2003-12-02 Stephane Eranian <eranian@hpl.hp.com>
	* release 3.0 

2003-11-18 Stephane Eranian <eranian@hpl.hp.com>
	* removed the --btb-all-mispredicted option for Itanium2.
	  it was confusing and could not be achieve in one run.

2003-11-12 Stephane Eranian <eranian@hpl.hp.com>
	* fix --btb-all-mispredicted brt setting bug

2003-11-12 Stephane Eranian <eranian@hpl.hp.com>
	* cleanup 64bits prints

2003-11-11 Stephane Eranian <eranian@hpl.hp.com>
	* fix race condition in pfmon_task.c:task_create.c wrt sigchild
	  and waitpid.

2003-11-05 Stephane Eranian <eranian@hpl.hp.com>
	* fixed broken --cpu-list option
	* reinforce parameter check for all options with parameters
	* fixed --exec-split handling in pfmon_task.c
	* updated pfmon_usersguide.txt, pfmon_itanium.txt, pfmon_itanium2.txt
	* fixed -outfile and --smpl-outfile to use the same naming convention
	* added missing call to print_smpl_header() is case of aggregation and sampling
	* fixed a bug in which no samples were reported in system-wide aggregation mode
	* fixed bugs in print_standard_header() for system-wide sessions
	* cosmetic changes to some prints in verbose mode
	* fixed a missing len+1 string bug in load_kallsyms_symbol() 
	* cleaned some bogus return values in pfmon_utils range_*() functions
	* added default size to options print in some sampling modules
	* don't put trailing spaces after event names when not needed (helps i2prof.pl)

2003-10-29 Stephane Eranian <eranian@hpl.hp.com>
	* release beta of 3.0 to selected users
	* changelog has not been updated!

2002-12-06 Stephane Eranian <eranian@hpl.hp.com>
	* fix DESTDIR because the wrong final root
	  was used when looking for pfmon.conf. Added DESTROOT

2002-12-05 Stephane Eranian <eranian@hpl.hp.com>
	* release 2.0 beta for testing

2002-12-04 Stephane Eranian <eranian@hpl.hp.com>
	* updated entire source code to reflect change in the 
	  library for pfmlib_param_t and pfm_dispatch_events().

2002-12-03 Stephane Eranian <eranian@hpl.hp.com>
	* cleanup management of per-event instruction set in pfmon_itanium.c
	  and pfmon_itanium2.c.
	* got rid of gen_priv_levels() in pfmon_util.c, consolidated in
          pfmon.c
	* fix a bug in the drange plm setting in both pfmon_itanium.c
	  and pfmon_itanium2.c
	* added restriction that we cannot mix per-event priv level
	  with any range restrictions.
	* updated pfmon_itaniu.txt range restriction section
	* added missing range restriction section in pfmon_itanium2.txt

2002-11-27 Stephane Eranian <eranian@hpl.hp.com>
	* cleanup the sampling output formats, especially compact
	* simplified management of entry number when sampling.

2002-11-21 Stephane Eranian <eranian@hpl.hp.com>
	* added --smpl-print-counts option. By default counts are not
	  printed when sampling unless this option is specified.
	* added warning in compact sampling output mode  header when
	  randomization is not supported by kernel. We keep the same 
	  number of columns but the value is zero instead of the initial
	  value of first overflowed PMD.
	* added --trigger-start-delay option to both per-process and
	  system-wide modes.

2002-11-20 Stephane Eranian <eranian@hpl.hp.com>
	* separated the library, include files, and examples
	  from pfmon.

2002-11-19 Stephane Eranian <eranian@hpl.hp.com>
	* added support for monitoring at priv level 1 2
	* added support for sampling period randomization (--smpl-periods-random)
	* fix to forbid using a command and timeout to delimit system-wide session
	* fixed signal/ptrace problem when using --trigger-address option in
	  per-process mode. We must resend all signals back to the monitored 
	  task unless it is SIGTRAP.

2002-11-14 Stephane Eranian <eranian@hpl.hp.com>
	* cleanup the DPRINT/debug_print macros. Added dprintf() to pfmon_util.c

2002-11-11 Stephane Eranian <eranian@hpl.hp.com>
	* fixed a sampling buffer bug in which the last set of samples
	  may be reported twice for non-blocking (default) per-process
	  sampling sessions. This is a kernel perfmon buf in 2.4.18 and
	  2.4.19. Need a way to identify each set of samples.

2002-09-25 Stephane Eranian <eranian@hpl.hp.com>
	* fixed invalid order of parameters for warning() all sampling format
	  validate routines.

2002-09-16 Stephane Eranian <eranian@hpl.hp.com>
	* added generation counter to barrier structure in pfmon/pfmon_system.c
	  to avoid possible wake-up race condition.

2002-08-14 Stephane Eranian <eranian@hpl.hp.com>
	* fixed a problem when using BTB on Itanium2, it would not setup
	  pmc12 correctly
	* fixed btb_smpl.c to do the bundle address adjustment on Itanium2

2002-07-03 Stephane Eranian <eranian@hpl.hp.com>
	* release version 1.1 which includes full Itanium2 support

2002-03-01 Stephane Eranian <eranian@hpl.hp.com>
	* added + symbol to getopt_long() in pfmon.c
	* fixed missing -v option (verbose) support 

2002-01-22 Stephane Eranian <eranian@hpl.hp.com>
	* release version 1.0
	* system wide support for UP and SMP
	* Itanium address range restriction support
	* possibility to sample on any events (not just Itanium EAR/BTB)
	* lots of new command line options
	* a new set of examples  and simple test programs
	* a new compact sampling format
	* completely restructured library code: prepare for other CPU models
	* completely restructured pfmon code: prepare for other CPU models
	* a user's guide (1000 lines)
	* updated event names for Itanium (update to latest specs)
	* support for Intel ecc compiler (beta6.0) 
	* support for IA-32 execution monitoring

2001-08-06 Stephane Eranian <eranian@hpl.hp.com>
	* release 0.06a

	* added more explicit message when kernel does not support
	  perfmon (CONFIG_PERFMON)

	* fixed libpfm Makefile for missing . in front of the
	  python script.

	* reverted a bogus patch introduce in a temporary 0.06 release
	  of pfmon. You must update from pfmon-0.06 if downloaded after
	 June 5th 2001.

	* fixed missing "h" in the getopt() call.

	* added shortcut "-V" for version

2001-01-29 Stephane Eranian <eranian@hpl.hp.com>

	* added minimal README

	* release 0.06 to Trillian members

2001-01-25 Stephane Eranian <eranian@hpl.hp.com>

	* fixed the L3_MISSES event. The code was incorrect. We were
	  counting the number of loads retired.

	* cleaned the perfmon.database: unified name of some events

	* remove some of the extra printf. Converted into debug or verbose
	  mode prints.

	* release 0.05b
