:1,2,3 um:zero minimum:500 name:ISSUE_L1 :Issue to L0
+event:42 counters:1,2,3 um:zero minimum:500 name:ISSUE_E0 :Issue to E0
+event:43 counters:1,2,3 um:zero minimum:500 name:ISSUE_E1 :Issue to E1
+
+# Explaining Sub-Peak Performance: Pipeline Traps
+event:47 counters:1,2,3 um:zero minimum:500 name:BRANCH_MISSPREDICTS :Branch mispredicts
+event:29 counters:0,1,2,3 um:zero minimum:500 name:MBOX_REPLAY :MBOX replay
+event:28 counters:0,1,2,3 um:zero minimum:500 name:DCFIFO :DCFIFO
+event:30 counters:0,1,2,3 um:zero minimum:500 name:DATA_DEPENDENCY_REPLAY :Data dependency replay
+event:27 counters:0,1,2,3 um:zero minimum:500 name:DCACHE_FILL_REPLAY :Dcache fill replay
+event:31 counters:0,1,2,3 um:zero minimum:500 name:ANY_REPLAY :Any replay except mispredict
+
+
+# Explaining Sub-Peak Performance: static and dynamic stalls
+event:32 counters:1,2,3 um:zero minimum:500 name:MAX_ISSUE :Max issue
+event:33 counters:1,2,3 um:zero minimum:500 name:NO_VALID_INSN :No valid instr to issue
+event:34 counters:1,2,3 um:zero minimum:500 name:CONSUMER_WAITING_FOR_LOAD :load consumer waiting for dfill
+event:35 counters:1,2,3 um:zero minimum:500 name:NOT_DATA_READY :Not data ready
+event:36 counters:1,2,3 um:zero minimum:500 name:RESOURCE_CONSTRAINT :Resource (L0/1 E0/1) constraint
+event:37 counters:1,2,3 um:zero minimum:500 name:ISSUE_CONFLICT_DUE_IMISS :issue conflict due to imiss using LS0
+event:38 counters:1,2,3 um:zero minimum:500 name:ISSUE_CONFLICT_DUE_DFILL :issue conflict due to dfill using LS0/1
+
+# Grouping Co-issued Instructions
+event:39 counters:1,2,3 um:zero minimum:500 name:INSN_STAGE4 :One or more instructions survives stage 4
+
+# Branch information
+event:44 counters:1,2,3 um:zero minimum:500 name:BRANCH_STAGE4 :Branch survived stage 4
+event:45 counters:1,2,3 um:zero minimum:500 name:BRANCH_REALLY_TAKEN :Conditional branch was really taken
+event:46 counters:1,2,3 um:zero minimum:500 name:BRANCH_PREDICTED_TAKEN :Predicted taken conditional branch
+
+# Cache access
+event:1 counters:1,2,3 um:zero minimum:500 name:RQ_LENGTH :Read queue length
+event:2 counters:1,2,3 um:zero minimum:500 name:UNCACHED_RQ_LENGTH :Number of valid uncached entries in read queue
+event:3 counters:1,2,3 um:zero minimum:500 name:DCACHE_READ_MISS :Dcache read results in a miss
+
+event:10 counters:1,2,3 um:zero minimum:500 name:DCACHE_FILLED_SHD_NONC_EXC :Dcache is filled (shared, nonc, exclusive)
+event:11 counters:1,2,3 um:zero minimum:500 name:DCACHE_FILL_SHARED_LINE :Dcache is filled with shared line
+event:12 counters:1,2,3 um:zero minimum:500 name:DCACHE_READ_MISS :Dcache read results in a miss
+event:15 counters:1,2,3 um:zero minimum:500 name:WRITEBACK_RETURNS :Number of instruction returns
+event:13 counters:1,2,3 um:zero minimum:500 name:VICTIM_WRITEBACK :A writeback occurs due to replacement
+event:7 counters:1,2,3 um:zero minimum:500 name:UPGRADE_SHARED_TO_EXCLUSIVE :A line is upgraded from shared to exclusive
+event:6 counters:1,2,3 um:zero minimum:500 name:LD_ST_HITS_PREFETCH_IN_QUEUE :Load/store hits prefetch in read queue
+event:5 counters:1,2,3 um:zero minimum:500 name:PREFETCH_HITS_CACHE_OR_READ_Q :Prefetch hits in cache or read queue
+event:4 counters:1,2,3 um:zero minimum:500 name:READ_HITS_READ_Q :Read hits in read queue
+
+# BIU
+
+event:17 counters:1,2,3 um:zero minimum:500 name:BIU_STALLS_ON_ZB_ADDR_BUS :BIU stalls on ZB addr bus
+event:18 counters:1,2,3 um:zero minimum:500 name:BIU_STALLS_ON_ZB_DATA_BUS :BIU stalls on ZB data bus
+event:19 counters:1,2,3 um:zero minimum:500 name:READ_RQ_SENT_TO_ABUS :Requests sent to ZB Abus
+event:20 counters:1,2,3 um:zero minimum:500 name:READ_RQ_NOPS_SENT_TO_ABUS :Read requests and NOPs sent to ZB Abus
+event:21 counters:1,2,3 um:zero minimum:500 name:READ_RQ_SENT_TO_ABUS :Read requests sent to ZB Abus
+event:14 counters:1,2,3 um:zero minimum:500 name:MBOX_RQ_WHEN_BIU_BUSY :MBOX requests to BIU when BIU busy
+
+# Multiprocessor
+event:26 counters:1,2,3 um:zero minimum:500 name:STORE_COND_FAILED :Failed store conditional
+event:22 counters:1,2,3 um:zero minimum:500 name:SNOOP_RQ_HITS :Snoop request hits anywhere
+event:23 counters:1,2,3 um:zero minimum:500 name:SNOOP_ADDR_Q_FULL :Snoop address queue is full
+event:24 counters:1,2,3 um:zero minimum:500 name:R_RESP_OTHER_CORE :Read response comes from the other core
+event:25 counters:1,2,3 um:zero minimum:500 name:R_RESP_OTHER_CORE_D_MOD :Read response comes from the other core with D_MOD set
+
+# Instruction Counts
+event:8 counters:1,2,3 um:zero minimum:500 name:LOAD_SURVIVED_STAGE4 :Load survived stage 4
+event:9 counters:1,2,3 um:zero minimum:500 name:STORE_SURVIVED_STAGE4  :Store survived stage 4
+event:0 counters:1,2,3 um:zero minimum:500 name:INSN_SURVIVED_STAGE7 :Instruction survived stage 7
diff -purN oprofile-0.8.1.orig/events/mips/sb1/unit_masks oprofile-0.8.1/events/mips/sb1/unit_masks
--- oprofile-0.8.1.orig/events/mips/sb1/unit_masks	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/sb1/unit_masks	Sun Dec 12 22:18:49 2004
@@ -0,0 +1,5 @@
+#
+# Sibyte SB1 possible unit masks
+#
+name:zero type:mandatory default:0x0
+	0x0 No unit mask
diff -purN oprofile-0.8.1.orig/events/mips/vr5432/events oprofile-0.8.1/events/mips/vr5432/events
--- oprofile-0.8.1.orig/events/mips/vr5432/events	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/vr5432/events	Sun Dec 12 22:09:02 2004
@@ -0,0 +1,14 @@
+#
+# VR5432 events
+#
+event:0 counters:0,1 um:zero minimum:500 name:CYCLES : Processor cycles (PClock)
+event:1 counters:0,1 um:zero minimum:500 name:INSTRUCTIONS_EXECUTED : (Instructions executed)/2 and truncated
+event:2 counters:0,1 um:zero minimum:500 name:LOAD_PREF_CACHE_INSTRUCTIONS : Load, prefetch/CacheOps execution (no sync)
+event:3 counters:0,1 um:zero minimum:500 name:STORES : Store execution
+event:4 counters:0,1 um:zero minimum:500 name:BRANCHES : Branch execution (no jumps or jump registers)
+event:5 counters:0,1 um:zero minimum:500 name:FP_INSTRUCTIONS : (FP instruction execution) / 2 and truncated excluding cp1 loads and stores
+event:6 counters:0,1 um:zero minimum:500 name:DOUBLEWORDS_FLUSHED : Doublewords flushed to main memory (no uncached stores)
+event:7 counters:0,1 um:zero minimum:500 name:JTLB_REFILLS : JTLB refills
+event:8 counters:0,1 um:zero minimum:500 name:DCACHE_MISSES : Data cache misses (no I-cache misses)
+event:9 counters:0,1 um:zero minimum:500 name:ICACHE_MISSES : Instruction cache misses (no D-cache misses)
+event:10 counters:0,1 um:zero minimum:500 name:BRANCHES_MISPREDICTED : Branches mispredicted
diff -purN oprofile-0.8.1.orig/events/mips/vr5432/unit_masks oprofile-0.8.1/events/mips/vr5432/unit_masks
--- oprofile-0.8.1.orig/events/mips/vr5432/unit_masks	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/vr5432/unit_masks	Sun Dec 12 22:16:51 2004
@@ -0,0 +1,5 @@
+#
+# MIPS VR5432 possible unit masks
+#
+name:zero type:mandatory default:0x0
+	0x0 No unit mask
diff -purN oprofile-0.8.1.orig/events/mips/vr5500/events oprofile-0.8.1/events/mips/vr5500/events
--- oprofile-0.8.1.orig/events/mips/vr5500/events	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/vr5500/events	Sun Dec 12 22:10:11 2004
@@ -0,0 +1,16 @@
+#
+# VR5500, VR5532 and VR7701 events
+#
+# Very similar to what the VR5432 provides.
+#
+event:0 counters:0,1 um:zero minimum:500 name:CYCLES : Processor clock cycles
+event:1 counters:0,1 um:zero minimum:500 name:INSTRUCTIONS_EXECUTED : Instructions executed
+event:2 counters:0,1 um:zero minimum:500 name:LOAD_PREF_CACHE_INSTRUCTIONS : Execution of load/prefetch/cache instruction
+event:3 counters:0,1 um:zero minimum:500 name:STORES : Execution of store instruction
+event:4 counters:0,1 um:zero minimum:500 name:BRANCHES : Execution of branch instruction
+event:5 counters:0,1 um:zero minimum:500 name:FP_INSTRUCTIONS : Execution of floating-point instruction
+event:6 counters:0,1 um:zero minimum:500 name:DOUBLEWORDS_FLUSHED : Doubleword flush to main memory
+event:7 counters:0,1 um:zero minimum:500 name:JTLB_REFILLS : TLB refill
+event:8 counters:0,1 um:zero minimum:500 name:DCACHE_MISSES : Data cache miss
+event:9 counters:0,1 um:zero minimum:500 name:ICACHE_MISSES : Instruction cache miss
+event:10 counters:0,1 um:zero minimum:500 name:BRANCHES_MISPREDICTED : Branch prediction miss
diff -purN oprofile-0.8.1.orig/events/mips/vr5500/unit_masks oprofile-0.8.1/events/mips/vr5500/unit_masks
--- oprofile-0.8.1.orig/events/mips/vr5500/unit_masks	Thu Jan  1 01:00:00 1970
+++ oprofile-0.8.1/events/mips/vr5500/unit_masks	Sun Dec 12 22:16:57 2004
@@ -0,0 +1,5 @@
+#
+# MIPS VR5500 possible unit masks
+#
+name:zero type:mandatory default:0x0
+	0x0 No unit mask
diff -purN oprofile-0.8.1.orig/libop/op_cpu_type.c oprofile-0.8.1/libop/op_cpu_type.c
--- oprofile-0.8.1.orig/libop/op_cpu_type.c	Sat May 29 19:02:59 2004
+++ oprofile-0.8.1/libop/op_cpu_type.c	Sun Dec 12 21:56:54 2004
@@ -43,6 +43,13 @@ static struct cpu_descr const cpu_descrs
 	{ "Pentium M (P6 core)", "i386/p6_mobile", CPU_P6_MOBILE, 2 },
 	{ "ARM/XScale PMU1", "arm/xscale1", CPU_ARM_XSCALE1, 3 },
 	{ "ARM/XScale PMU2", "arm/xscale2", CPU_ARM_XSCALE2, 5 },
+	{ "MIPS R10000", "mips/r10000", CPU_MIPS_R10000, 2},
+	{ "MIPS R12000", "mips/r12000", CPU_MIPS_R12000, 4},
+	{ "QED RM7000", "mips/rm7000", CPU_MIPS_RM7000, 1},
+	{ "PMC-Sierra RM9000", "mips/rm9000", CPU_MIPS_RM9000, 2},
+	{ "Sibyte SB1", "mips/sb1", CPU_MIPS_SB1, 4},
+	{ "NEC VR5432", "mips/vr5432", CPU_MIPS_VR5432, 2},
+	{ "NEC VR5500", "mips/vr5500", CPU_MIPS_VR5500, 2},
 };
  
 static size_t const nr_cpu_descrs = sizeof(cpu_descrs) / sizeof(struct cpu_descr);
diff -purN oprofile-0.8.1.orig/libop/op_cpu_type.h oprofile-0.8.1/libop/op_cpu_type.h
--- oprofile-0.8.1.orig/libop/op_cpu_type.h	Sat May 29 19:02:59 2004
+++ oprofile-0.8.1/libop/op_cpu_type.h	Sun Dec 12 21:41:21 2004
@@ -39,6 +39,13 @@ typedef enum {
 	CPU_P6_MOBILE, /**< Pentium M series */
 	CPU_ARM_XSCALE1, /**< ARM XScale 1 */
 	CPU_ARM_XSCALE2, /**< ARM XScale 2 */
+	CPU_MIPS_R10000, /**< MIPS R10000 */
+	CPU_MIPS_R12000, /**< MIPS R12000 */
+	CPU_MIPS_RM7000, /**< QED  RM7000 */
+	CPU_MIPS_RM9000, /**< PMC-Sierra RM9000 */
+	CPU_MIPS_SB1, /**< Broadcom SB1 */
+	CPU_MIPS_VR5432, /**< NEC VR5432 */
+	CPU_MIPS_VR5500, /**< MIPS VR5500, VR5532 and VR7701 */
 	MAX_CPU_TYPE
 } op_cpu;
 
diff -purN oprofile-0.8.1.orig/libop/op_events.c oprofile-0.8.1/libop/op_events.c
--- oprofile-0.8.1.orig/libop/op_events.c	Sat May 29 19:02:59 2004
+++ oprofile-0.8.1/libop/op_events.c	Sun Dec 12 22:10:30 2004
@@ -651,6 +651,25 @@ void op_default_event(op_cpu cpu_type, s
 			descr->name = "CPU_CYCLES";
 			break;
 
+		case CPU_MIPS_R10000:
+		case CPU_MIPS_R12000:
+			descr->name = "INSTRUCTIONS_GRADUATED";
+			break;
+
+		case CPU_MIPS_RM7000:
+		case CPU_MIPS_RM9000:
+			descr->name = "INSTRUCTIONS_ISSUED";
+			break;
+
+		case CPU_MIPS_SB1:
+			descr->name = "INSN_SURVIVED_STAGE7";
+			break;
+
+		case CPU_MIPS_VR5432:
+		case CPU_MIPS_VR5500:
+			descr->name = "INSTRUCTIONS_EXECUTED";
+			break;
+
 		// don't use default, if someone add a cpu he wants a compiler
 		// warning if he forgets to handle it here.
 		case CPU_TIMER_INT:

Hi,
Are there any guidelines while setting counter values of events while
profiling? Say for example should I be targeting to measure 2000
samples per second and then calculate the counter value based on the
speed of my CPU, i.e if I have a 2 Ghz CPU I should use a counter
value of 2G/2000 = 1000000 and in case of a 3GHz CPU set it to
1500000?
Also is it a good (and safe) idea to set the counter for events like
branch mispredictions and cache misses to  4-5 times lower value than
the CPU (time) counter, expecting that fewer such events would be
generated than compared to the CPU event.
I would be highly obliged if someone can throw some light on this issue.  
Thanking in advance.
Shourya

Hello all,
I am unable to install the opofile callgraph patch for x86
architecture on kernel 2.6.4. I will highly grateful if some one can
tell me what I am doing wrong. These are the steps I have followed.
1. In the source directory of kernel 2.6.4, I did "patch -p1 <
callgraph-0.3.diff". It complained for some missing files in non-x86
architectures. I ignored those and continued.
2. I then configured the kernel with support for Oprofile and framepointers.
3. On doing "make bzImage" I get the following error.
========================================
drivers/oprofile/cpu_buffer.c: In function `log_sample':
drivers/oprofile/cpu_buffer.c:189: warning: `return' with no value, in
function returning non-void
drivers/oprofile/cpu_buffer.c:192: error: `eip' undeclared (first use
in this fu nction)
drivers/oprofile/cpu_buffer.c:192: error: (Each undeclared identifier
is reporte d only once
drivers/oprofile/cpu_buffer.c:192: error: for each function it appears in.)
drivers/oprofile/cpu_buffer.c: At top level:
drivers/oprofile/cpu_buffer.c:200: error: redefinition of `cpu_buffer_reset'
drivers/oprofile/cpu_buffer.c:87: error: `cpu_buffer_reset' previously
defined h ere
drivers/oprofile/cpu_buffer.c:154: warning: `log_sample' defined but not used
make[1]: *** [arch/i386/oprofile/../../../drivers/oprofile/cpu_buffer.o] Error 1
make: *** [arch/i386/oprofile] Error 2
========================================

Once again, It will be extremely helpful if some one can help me out here. 

Shourya

Re: Oprofile callgraph patch problems From: Greg Banks <gnb@me...> - 2004-12-06 09:22 On Mon, 2004-12-06 at 20:18, Shourya P. Bhattacharya wrote: > 1. In the source directory of kernel 2.6.4, I did "patch -p1 < > callgraph-0.3.diff". It complained for some missing files in non-x86 > architectures. I ignored those and continued. Can you post the actual output from patch? I suspect some important hunk went astray. Greg. -- Greg Banks, R&D Software Engineer, SGI Australian Software Group. I don't speak for SGI.

Re: Oprofile callgraph patch problems From: Shourya P. Bhattacharya <sbhattacharya@gm...> - 2004-12-06 09:44 On Mon, 06 Dec 2004 20:21:47 +1100, Greg Banks <gnb@...> wrote: > On Mon, 2004-12-06 at 20:18, Shourya P. Bhattacharya wrote: > > 1. In the source directory of kernel 2.6.4, I did "patch -p1 < > > callgraph-0.3.diff". It complained for some missing files in non-x86 > > architectures. I ignored those and continued. > > Can you post the actual output from patch? I suspect some > important hunk went astray. Thank you for your reply. On going through the patch output I found the following -------------------------------------------------------------------------------------- patching file drivers/oprofile/buffer_sync.c patching file drivers/oprofile/cpu_buffer.c Hunk #2 succeeded at 61 with fuzz 2. Hunk #5 FAILED at 186. 1 out of 5 hunks FAILED -- saving rejects to file drivers/oprofile/cpu_buffer.c.rej ....... patching file drivers/oprofile/oprof.c patching file drivers/oprofile/oprof.h patching file drivers/oprofile/oprofile_files.c Hunk #1 succeeded at 18 with fuzz 2. Hunk #2 succeeded at 63 with fuzz 2. Hunk #3 FAILED at 78. Hunk #4 FAILED at 107. 2 out of 5 hunks FAILED -- saving rejects to file drivers/oprofile/oprofile_files.c.rej ------------------------------------------------------------------- Could this be the cause of the problem? I am also attaching the full text of patch output. Thank you once again. Shourya OUTPUT of the Patch **************************************************************** patching file arch/alpha/oprofile/common.c patching file arch/alpha/oprofile/op_model_ev4.c patching file arch/alpha/oprofile/op_model_ev5.c patching file arch/alpha/oprofile/op_model_ev6.c patching file arch/alpha/oprofile/op_model_ev67.c can't find file to patch at input line 147 Perhaps you used the wrong -p or --strip option? The text leading up to this was: -------------------------- |Index: linux-cvs/arch/arm/oprofile/common.c |=================================================================== |RCS file: /home/moz/cvs//linux-2.5/arch/arm/oprofile/common.c,v |retrieving revision 1.2 |diff -u -a -p -r1.2 common.c |--- linux-cvs/arch/arm/oprofile/common.c 12 Apr 2004 22:07:11 -0000 1.2 |+++ linux-cvs/arch/arm/oprofile/common.c 15 Jul 2004 15:20:24 -0000 -------------------------- File to patch: Skip this patch? [y] Skipping patch. 2 out of 2 hunks ignored patching file arch/arm/oprofile/init.c Hunk #1 FAILED at 12. 1 out of 1 hunk FAILED -- saving rejects to file arch/arm/oprofile/init.c.rej can't find file to patch at input line 221 Perhaps you used the wrong -p or --strip option? The text leading up to this was: -------------------------- |Index: linux-cvs/arch/arm/oprofile/op_model_xscale.c |=================================================================== |RCS file: /home/moz/cvs//linux-2.5/arch/arm/oprofile/op_model_xscale.c,v |retrieving revision 1.3 |diff -u -a -p -r1.3 op_model_xscale.c |--- linux-cvs/arch/arm/oprofile/op_model_xscale.c 29 Apr 2004 14:10:01 -0000 1.3 |+++ linux-cvs/arch/arm/oprofile/op_model_xscale.c 15 Jul 2004 02:52:21 -0000 -------------------------- File to patch: Skip this patch? [y] Skipping patch. 2 out of 2 hunks ignored patching file arch/i386/oprofile/Makefile patching file arch/i386/oprofile/backtrace.c patching file arch/i386/oprofile/init.c patching file arch/i386/oprofile/nmi_int.c Hunk #2 succeeded at 297 (offset -3 lines). Hunk #3 succeeded at 305 (offset -3 lines). Hunk #4 succeeded at 328 (offset -3 lines). Hunk #5 succeeded at 336 (offset -3 lines). Hunk #6 succeeded at 352 (offset -3 lines). Hunk #7 succeeded at 370 (offset -3 lines). Hunk #8 succeeded at 385 (offset -3 lines). Hunk #9 succeeded at 406 (offset -3 lines). patching file arch/i386/oprofile/nmi_timer_int.c patching file arch/i386/oprofile/op_model_athlon.c patching file arch/i386/oprofile/op_model_p4.c Hunk #1 succeeded at 592 (offset 3 lines). Hunk #2 succeeded at 627 (offset 3 lines). patching file arch/i386/oprofile/op_model_ppro.c patching file arch/i386/oprofile/op_x86_model.h patching file arch/ia64/oprofile/init.c patching file arch/parisc/oprofile/init.c can't find file to patch at input line 762 Perhaps you used the wrong -p or --strip option? The text leading up to this was: -------------------------- |Index: linux-cvs/arch/ppc64/oprofile/common.c |=================================================================== |RCS file: /home/moz/cvs//linux-2.5/arch/ppc64/oprofile/common.c,v |retrieving revision 1.4 |diff -u -a -p -r1.4 common.c |--- linux-cvs/arch/ppc64/oprofile/common.c 27 Jun 2004 17:52:30 -0000 1.4 |+++ linux-cvs/arch/ppc64/oprofile/common.c 15 Jul 2004 03:26:53 -0000 -------------------------- File to patch: Skip this patch? [y] Skipping patch. 3 out of 3 hunks ignored can't find file to patch at input line 848 Perhaps you used the wrong -p or --strip option? The text leading up to this was: -------------------------- |Index: linux-cvs/arch/ppc64/oprofile/op_model_power4.c |=================================================================== |RCS file: /home/moz/cvs//linux-2.5/arch/ppc64/oprofile/op_model_power4.c,v |retrieving revision 1.4 |diff -u -a -p -r1.4 op_model_power4.c |--- linux-cvs/arch/ppc64/oprofile/op_model_power4.c 27 Jun 2004 17:52:30 -0000 1.4 |+++ linux-cvs/arch/ppc64/oprofile/op_model_power4.c 15 Jul 2004 03:26:53 -0000 -------------------------- File to patch: Skip this patch? [y] Skipping patch. 2 out of 2 hunks ignored can't find file to patch at input line 876 Perhaps you used the wrong -p or --strip option? The text leading up to this was: -------------------------- |Index: linux-cvs/arch/ppc64/oprofile/op_model_rs64.c |=================================================================== |RCS file: /home/moz/cvs//linux-2.5/arch/ppc64/oprofile/op_model_rs64.c,v |retrieving revision 1.2 |diff -u -a -p -r1.2 op_model_rs64.c |--- linux-cvs/arch/ppc64/oprofile/op_model_rs64.c 12 Apr 2004 19:47:59 -0000 1.2 |+++ linux-cvs/arch/ppc64/oprofile/op_model_rs64.c 21 Jun 2004 03:42:20 -0000 -------------------------- File to patch: Skip this patch? [y] Skipping patch. 2 out of 2 hunks ignored can't find file to patch at input line 900 Perhaps you used the wrong -p or --strip option? The text leading up to this was: -------------------------- |Index: linux-cvs/arch/s390/oprofile/init.c |=================================================================== |RCS file: /home/moz/cvs//linux-2.5/arch/s390/oprofile/init.c,v |retrieving revision 1.2 |diff -u -a -p -r1.2 init.c |--- linux-cvs/arch/s390/oprofile/init.c 29 Apr 2004 14:20:11 -0000 1.2 |+++ linux-cvs/arch/s390/oprofile/init.c 15 Jul 2004 03:11:58 -0000 -------------------------- File to patch: Skip this patch? [y] Skipping patch. 1 out of 1 hunk ignored patching file arch/sh/oprofile/op_model_null.c patching file arch/sparc64/oprofile/init.c patching file drivers/oprofile/buffer_sync.c patching file drivers/oprofile/cpu_buffer.c Hunk #2 succeeded at 61 with fuzz 2. Hunk #5 FAILED at 186. 1 out of 5 hunks FAILED -- saving rejects to file drivers/oprofile/cpu_buffer.c.rej patching file drivers/oprofile/cpu_buffer.h patching file drivers/oprofile/event_buffer.h patching file drivers/oprofile/oprof.c patching file drivers/oprofile/oprof.h patching file drivers/oprofile/oprofile_files.c Hunk #1 succeeded at 18 with fuzz 2. Hunk #2 succeeded at 63 with fuzz 2. Hunk #3 FAILED at 78. Hunk #4 FAILED at 107. 2 out of 5 hunks FAILED -- saving rejects to file drivers/oprofile/oprofile_files.c.rej patching file drivers/oprofile/oprofile_stats.c patching file drivers/oprofile/oprofile_stats.h patching file drivers/oprofile/timer_int.c patching file include/linux/oprofile.h patching file mm/memory.c Hunk #1 succeeded at 674 (offset 14 lines). > > Greg. > -- > Greg Banks, R&D Software Engineer, SGI Australian Software Group. > I don't speak for SGI. > >

Re: Oprofile callgraph patch problems From: Greg Banks <gnb@me...> - 2004-12-06 11:02 On Mon, 2004-12-06 at 20:44, Shourya P. Bhattacharya wrote: > [...] On going through the patch output I found > the following [...] > patching file drivers/oprofile/cpu_buffer.c [...] > Hunk #5 FAILED at 186. > [...] > patching file drivers/oprofile/oprofile_files.c [...] > Hunk #3 FAILED at 78. > Hunk #4 FAILED at 107. > ------------------------------------------------------------------- > Could this be the cause of the problem? Yes. I don't know if someone has the current callgraph patch ported to 2.6.4. The previous revision of the patch (0.2) did apply to 2.6.4. Greg. -- Greg Banks, R&D Software Engineer, SGI Australian Software Group. I don't speak for SGI.

Re: Oprofile callgraph patch problems From: Greg Banks <gnb@me...> - 2004-12-06 11:19 Attachments: callgraph-pre-fix      On Mon, 2004-12-06 at 20:44, Shourya P. Bhattacharya wrote: > On Mon, 06 Dec 2004 20:21:47 +1100, Greg Banks <gnb@...> wrote: > > On Mon, 2004-12-06 at 20:18, Shourya P. Bhattacharya wrote: > > > 1. In the source directory of kernel 2.6.4, I did "patch -p1 < > > > callgraph-0.3.diff". It complained for some missing files in non-x86 > > > architectures. I ignored those and continued. > > > > Can you post the actual output from patch? I suspect some > > important hunk went astray. > > Thank you for your reply. On going through the patch output I found > the following Try applying the attached patch to a clean 2.6.4 tree *before* applying the callgraph-0.3 patch. The latter should then apply. Greg. -- Greg Banks, R&D Software Engineer, SGI Australian Software Group. I don't speak for SGI.

Re: Oprofile callgraph patch problems From: Shourya P. Bhattacharya <sbhattacharya@gm...> - 2004-12-06 12:54 > > Try applying the attached patch to a clean 2.6.4 tree *before* > applying the callgraph-0.3 patch. The latter should then apply. > The pre-patch did work. Thanks a lot. Why don't you put it up on the official site along with the callgraph-0.3 patch? I think it will be helpful for others as well. Once again thanks a lot. Shourya

Re: Oprofile callgraph patch problems From: John Levon <levon@mo...> - 2004-12-06 12:58 On Mon, Dec 06, 2004 at 06:24:40PM +0530, Shourya P. Bhattacharya wrote: > > Try applying the attached patch to a clean 2.6.4 tree *before* > > applying the callgraph-0.3 patch. The latter should then apply. > > The pre-patch did work. Thanks a lot. Why don't you put it up on the > official site along with the callgraph-0.3 patch? I think it will be > helpful for others as well. Once again thanks a lot. Because we don't have time to maintain correct patches for completely outdated kernel versions john

On Nov 30, 2004, at 13:51, John Levon wrote:

> On Tue, Nov 30, 2004 at 11:41:45AM -0600, Andy Fleming wrote:
>
>> Oops.  I had a Changelog entry, but it seems to have disappeared.  It
>> is now added.
>
> Thanks (not in the proper format, but I'll tidy that up)

Oops.  Thanks, I'll do better next time

>
>>> What's the status of the kernel side patch for this?
>>
>> The kernel side patch has been submitted, and awaits approval.
>
> I'd rather wait for it to end up in some official tree first (e.g.
> ppc maintainer's tree, etc.) before applying the userspace changes

Ok, that's reasonable (as long as they don't say the same thing!)

UPDATE: Dec 3, 2004:
Patch has been submitted to Andrew Morton.
Sorry about the delayed reply, this window got hidden.

Andy

2 messages have been excluded from this view by a project administrator.

Next Messages