|
Packit |
577717 |
2011-05-10
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/Rules.pfm_pe: The --with-bitmode parameter was not being
|
|
Packit |
577717 |
passed along to libpfm3, so it was not possible to build
|
|
Packit |
577717 |
perf_event PAPI on non-default bitmodes. This change passes
|
|
Packit |
577717 |
along the $(BITFLAGS) value to the libpfm3 make invocation.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_pfm_events.c, papi_pfm_events.h, perf_events.c: The
|
|
Packit |
577717 |
perf_events code was using __u64 instead of uint64_t and this was
|
|
Packit |
577717 |
causing a warning when compiling for 64-bit Power.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/libpfm-3.y/lib/amd64_events_fam15h.h: Added Robert Richter's
|
|
Packit |
577717 |
patch with a few new events for AMD Family 15h.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-05-06
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* INSTALL.txt: Load the 'gcc' module not 'gnu' module for Cray.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* INSTALL.txt: Update the install instructions for Cray XT and XE
|
|
Packit |
577717 |
systems.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/: multiattach.c, multiattach2.c: Make the multiattach
|
|
Packit |
577717 |
and multiattach2 failures into warnings.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
I have a proposed fix that makes the failures go away, but it has
|
|
Packit |
577717 |
not been tested much and also causes some new fcntl() error
|
|
Packit |
577717 |
messages under perfctr.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
So temporarily make the tests only warn for the release and I'll
|
|
Packit |
577717 |
work on a proper fix for after. The behavior in these tests has
|
|
Packit |
577717 |
been broken for a long time so it is not a recent regression.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_memory.c: Band-aid for the leak debugging statement in
|
|
Packit |
577717 |
papi_memory.c on NO_VARARG_MACRO systems. (aix currently)
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-05-05
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/multiattach.c: Had the division backwards on the
|
|
Packit |
577717 |
validation.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/multiattach.c: Update the multiattach test to fail if
|
|
Packit |
577717 |
the results aren't in the proper ratio. This was failing on
|
|
Packit |
577717 |
perf_event kernels but since the results weren't checked it was
|
|
Packit |
577717 |
never reported as an error.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* delete_before_release.sh: delete cvs2cl.pl before release
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* ChangeLogP413.txt: First cut change log for the 4.1.3 release.
|
|
Packit |
577717 |
Nothing's frozen yet...
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* cvs2cl.pl: Perl script to generate change logs. Keeping it with
|
|
Packit |
577717 |
the project makes life easier.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* INSTALL.txt: Change INSTALL to reflect that we support power7.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/Makefile.in, src/configure, src/configure.in, src/papi.h,
|
|
Packit |
577717 |
doc/Doxyfile, doc/Doxyfile-everything, papi.spec: Modfy version
|
|
Packit |
577717 |
number for pending release: 4.1.3.0
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-05-03
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_internal.c, papi_internal.h, sys_perf_event_open.c,
|
|
Packit |
577717 |
ctests/attach2.c: Cleanup the _papi_hwi_cleanup_eventset()
|
|
Packit |
577717 |
function in papi_internal.c
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This function was re-using existing functionality to remove one
|
|
Packit |
577717 |
event at a time before cleaning out the eventset. This is not
|
|
Packit |
577717 |
strictly necessary and was breaking on perf_event eventsets that
|
|
Packit |
577717 |
were attached to finished processes, as a call to
|
|
Packit |
577717 |
update_control_state() would close/reopen the perf_event fd,
|
|
Packit |
577717 |
failing when the finished process went away after the close.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
The new code removes all events from the eventset in one go
|
|
Packit |
577717 |
before calling update_control_state.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
The change here also updates code comments as necessary, as some
|
|
Packit |
577717 |
of the code in papi_internal.c can be a bit obscure.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
It also updates some of the comments in ctests/attach2.c to give
|
|
Packit |
577717 |
better debugging info.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-28
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/threads.c: Uncomment the actual signal passing functionality
|
|
Packit |
577717 |
in _papi_hwi_broadcast_signal
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_debug.h: Include files added to papi_debug.h
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/README: Added detailed instructions on how to
|
|
Packit |
577717 |
build PAPI with the CUDA component
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-27
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/threads.c: Move an escape test to the outer loop in
|
|
Packit |
577717 |
_papi_hwi_broadcast_signal.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This cleans up an infinite loop where before we would only break
|
|
Packit |
577717 |
out of the component look, not the thread list walking loop.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi.c, papi_internal.c, papi_internal.h, papi_protos.h:
|
|
Packit |
577717 |
Clean up papi_internal.c so that functions not used outside are
|
|
Packit |
577717 |
marked static.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_pfm_events.c, papi_preset.c, pmapi-ppc64_events.c:
|
|
Packit |
577717 |
papi: Fix some memory leaks
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Signed-off-by: Robert Richter <robert.richter@amd.com>
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perf_events.c: papi: Make functions and variables static in
|
|
Packit |
577717 |
perf_events.c
|
|
Packit |
577717 |
|
|
Packit |
577717 |
All this functions and variables are not used outside
|
|
Packit |
577717 |
perf_events.c. Making them static.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Signed-off-by: Robert Richter <robert.richter@amd.com>
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_pfm_events.c: papi: Fix crash in error handler for
|
|
Packit |
577717 |
pfm_get_event_code_counter()
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Signed-off-by: Robert Richter <robert.richter@amd.com>
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/utils/native_avail.c: papi: Fix error check in native_avail.c
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Signed-off-by: Robert Richter <robert.richter@amd.com>
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-26
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/libpfm-3.y/: include/perfmon/pfmlib_amd64.h,
|
|
Packit |
577717 |
lib/pfmlib_amd64.c: AMD architectural PMU could not be detected
|
|
Packit |
577717 |
for family 15h as there was a strict check for AMD family 10h.
|
|
Packit |
577717 |
Enabling it now for all families from 10h.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Signed-off-by: Robert Richter
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/libpfm-3.y/lib/amd64_events_fam15h.h: There is no kernel
|
|
Packit |
577717 |
support for AMD family 15h northbridge events, disabling them in
|
|
Packit |
577717 |
libpfm3 to not report them as available native events.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Robert Richter
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in, linux-common.c: Add some extra
|
|
Packit |
577717 |
debug messages for better tracking of the --with-assumed-kernel
|
|
Packit |
577717 |
configure option.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-25
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in, linux-common.c: Add a new
|
|
Packit |
577717 |
configure option: --with-assumed-kernel=<ver> This allows you
|
|
Packit |
577717 |
to specify a kernel revision to (instead of being autodetected
|
|
Packit |
577717 |
with uname) for perf_event workaround purposes. With this you
|
|
Packit |
577717 |
can force PAPI to not use workarounds on kernels with
|
|
Packit |
577717 |
backported versions of perf_event features.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-19
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: Makefile.inc, configure, configure.in, papi_debug.h,
|
|
Packit |
577717 |
papi_internal.h, sys_perf_event_open.c: Add debugging to
|
|
Packit |
577717 |
sys_perf_event_open.c to show exactly what values are being
|
|
Packit |
577717 |
passed to the perf_event_open syscall.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-18
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: run_tests.sh, ctests/attach2.c, ctests/attach3.c: Fix for
|
|
Packit |
577717 |
finding attach_target with execlp to search the path.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-14
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: Rules.pfm, configure, configure.in, linux-ia64-pfm.h,
|
|
Packit |
577717 |
linux-ia64.c, linux-ia64.h, perfmon-ia64-pfm.h, perfmon-ia64.c,
|
|
Packit |
577717 |
perfmon-ia64.h, perfmon.h: Rename the linux-ia64-* files to be
|
|
Packit |
577717 |
called perfmon-ia64-*
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This is a more descriptive name, and makes it more obvious what
|
|
Packit |
577717 |
the files are for.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/libpfm-3.y/: include/perfmon/pfmlib_amd64.h,
|
|
Packit |
577717 |
lib/pfmlib_amd64.c, lib/pfmlib_amd64_priv.h: Patch to have
|
|
Packit |
577717 |
libpfm3 use 6 counters on Interlagos.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch provided by Robert Richter
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/linux-memory.c: Fix the POWER cache detection routines to
|
|
Packit |
577717 |
work properly on POWER7.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch provided by Corey Ashford
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Have configure check for ifort if
|
|
Packit |
577717 |
gfortran, etc, not found.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch by Gary Mohr
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/johnmay2.c: Update the validation message on the
|
|
Packit |
577717 |
ctests/johnmay2.c test to be less confusing. Also add some
|
|
Packit |
577717 |
comments to the source code.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Problem reported by Steve Kaufmann.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-13
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/: multiattach2, multiattach2.c: Remove the
|
|
Packit |
577717 |
accidentally added ctests/multiattach2 and add instead the proper
|
|
Packit |
577717 |
ctests/multiattach2.c
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/Makefile.inc: components_config.h is cleaned out with make
|
|
Packit |
577717 |
clobber, not make clean this should fix the build bot issues.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/: Makefile, attach3.c, multiattach.c, multiattach2,
|
|
Packit |
577717 |
zero_attach.c: Minor typos in comments. Discovered another bug in
|
|
Packit |
577717 |
attach code demonstrated by multiattach2. You cannot have an
|
|
Packit |
577717 |
eventset running that is self counting as well as one that is
|
|
Packit |
577717 |
attached. PAPI thinks that both are running and throws an error.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perf_events.c: We must update the control state after
|
|
Packit |
577717 |
attaching for perf_events, zero_attach now passes
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/: Makefile, attach2.c, attach3.c, do_loops.c: This
|
|
Packit |
577717 |
commit adds testing of attaching to fork/exec'd executables.
|
|
Packit |
577717 |
zero_attach and multiattach just test forks. This also modifies
|
|
Packit |
577717 |
do_loops.c to be able to generate a test driver when
|
|
Packit |
577717 |
-DDUMMY_DRIVER is defined so we can use it to generate flops as a
|
|
Packit |
577717 |
sub process.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Attach2 and attach3 have one important difference.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Attach3 does a 'assign component' before attaching and then
|
|
Packit |
577717 |
adding events. Attach2 does not assign a component and thus
|
|
Packit |
577717 |
should inherit the default component.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
The current bug in PAPI is that: * The default component is not
|
|
Packit |
577717 |
assigned until you add an event. * However, attaching an
|
|
Packit |
577717 |
eventset without events is perfectly valid, but we get an error.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Possible solution is that the default component should be
|
|
Packit |
577717 |
assigned at create time.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-12
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/multiattach.c: Make sure the two processes compute
|
|
Packit |
577717 |
different numbers of flops to test attach
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-05
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/power7_events.h: Turns out Maynard Johnson answered my
|
|
Packit |
577717 |
questions about the native_name enum back in December. ( this is
|
|
Packit |
577717 |
a correct version of the events file )
|
|
Packit |
577717 |
|
|
Packit |
577717 |
As I found out, the AIX substrates do not use the native_name
|
|
Packit |
577717 |
enum. But a hypothetical perfctr build would.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-04
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/Makefile.inc: Clear out the components_config.h file on make
|
|
Packit |
577717 |
clobber
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: aix.c, power7_events.h: Initial support for power7 aix, the
|
|
Packit |
577717 |
events file is a copy of power6_events.h with the number of
|
|
Packit |
577717 |
groups changed. The native_name enum is unchanged, but unused?
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-04-01
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/configure.in: Commited wrong configure.in
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Clean up setting bitmode flags for
|
|
Packit |
577717 |
non-gcc (xlc in this case) compilers.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_events.csv: Change the Nehalem PAPI_FP_OPS event from
|
|
Packit |
577717 |
FP_COMP_OPS_EXE:SSE_SINGLE_PRECISION+FP_COMP_OPS_EXE:DOUBLE_PRECISION
|
|
Packit |
577717 |
to FP_COMP_OPS_EXE:SSE+FP_COMP_OPS_EXE:X87
|
|
Packit |
577717 |
|
|
Packit |
577717 |
The new event gives the same results as the previous one, with
|
|
Packit |
577717 |
the added benefit of also counting 32-bit compiled x87 fp ops
|
|
Packit |
577717 |
properly.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
More detailed analysis can be found here:
|
|
Packit |
577717 |
http://web.eecs.utk.edu/~vweaver1/projects/nehalem-fp_ops/
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-03-28
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/utils/multiplex_cost.c: Turns out that getopt_long isn't as
|
|
Packit |
577717 |
standard as I had hoped.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Convert multiplex_cost to use only getopt. -s disables software
|
|
Packit |
577717 |
multiplexing -k disables kernel multiplexing
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-03-25
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, utils/Makefile, utils/multiplex_cost.c,
|
|
Packit |
577717 |
configure.in: Multiplex_cost utility.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/utils/: Makefile, cost.c, cost_utils.c, cost_utils.h: Split
|
|
Packit |
577717 |
off the statistics functions from cost.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-03-22
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: run_tests_exclude_cuda.txt, run_tests.sh: Exclude some
|
|
Packit |
577717 |
fork/thread tests from fulltest that won't run with CUDA (reason:
|
|
Packit |
577717 |
cannot invoke same GPU from different threads)
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-03-21
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/utils/cost.c: Add a test for DERIVED_[ADD | SUB ] events to
|
|
Packit |
577717 |
papi_cost.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-03-18
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/cuda/linux-cuda.c: all_native_events ctest failed
|
|
Packit |
577717 |
when CUDA Component is used. Reason: removing cuda events from
|
|
Packit |
577717 |
the eventset is currently not supported. According to the NVIDIA
|
|
Packit |
577717 |
folks this is a bug in cuda 4.0rc and will be fixed in rc2. Note
|
|
Packit |
577717 |
also, several fork and thread tests fail since it's illegal to
|
|
Packit |
577717 |
invoke the same GPU device from different processes / threads. We
|
|
Packit |
577717 |
need a mechanism that allows us to run tests for the CPU
|
|
Packit |
577717 |
component only.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-03-15
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/utils/cost.c: Add a test case to cost util, look for a
|
|
Packit |
577717 |
derived-postfix event and if found, give timing information for
|
|
Packit |
577717 |
read calls to it.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This is just a first run at the test, Core2 and AMD have
|
|
Packit |
577717 |
candidate events and the test runs, but that is the extent of my
|
|
Packit |
577717 |
testing so far.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-03-11
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/: README, cuda/Makefile.cuda.in, cuda/Rules.cuda,
|
|
Packit |
577717 |
cuda/configure, cuda/configure.in, cuda/linux-cuda.c,
|
|
Packit |
577717 |
cuda/linux-cuda.h: Added CUDA component, a hardware performance
|
|
Packit |
577717 |
counter measurement technology for the NVIDIA CUDA platform which
|
|
Packit |
577717 |
provides access to the hardware counters inside the GPU. PAPI
|
|
Packit |
577717 |
CUDA is based on CUPTI support - shipped with CUDA 4.0rc - in the
|
|
Packit |
577717 |
NVIDIA driver library. In any environment where the CUPTI-enabled
|
|
Packit |
577717 |
driver is installed, the PAPI CUDA component can provide detailed
|
|
Packit |
577717 |
performance counter information regarding the execution of GPU
|
|
Packit |
577717 |
kernels.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/: coretemp/linux-coretemp.c,
|
|
Packit |
577717 |
lustre/linux-lustre.c: Add some missing includes to components.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Thanks to Will Cohen for reminding us warnings matter. :)
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in, perf_events.c: The SYNC_READ
|
|
Packit |
577717 |
workaround in perf_events.c was being handled at compile time,
|
|
Packit |
577717 |
rather than at run-time like all of our other workarounds.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Change it to be like our other kernel-version related
|
|
Packit |
577717 |
workarounds.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-03-09
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/multiplex1_pthreads.c: Between 4.0.0 and 4.1.0 a
|
|
Packit |
577717 |
pthread_exit() call was added to ctest/multiplex1_pthreads.c that
|
|
Packit |
577717 |
caused the test to exit partway through the test and without
|
|
Packit |
577717 |
doing a proper PASS/FAIL result.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This changeset backs out that change, though the original change
|
|
Packit |
577717 |
was marked as a memory leak fix so a different fix may be needed.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Reported by Steve Kaufmann
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/linux-timer.c: Add missing header needed by
|
|
Packit |
577717 |
--with-virtualtimer=times build.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Reported by Steve Kaufmann
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-03-01
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_pfm_events.c, perf_events.c: Fix broken Linux/PPC
|
|
Packit |
577717 |
build caused by my pfm_events code movement changes.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-25
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_pfm_events.c, papi_pfm_events.h, perfctr-x86.h: My
|
|
Packit |
577717 |
changes yesterday broke the perfctr build. This should fix it.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-24
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/inherit.c: Make the inherit test respect TESTS_QUIET
|
|
Packit |
577717 |
so that it does not print extra output during a run_tests.sh run
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/overflow.c: Fix missing newline in the overflow
|
|
Packit |
577717 |
output.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Reported by Gary Mohr
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_pfm_events.c, papi_pfm_events.h, perf_events.c: Move
|
|
Packit |
577717 |
the libpfm3 specific functions from perf_events.c into
|
|
Packit |
577717 |
papi_pfm_events.c
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perf_events.c: Separate the libpfm3-specific code from
|
|
Packit |
577717 |
_papi_pe_init_substrate() and _papi_pe_update_control_state()
|
|
Packit |
577717 |
into their own functions. This will allow eventual code sharing
|
|
Packit |
577717 |
and also make the libpfm4 merge easier.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perf_events.c: Some minor cleanups I found after reviewing
|
|
Packit |
577717 |
the inherit merge. + Add missing "static inline" to the new
|
|
Packit |
577717 |
kernel-version codes + Remove duplicated test for Pentium 4
|
|
Packit |
577717 |
+ Fix a warning only seen if --with-debug is enabled
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi.c, papi.h, papi_internal.h, perf_events.c,
|
|
Packit |
577717 |
perf_events.h, ctests/Makefile, ctests/inherit.c,
|
|
Packit |
577717 |
ctests/test_utils.c: Merging Gary Mohr's re-implementation of
|
|
Packit |
577717 |
inherit into the code base. Thanks, Gary!
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-23
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: any-null.h, freebsd.h, linux-bgp.h, linux-common.c,
|
|
Packit |
577717 |
linux-common.h, linux-context.h, linux-ia64.c, linux-ia64.h,
|
|
Packit |
577717 |
linux-lock.h, linux-memory.c, linux-ppc64.h, linux-timer.c,
|
|
Packit |
577717 |
papi_internal.h, papi_pfm_events.c, perf_events.c, perf_events.h,
|
|
Packit |
577717 |
perfctr-x86.h, perfctr.c, perfmon.h, solaris-niagara2.h,
|
|
Packit |
577717 |
solaris-ultra.h, solaris.h, x86_cache_info.c: Move some more
|
|
Packit |
577717 |
duplicated OS common code (in this case the locking code and the
|
|
Packit |
577717 |
context accessing code) out of the various substrate include
|
|
Packit |
577717 |
files and into a common location.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-22
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perf_events.c: Separate out the kernel-version dependent
|
|
Packit |
577717 |
checks and group them together near the beginning of the code.
|
|
Packit |
577717 |
This not only allows us to easily see which routines are
|
|
Packit |
577717 |
kernel-version dependent, but it makes it easier to disable the
|
|
Packit |
577717 |
checks one-by-one when debugging kernel-version related issues
|
|
Packit |
577717 |
like those found with the inherit patches.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-21
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_internal.c: Extend _papi_hwi_cleanup_eventset to free
|
|
Packit |
577717 |
memory and better cleanup after us.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-18
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi.c: PAPI_assign_eventset_component changed; refuses to
|
|
Packit |
577717 |
reassign components.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-17
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_events.csv, libpfm-3.y/include/perfmon/pfmlib.h,
|
|
Packit |
577717 |
libpfm-3.y/lib/amd64_events.h,
|
|
Packit |
577717 |
libpfm-3.y/lib/amd64_events_fam10h.h,
|
|
Packit |
577717 |
libpfm-3.y/lib/amd64_events_fam15h.h,
|
|
Packit |
577717 |
libpfm-3.y/lib/pfmlib_amd64.c,
|
|
Packit |
577717 |
libpfm-3.y/lib/pfmlib_amd64_priv.h: Add support for AMD Family
|
|
Packit |
577717 |
15h processors. Also adds suport for Family 10h RevE
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patches provided by Robert Richter
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/utils/native_avail.c: Modify papi_native_avail to properly
|
|
Packit |
577717 |
handle event names with libpfm4-style "::" separators in them.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-15
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/Makefile.inc: make install-doxyman will build/install the
|
|
Packit |
577717 |
doxygen version of the manpages.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Note that these pages are very rough right now, much work is
|
|
Packit |
577717 |
needed to get them to be a drop in replacement for the current
|
|
Packit |
577717 |
man pages. (mostly formatting related/use related issues, eg man
|
|
Packit |
577717 |
PAPI_start will not work yet; the content is there.)
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* doc/Makefile: Add install target for doxygen generated man pages.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-11
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: perfctr-x86.c, perfctr.c: perfctr-2.6.42 introduced
|
|
Packit |
577717 |
PERFCTR_X86_INTEL_WSTMR PAPI added support for
|
|
Packit |
577717 |
PERFCTR_X86_INTEL_WSMR notice the missing T
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Fix PAPI to use the proper define. This should fix Westmere
|
|
Packit |
577717 |
support on perfctr kernels.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-09
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_protos.h, papi_vector.c, papi_vector.h,
|
|
Packit |
577717 |
papi_vector_redefine.h: Added function pointer destroy_eventset
|
|
Packit |
577717 |
to the PAPI vector table. Needed for the CUDA Component to
|
|
Packit |
577717 |
disable CUDA eventGroups, to destroy floating CUDA context, and
|
|
Packit |
577717 |
to free perfmon hardware on the GPU. (Note: the CUDA Component
|
|
Packit |
577717 |
cannot be released yet since we are still under NDA with NVIDIA.
|
|
Packit |
577717 |
Stay tuned.)
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-07
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/x86_cache_info.c: The cpuid leaf2 code was printing a message
|
|
Packit |
577717 |
to stderr if leaf4 was needed (only happens on Westmere
|
|
Packit |
577717 |
currently). Change this to be a MEMDBG() debug message instead.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-02-03
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_events.csv, perfctr-x86.c: perfctr-x86 was reporting
|
|
Packit |
577717 |
"Core i7" instead of "Nehalem". i7 can mean Westmere or Sandy
|
|
Packit |
577717 |
Bridge too, so change the code to properly report Nehalem.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-01-27
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/all_native_events.c:
|
|
Packit |
577717 |
Fix this ctest. It failed when the package was built with several
|
|
Packit |
577717 |
components because the eventset was reused and failed to add
|
|
Packit |
577717 |
events that were not from the first component.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
In order to fix it, I recreate & destroy the eventset when the
|
|
Packit |
577717 |
current event does not belong to the previous component.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-01-26
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in, linux-timer.c, perfmon.c: Fix Cray
|
|
Packit |
577717 |
CLE build.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Putting -Wall in cflags now
|
|
Packit |
577717 |
requires CC = gcc
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: aix.c, freebsd.c, linux-bgp.c, linux-common.c,
|
|
Packit |
577717 |
linux-memory.c, linux-memory.h, papi.c, papi_protos.h,
|
|
Packit |
577717 |
papi_vector.c, papi_vector.h, solaris-niagara2.c,
|
|
Packit |
577717 |
solaris-ultra.c, windows-common.c, windows-memory.c: Change the
|
|
Packit |
577717 |
paramaters passed to update_shlib_info() to match better with
|
|
Packit |
577717 |
those passed to get_system_info(). This only affects the
|
|
Packit |
577717 |
substrates, outside users of PAPI will not notice this change.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-01-25
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Make sure that aix gets -g.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Give everyone else -g when
|
|
Packit |
577717 |
configuring with debug.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
To wit, we pass gcc -g3 but neglected platforms where CC!=gcc.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/aix.c: First run at supporting power7. NOTE: this code is
|
|
Packit |
577717 |
only good for getting event listings eg papi_native_avail,
|
|
Packit |
577717 |
|
|
Packit |
577717 |
passing PM_GET_GROUPS causes our code to segfault later on, a
|
|
Packit |
577717 |
buffer overflow I'm still tracking down.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perfctr-x86.c: Accidentally converted a function to _perfctr_
|
|
Packit |
577717 |
that should have stayed _linux_.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: perfctr-x86.c, perfctr.c: Rename the various perfctr
|
|
Packit |
577717 |
functions to be _perfctr_ rather than _linux_. This way _linux_
|
|
Packit |
577717 |
is reserved for the common functions used by all.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: linux-common.c, linux-memory.c, linux-timer.c,
|
|
Packit |
577717 |
perf_events.c, windows-common.c, windows-memory.c,
|
|
Packit |
577717 |
windows-timer.c: Split the WIN32 specific code out from the new
|
|
Packit |
577717 |
linux common code.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
In most cases very little code was shared (it tended to be a big
|
|
Packit |
577717 |
#ifdef block) and it is confusing to have windows-specific code
|
|
Packit |
577717 |
in files named linux-*
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-01-24
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/linux-timer.c: Fix a compile error that only shows up on PPC.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/linux-timer.c: Fix compile warning if mmtimer is enabled.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perfctr-x86.c: Missing comma in the perfctr code.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: Makefile.inc, aix.c, configure, configure.in,
|
|
Packit |
577717 |
hwinfo_linux.c, linux-bgp.c, linux-common.c, linux-common.h,
|
|
Packit |
577717 |
linux-ia64.c, linux-timer.c, linux-timer.h, papi_vector.h,
|
|
Packit |
577717 |
perf_events.c, perfctr-x86.c, perfctr.c, perfmon.c,
|
|
Packit |
577717 |
solaris-niagara2.c, solaris-ultra.c: One last batch of
|
|
Packit |
577717 |
consolidation changes.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This one moves get_system_info and get_cpu_info into
|
|
Packit |
577717 |
linux-common.c, plus moves some other routines from perf_events.c
|
|
Packit |
577717 |
there that are shared by the future libpfm4 version.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Some non-linux substrates are touched here; these are just short
|
|
Packit |
577717 |
fixes to make sure the get_system_info() function pointed to by
|
|
Packit |
577717 |
the papi_vector has the same format on all substrates.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: Makefile.inc, configure, configure.in, linux-memory.c,
|
|
Packit |
577717 |
linux-memory.h, perf_events.c, perfctr-x86.c, perfctr.c,
|
|
Packit |
577717 |
perfmon.c: Move the various Linux update_shlib_info() functions
|
|
Packit |
577717 |
into a common place.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: Makefile.inc, linux-timer.c, linux-timer.h, perf_events.c,
|
|
Packit |
577717 |
perfctr-x86.c, perfctr.c, perfmon.c: Move the various
|
|
Packit |
577717 |
timer-related functions to linux-timer.c This gets rid of the
|
|
Packit |
577717 |
duplicated code spread throughout the substrates.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-01-21
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* delete_before_release.sh, release_procedure.txt: Updated the
|
|
Packit |
577717 |
release docs with what I learned when making the 4.1.2.1 release.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in, freebsd-memory.c,
|
|
Packit |
577717 |
linux-ia64-memory.c, linux-memory.c, linux-memory.h,
|
|
Packit |
577717 |
linux-mx-memory.c, linux-ppc64-memory.c, perf_events.c,
|
|
Packit |
577717 |
perfctr-x86.c, perfmon-memory.c, perfmon.c: Currently there are
|
|
Packit |
577717 |
at least 3 identical copies of the linux memory detection code
|
|
Packit |
577717 |
spread throughout the PAPI source code.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This change puts them all in linux-memory.c, and then has all the
|
|
Packit |
577717 |
individual substrates use the common code.
|