|
Packit |
577717 |
2011-10-25
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* doc/: Makefile, doxygen_procedure.txt: Update doxygen_procedure
|
|
Packit |
577717 |
to note that we need a recent version of doxygen.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* man/: man1/avail.c.1, man1/clockres.c.1, man1/command_flags_t.1,
|
|
Packit |
577717 |
man1/command_line.c.1, man1/component.c.1, man1/cost.c.1,
|
|
Packit |
577717 |
man1/decode.c.1, man1/error_codes.c.1, man1/event_chooser.c.1,
|
|
Packit |
577717 |
man1/mem_info.c.1, man1/native_avail.c.1, man1/options_t.1,
|
|
Packit |
577717 |
man1/papi_avail.1, man1/papi_clockres.1,
|
|
Packit |
577717 |
man1/papi_command_line.1, man1/papi_component_avail.1,
|
|
Packit |
577717 |
man1/papi_cost.1, man1/papi_decode.1, man1/papi_error_codes.1,
|
|
Packit |
577717 |
man1/papi_event_chooser.1, man1/papi_mem_info.1,
|
|
Packit |
577717 |
man1/papi_multiplex_cost.1, man1/papi_native_avail.1, man3/CDI.3,
|
|
Packit |
577717 |
man3/HighLevelInfo.3, man3/PAPIF.3, man3/PAPIF_accum.3,
|
|
Packit |
577717 |
man3/PAPIF_accum_counters.3, man3/PAPIF_add_event.3,
|
|
Packit |
577717 |
man3/PAPIF_add_events.3, man3/PAPIF_assign_eventset_component.3,
|
|
Packit |
577717 |
man3/PAPIF_cleanup_eventset.3, man3/PAPIF_create_eventset.3,
|
|
Packit |
577717 |
man3/PAPIF_destroy_eventset.3, man3/PAPIF_enum_event.3,
|
|
Packit |
577717 |
man3/PAPIF_event_code_to_name.3, man3/PAPIF_event_name_to_code.3,
|
|
Packit |
577717 |
man3/PAPIF_flips.3, man3/PAPIF_flops.3,
|
|
Packit |
577717 |
man3/PAPIF_get_clockrate.3, man3/PAPIF_get_dmem_info.3,
|
|
Packit |
577717 |
man3/PAPIF_get_domain.3, man3/PAPIF_get_event_info.3,
|
|
Packit |
577717 |
man3/PAPIF_get_exe_info.3, man3/PAPIF_get_granularity.3,
|
|
Packit |
577717 |
man3/PAPIF_get_hardware_info.3, man3/PAPIF_get_multiplex.3,
|
|
Packit |
577717 |
man3/PAPIF_get_preload.3, man3/PAPIF_get_real_cyc.3,
|
|
Packit |
577717 |
man3/PAPIF_get_real_nsec.3, man3/PAPIF_get_real_usec.3,
|
|
Packit |
577717 |
man3/PAPIF_get_virt_cyc.3, man3/PAPIF_get_virt_usec.3,
|
|
Packit |
577717 |
man3/PAPIF_ipc.3, man3/PAPIF_is_initialized.3,
|
|
Packit |
577717 |
man3/PAPIF_library_init.3, man3/PAPIF_lock.3,
|
|
Packit |
577717 |
man3/PAPIF_multiplex_init.3, man3/PAPIF_num_cmp_hwctrs.3,
|
|
Packit |
577717 |
man3/PAPIF_num_counters.3, man3/PAPIF_num_events.3,
|
|
Packit |
577717 |
man3/PAPIF_num_hwctrs.3, man3/PAPIF_perror.3,
|
|
Packit |
577717 |
man3/PAPIF_query_event.3, man3/PAPIF_read.3,
|
|
Packit |
577717 |
man3/PAPIF_read_ts.3, man3/PAPIF_register_thread.3,
|
|
Packit |
577717 |
man3/PAPIF_remove_event.3, man3/PAPIF_remove_events.3,
|
|
Packit |
577717 |
man3/PAPIF_reset.3, man3/PAPIF_set_cmp_domain.3,
|
|
Packit |
577717 |
man3/PAPIF_set_cmp_granularity.3, man3/PAPIF_set_debug.3,
|
|
Packit |
577717 |
man3/PAPIF_set_domain.3, man3/PAPIF_set_event_domain.3,
|
|
Packit |
577717 |
man3/PAPIF_set_granularity.3, man3/PAPIF_set_inherit.3,
|
|
Packit |
577717 |
man3/PAPIF_set_multiplex.3, man3/PAPIF_shutdown.3,
|
|
Packit |
577717 |
man3/PAPIF_start.3, man3/PAPIF_start_counters.3,
|
|
Packit |
577717 |
man3/PAPIF_state.3, man3/PAPIF_stop.3,
|
|
Packit |
577717 |
man3/PAPIF_stop_counters.3, man3/PAPIF_thread_id.3,
|
|
Packit |
577717 |
man3/PAPIF_thread_init.3, man3/PAPIF_unlock.3,
|
|
Packit |
577717 |
man3/PAPIF_unregister_thread.3, man3/PAPIF_write.3,
|
|
Packit |
577717 |
man3/PAPI_accum.3, man3/PAPI_accum_counters.3,
|
|
Packit |
577717 |
man3/PAPI_add_event.3, man3/PAPI_add_events.3,
|
|
Packit |
577717 |
man3/PAPI_addr_range_option_t.3, man3/PAPI_address_map_t.3,
|
|
Packit |
577717 |
man3/PAPI_all_thr_spec_t.3,
|
|
Packit |
577717 |
man3/PAPI_assign_eventset_component.3, man3/PAPI_attach.3,
|
|
Packit |
577717 |
man3/PAPI_attach_option_t.3, man3/PAPI_cleanup_eventset.3,
|
|
Packit |
577717 |
man3/PAPI_component_info_t.3, man3/PAPI_cpu_option_t.3,
|
|
Packit |
577717 |
man3/PAPI_create_eventset.3, man3/PAPI_debug_option_t.3,
|
|
Packit |
577717 |
man3/PAPI_descr_error.3, man3/PAPI_destroy_eventset.3,
|
|
Packit |
577717 |
man3/PAPI_detach.3, man3/PAPI_dmem_info_t.3,
|
|
Packit |
577717 |
man3/PAPI_domain_option_t.3, man3/PAPI_enum_event.3,
|
|
Packit |
577717 |
man3/PAPI_event_code_to_name.3, man3/PAPI_event_info_t.3,
|
|
Packit |
577717 |
man3/PAPI_event_name_to_code.3, man3/PAPI_exe_info_t.3,
|
|
Packit |
577717 |
man3/PAPI_flips.3, man3/PAPI_flops.3, man3/PAPI_get_cmp_opt.3,
|
|
Packit |
577717 |
man3/PAPI_get_component_info.3, man3/PAPI_get_dmem_info.3,
|
|
Packit |
577717 |
man3/PAPI_get_event_info.3, man3/PAPI_get_executable_info.3,
|
|
Packit |
577717 |
man3/PAPI_get_hardware_info.3, man3/PAPI_get_multiplex.3,
|
|
Packit |
577717 |
man3/PAPI_get_opt.3, man3/PAPI_get_overflow_event_index.3,
|
|
Packit |
577717 |
man3/PAPI_get_real_cyc.3, man3/PAPI_get_real_nsec.3,
|
|
Packit |
577717 |
man3/PAPI_get_real_usec.3, man3/PAPI_get_shared_lib_info.3,
|
|
Packit |
577717 |
man3/PAPI_get_thr_specific.3, man3/PAPI_get_virt_cyc.3,
|
|
Packit |
577717 |
man3/PAPI_get_virt_nsec.3, man3/PAPI_get_virt_usec.3,
|
|
Packit |
577717 |
man3/PAPI_granularity_option_t.3, man3/PAPI_hw_info_t.3,
|
|
Packit |
577717 |
man3/PAPI_inherit_option_t.3, man3/PAPI_ipc.3,
|
|
Packit |
577717 |
man3/PAPI_is_initialized.3, man3/PAPI_itimer_option_t.3,
|
|
Packit |
577717 |
man3/PAPI_library_init.3, man3/PAPI_list_events.3,
|
|
Packit |
577717 |
man3/PAPI_list_threads.3, man3/PAPI_lock.3,
|
|
Packit |
577717 |
man3/PAPI_mh_cache_info_t.3, man3/PAPI_mh_info_t.3,
|
|
Packit |
577717 |
man3/PAPI_mh_level_t.3, man3/PAPI_mh_tlb_info_t.3,
|
|
Packit |
577717 |
man3/PAPI_mpx_info_t.3, man3/PAPI_multiplex_init.3,
|
|
Packit |
577717 |
man3/PAPI_multiplex_option_t.3, man3/PAPI_num_cmp_hwctrs.3,
|
|
Packit |
577717 |
man3/PAPI_num_components.3, man3/PAPI_num_counters.3,
|
|
Packit |
577717 |
man3/PAPI_num_events.3, man3/PAPI_num_hwctrs.3,
|
|
Packit |
577717 |
man3/PAPI_option_t.3, man3/PAPI_overflow.3, man3/PAPI_perror.3,
|
|
Packit |
577717 |
man3/PAPI_preload_info_t.3, man3/PAPI_profil.3,
|
|
Packit |
577717 |
man3/PAPI_query_event.3, man3/PAPI_read.3,
|
|
Packit |
577717 |
man3/PAPI_read_counters.3, man3/PAPI_read_ts.3,
|
|
Packit |
577717 |
man3/PAPI_register_thread.3, man3/PAPI_remove_event.3,
|
|
Packit |
577717 |
man3/PAPI_remove_events.3, man3/PAPI_reset.3,
|
|
Packit |
577717 |
man3/PAPI_set_cmp_domain.3, man3/PAPI_set_cmp_granularity.3,
|
|
Packit |
577717 |
man3/PAPI_set_debug.3, man3/PAPI_set_domain.3,
|
|
Packit |
577717 |
man3/PAPI_set_granularity.3, man3/PAPI_set_multiplex.3,
|
|
Packit |
577717 |
man3/PAPI_set_opt.3, man3/PAPI_set_thr_specific.3,
|
|
Packit |
577717 |
man3/PAPI_shlib_info_t.3, man3/PAPI_shutdown.3,
|
|
Packit |
577717 |
man3/PAPI_sprofil.3, man3/PAPI_sprofil_t.3, man3/PAPI_start.3,
|
|
Packit |
577717 |
man3/PAPI_start_counters.3, man3/PAPI_state.3, man3/PAPI_stop.3,
|
|
Packit |
577717 |
man3/PAPI_stop_counters.3, man3/PAPI_strerror.3,
|
|
Packit |
577717 |
man3/PAPI_thread_id.3, man3/PAPI_thread_init.3,
|
|
Packit |
577717 |
man3/PAPI_unlock.3, man3/PAPI_unregister_thread.3,
|
|
Packit |
577717 |
man3/PAPI_write.3, man3/high_api.3, man3/low_api.3,
|
|
Packit |
577717 |
man3/papi_data_structures.3, man3/papi_vector_t.3,
|
|
Packit |
577717 |
man3/ret_codes.3: Update doxygen generated man-pages for the
|
|
Packit |
577717 |
pending release.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
In the future, we need to use a newer version of doxygen to
|
|
Packit |
577717 |
generate the pages (1.7 +) because locally installed verions
|
|
Packit |
577717 |
appear to have a bug.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/nmi_watchdog.c: The nmi_watchdog test should report a
|
|
Packit |
577717 |
Warning if nmi_watchdog is enabled not an error. (Since we do
|
|
Packit |
577717 |
work around it, even if performance is likely impacted).
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/: Makefile, nmi_watchdog.c: I think the nmi_watchdog
|
|
Packit |
577717 |
stuff is going to cause us problems down the road.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Thus add a test that will tell users about the issue.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perf_events.c: The nmi_watchdog workaround is needed for
|
|
Packit |
577717 |
multiplexing too.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
The kernel devs don't seem eager to fix this. Until they do,
|
|
Packit |
577717 |
we'll have to fall back to software multiplexing on recent
|
|
Packit |
577717 |
kernels that have nmi_watchdog enabled (most vendor kernels).
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/multiplex.c: Yesterday's coverity fix to make sure the
|
|
Packit |
577717 |
cleanup and destroy rerturn values were checked ended up
|
|
Packit |
577717 |
over-writing "retval" in a way that broke the sdsc4-mpx test.
|
|
Packit |
577717 |
Fix things so that doesn't happen.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi.c, perf_events.c, ctests/overflow_allcounters.c: Some
|
|
Packit |
577717 |
changes for perf_event MIPS support
|
|
Packit |
577717 |
|
|
Packit |
577717 |
+ Add __mips__ cases to the format_group, schedulability, and
|
|
Packit |
577717 |
broken multiplexing bug workarounds, as even new Linux mips
|
|
Packit |
577717 |
kernels have these bugs + fix overflow_allcounters to work
|
|
Packit |
577717 |
properly if the MHz value is zero. + Add some debugging to
|
|
Packit |
577717 |
PAPI_overflow() so that errors are more obvious than just
|
|
Packit |
577717 |
returning PAPI_EINVAL, which made the previous item a pain to
|
|
Packit |
577717 |
track down.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* man/: footer.htm, header.htm, manServer_papi.pl, papiman.bat,
|
|
Packit |
577717 |
html/papi.html, html/papi_accum.html,
|
|
Packit |
577717 |
html/papi_accum_counters.html, html/papi_add_event.html,
|
|
Packit |
577717 |
html/papi_add_events.html,
|
|
Packit |
577717 |
html/papi_assign_eventset_component.html, html/papi_attach.html,
|
|
Packit |
577717 |
html/papi_avail.html, html/papi_cleanup_eventset.html,
|
|
Packit |
577717 |
html/papi_clockres.html, html/papi_command_line.html,
|
|
Packit |
577717 |
html/papi_cost.html, html/papi_create_eventset.html,
|
|
Packit |
577717 |
html/papi_decode.html, html/papi_destroy_eventset.html,
|
|
Packit |
577717 |
html/papi_detach.html, html/papi_encode_events.html,
|
|
Packit |
577717 |
html/papi_enum_event.html, html/papi_event_chooser.html,
|
|
Packit |
577717 |
html/papi_event_code_to_name.html,
|
|
Packit |
577717 |
html/papi_event_name_to_code.html, html/papi_flips.html,
|
|
Packit |
577717 |
html/papi_flops.html, html/papi_get_component_info.html,
|
|
Packit |
577717 |
html/papi_get_dmem_info.html, html/papi_get_event_info.html,
|
|
Packit |
577717 |
html/papi_get_executable_info.html,
|
|
Packit |
577717 |
html/papi_get_hardware_info.html, html/papi_get_multiplex.html,
|
|
Packit |
577717 |
html/papi_get_opt.html, html/papi_get_overflow_event_index.html,
|
|
Packit |
577717 |
html/papi_get_real_cyc.html, html/papi_get_real_usec.html,
|
|
Packit |
577717 |
html/papi_get_shared_lib_info.html,
|
|
Packit |
577717 |
html/papi_get_substrate_info.html,
|
|
Packit |
577717 |
html/papi_get_thr_specific.html, html/papi_get_virt_cyc.html,
|
|
Packit |
577717 |
html/papi_get_virt_usec.html, html/papi_help.html,
|
|
Packit |
577717 |
html/papi_ipc.html, html/papi_is_initialized.html,
|
|
Packit |
577717 |
html/papi_library_init.html, html/papi_list_events.html,
|
|
Packit |
577717 |
html/papi_list_threads.html, html/papi_lock.html,
|
|
Packit |
577717 |
html/papi_mem_info.html, html/papi_multiplex_init.html,
|
|
Packit |
577717 |
html/papi_native.html, html/papi_native_avail.html,
|
|
Packit |
577717 |
html/papi_num_cmp_hwctrs.html, html/papi_num_components.html,
|
|
Packit |
577717 |
html/papi_num_counters.html, html/papi_num_events.html,
|
|
Packit |
577717 |
html/papi_num_hwctrs.html, html/papi_overflow.html,
|
|
Packit |
577717 |
html/papi_perror.html, html/papi_presets.html,
|
|
Packit |
577717 |
html/papi_profil.html, html/papi_query_event.html,
|
|
Packit |
577717 |
html/papi_read.html, html/papi_read_counters.html,
|
|
Packit |
577717 |
html/papi_register_thread.html, html/papi_remove_event.html,
|
|
Packit |
577717 |
html/papi_remove_events.html, html/papi_reset.html,
|
|
Packit |
577717 |
html/papi_set_cmp_domain.html,
|
|
Packit |
577717 |
html/papi_set_cmp_granularity.html, html/papi_set_debug.html,
|
|
Packit |
577717 |
html/papi_set_domain.html, html/papi_set_event_info.html,
|
|
Packit |
577717 |
html/papi_set_granularity.html, html/papi_set_multiplex.html,
|
|
Packit |
577717 |
html/papi_set_opt.html, html/papi_set_thr_specific.html,
|
|
Packit |
577717 |
html/papi_shutdown.html, html/papi_sprofil.html,
|
|
Packit |
577717 |
html/papi_start.html, html/papi_start_counters.html,
|
|
Packit |
577717 |
html/papi_state.html, html/papi_stop.html,
|
|
Packit |
577717 |
html/papi_stop_counters.html, html/papi_strerror.html,
|
|
Packit |
577717 |
html/papi_thread_id.html, html/papi_thread_init.html,
|
|
Packit |
577717 |
html/papi_unlock.html, html/papi_unregister_thread.html,
|
|
Packit |
577717 |
html/papi_write.html, html/papif.html,
|
|
Packit |
577717 |
html/papif_get_clockrate.html, html/papif_get_domain.html,
|
|
Packit |
577717 |
html/papif_get_exe_info.html, html/papif_get_granularity.html,
|
|
Packit |
577717 |
html/papif_get_preload.html, html/papif_set_event_domain.html,
|
|
Packit |
577717 |
images/cssigoff.gif, images/cssigon.gif, images/headertop.jpg,
|
|
Packit |
577717 |
images/line.gif, images/logobottom.jpg, images/logoleft.jpg,
|
|
Packit |
577717 |
images/menubg.jpg, images/menubg95.jpg, images/rd.jpg,
|
|
Packit |
577717 |
images/spinbg.jpg, images/spinlogo.gif, images/stable.gif,
|
|
Packit |
577717 |
images/stripes2.jpg, images/trans.gif, images/utsigoff.gif,
|
|
Packit |
577717 |
images/utsigon.gif, images/white.jpg: Remove the old html
|
|
Packit |
577717 |
documentation and assorted helper files.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/coretemp/linux-coretemp.c: Fix a possible
|
|
Packit |
577717 |
directory stream leak in the coretemp component.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
reported by coverity checker.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/calibrate.c: Properly free the arrays in calibrate,
|
|
Packit |
577717 |
introduced by yesterdays coverity fix.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch by Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-10-24
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/coretemp/linux-coretemp.c: Fix coretemp to not
|
|
Packit |
577717 |
fail if /sys/class/hwmon doesn't exist.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/coretemp/linux-coretemp.c: Patch coretemp to only
|
|
Packit |
577717 |
free the initialized data in shutdown_substrate (once per
|
|
Packit |
577717 |
PAPI_init) rather than shutdown (once per thread).
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This was causing double free errors.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/utils/multiplex_cost.c: Fix various calls to PAPI_start() and
|
|
Packit |
577717 |
PAPI_stop() in multiplex_cost that didn't check the return value.
|
|
Packit |
577717 |
Took care to try to avoid changing timing measurements. Noticed
|
|
Packit |
577717 |
by coverity checker.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/utils/cost.c: In one case, cost was not checking the return
|
|
Packit |
577717 |
of PAPI_start()/PAPI_stop(). This change makes it does so, while
|
|
Packit |
577717 |
being careful not to interfere with the timing that is going on.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/: pthrtough.c, pthrtough2.c: pthrtough and pthrtough2
|
|
Packit |
577717 |
were not checking the return value for pthread_attr_setscope().
|
|
Packit |
577717 |
Reported by coverity checker.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/multiplex1_pthreads.c: multiplex1_pthreads was not
|
|
Packit |
577717 |
checking the return from PAPI_library_init() as flagged by
|
|
Packit |
577717 |
coverity checker.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/inherit.c: inherit.c wasn't checking the result of the
|
|
Packit |
577717 |
waitpid() call, as reported by coverity checker.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/clockres_pthreads.c: Check the return of
|
|
Packit |
577717 |
pthread_create().
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Reported by coverity checker.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_libpfm4_events.c: Fix an actual bug (reported as
|
|
Packit |
577717 |
deadcode by coverity) where _papi_hwd_ntv_code_to_descr was
|
|
Packit |
577717 |
appending extraneous ", masks:" strings into an event
|
|
Packit |
577717 |
description.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
None of our utils/ctests exercise this function, which is
|
|
Packit |
577717 |
probably why the bug wasn't noticed.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: multiplex.c, papi.c: Fix cases where PAPI_*() functions
|
|
Packit |
577717 |
were called without checking the return for an error.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Reported by coverity.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* doc/Doxyfile.utils: Update version to 4.2.0 for pending release.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/multiplex.c: Fix some code that could potentially dereference
|
|
Packit |
577717 |
a null pointer.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Found by the coverity checker.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_vector.c: Remove a dead code case as reported by
|
|
Packit |
577717 |
coverity. Shouldn't break anything as I can't find anywhere that
|
|
Packit |
577717 |
vector_print_table() is actually called.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* release_procedure.txt: Update release_procedure to reflect
|
|
Packit |
577717 |
another file that needs a version number bump. (Doxyfile.utils)
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/calibrate.c: Fix some weird code that was sharing a
|
|
Packit |
577717 |
memory allocation for both double and floats. This was really
|
|
Packit |
577717 |
ugly and made the coverity checker sad.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch provided by Will Cohen.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/testlib/test_utils.c: Fix a signed/unsigned comparison bug I
|
|
Packit |
577717 |
introduced.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/coretemp/tests/coretemp_basic.c: Fix the test so
|
|
Packit |
577717 |
it correctly iterates all of the components.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/coretemp/: linux-coretemp.c, tests/Makefile,
|
|
Packit |
577717 |
tests/coretemp_basic.c: Fix a potential memory leak in coretemp
|
|
Packit |
577717 |
(flagged by coverity).
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Also added a test case for coretemp so I can actually test if
|
|
Packit |
577717 |
these changes are breaking anything.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/solaris-ultra.c: Remove const decleration from get_virt_* in
|
|
Packit |
577717 |
solaris substrate. Vince removed this from papi_vector.h back in
|
|
Packit |
577717 |
June.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/testlib/test_utils.c: Improce the add_two_events() code in
|
|
Packit |
577717 |
the test library. Before it was possible to overrun a buffer if
|
|
Packit |
577717 |
none of the potential predefined events were available.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Noticed by the coverity checker.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* papi.spec, doc/Doxyfile, doc/Doxyfile-everything, src/configure,
|
|
Packit |
577717 |
src/papi.h, src/Makefile.in, src/configure.in: Update version to
|
|
Packit |
577717 |
4.2.0 for pending release.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-10-21
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: Makefile.inc, configure, configure.in, papi.c, papi.h,
|
|
Packit |
577717 |
papi_internal.c, papi_user_events.c, papi_user_events.h: Merge in
|
|
Packit |
577717 |
the user events code , protected by a configure option. (
|
|
Packit |
577717 |
--with-user-events )
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/testlib/test_utils.c: We now ensure that test_fail() always
|
|
Packit |
577717 |
exits. There was some code around that tracked the number of
|
|
Packit |
577717 |
times test_fail() was called. Remove that, as I think it was
|
|
Packit |
577717 |
confusing the coverity checker and causing a huge number of false
|
|
Packit |
577717 |
positives for NULL pointer dereferences.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/acpi/linux-acpi.c: Some minor cleanups to the acpi
|
|
Packit |
577717 |
component. It was choking a bit if ACPI didn't provide thermal
|
|
Packit |
577717 |
information, and also fix a few coverity bugs involving not
|
|
Packit |
577717 |
checking the result of a dup() call.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/testlib/test_utils.c: Another problem with negative numbers,
|
|
Packit |
577717 |
this time one could potentially be passed to a malloc call.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
noticed by coverity
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/overflow_pthreads.c: We were indexing an array with a
|
|
Packit |
577717 |
returned value that could be negative on failure. Add a check to
|
|
Packit |
577717 |
avoid that.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
We're also indexing a per-thread array with an EventSet number,
|
|
Packit |
577717 |
which sounds suspect, should probably investigate that further.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perf_events.c: perf_events.c was setting variables to -1 and
|
|
Packit |
577717 |
then potentially using them to index arrays or call close() on
|
|
Packit |
577717 |
them.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This adds checks to avoid that.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Noticed by the coverity checker.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/lustre/linux-lustre.h: Include stdint.h and
|
|
Packit |
577717 |
ctype.h; needed for uint64_t and isspace() respectivly.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/coretemp/linux-coretemp.c: Fix problem where we
|
|
Packit |
577717 |
try to manipulate a NULL directory entry.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This fixes a segfault on a Nehalem machine we have here that has
|
|
Packit |
577717 |
a /sys/class/hwmon/hwmon0 directory without a "device"
|
|
Packit |
577717 |
subdirectory.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/coretemp/linux-coretemp.c: We were opening a file
|
|
Packit |
577717 |
but not checking for failure before reading from it.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Flagged by the coverity checker.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/coretemp/linux-coretemp.c: Both gcc and coverity
|
|
Packit |
577717 |
were complaining about using an uninitialized pointer. This
|
|
Packit |
577717 |
makes sure it's not dereferenced if not initialized.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/prof_utils.c: Stop doing unnecessary pointer math in a
|
|
Packit |
577717 |
print statement.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This was flagged as a problem by the coverity tool.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/coretemp/linux-coretemp.c: Fix some wrong buffer
|
|
Packit |
577717 |
sizes in the coretemp component.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/sdsc.c: add some extra debug info for sdsc test
|
|
Packit |
577717 |
failures.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_hl.c: Add comment to PAPI_num_counters() documentation
|
|
Packit |
577717 |
about use of PAPI_num_cmp_hwctrs() for component counters.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-10-19
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi.c: Correct documentation errors for PAPI_strerror.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Under a no-cpu-counters build,
|
|
Packit |
577717 |
still build all of the utils. We probably want to rethink some
|
|
Packit |
577717 |
of the cost util details.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-10-11
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/run_tests.sh: Remove an unneeded call to "cat". For some
|
|
Packit |
577717 |
reason it was printing pointless warnings that needlessly
|
|
Packit |
577717 |
cluttered the buildbot logs.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/: Makefile, multiplex1.c: -lpapi should never be a
|
|
Packit |
577717 |
dependency. -I.. is missing in makefile
|
|
Packit |
577717 |
|
|
Packit |
577717 |
You should be able to cd ctests and do: make <test> or make
|
|
Packit |
577717 |
multiplex.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Also, added the read after start multiplex case for multiplex1.
|
|
Packit |
577717 |
This triggers bugs in perf_events systems.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-10-10
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi.c, papi_internal.c, threads.c: The multiplex1_pthreads
|
|
Packit |
577717 |
test was reporting a memory leak.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This is because the test was calling PAPI_unregister_thread()
|
|
Packit |
577717 |
without destroying its EventSets.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This added change adds code that at unregister_thread time will
|
|
Packit |
577717 |
destroy any events belonging to that thread.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This works on all the current ctests but I should check some of
|
|
Packit |
577717 |
the various corner cases not currently tested.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-10-07
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/libpfm4/: config.mk, lib/pfmlib_amd64.c, lib/pfmlib_common.c,
|
|
Packit |
577717 |
lib/pfmlib_intel_x86.c, lib/events/intel_nhm_events.h,
|
|
Packit |
577717 |
lib/events/intel_wsm_events.h: Merge the "conflicts" from the
|
|
Packit |
577717 |
libpfm4 merge
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: threads.c, threads.h: Fix the MEMORY LEAK errors involving
|
|
Packit |
577717 |
the attach ctests (as seen on buildbot)
|
|
Packit |
577717 |
|
|
Packit |
577717 |
These came about when proper multiattach support was added. A
|
|
Packit |
577717 |
"fake" thread structure is created for each attached process.
|
|
Packit |
577717 |
These fake thread structures were not being cleaned up at
|
|
Packit |
577717 |
shutdown, hence the leak.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This fix adds support so at thread shutdown, if we have any
|
|
Packit |
577717 |
"fake" threads that we created, also shut them down too.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This was tricky, especially dealing with the circular-linked list
|
|
Packit |
577717 |
the thread info structs are in. This fix seems to work without
|
|
Packit |
577717 |
negatively affecting the pthread cases.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
ctests/multiplex1_pthreads still reports MEMORY LEAK but that
|
|
Packit |
577717 |
seems to be an eventset issue, not a thread issue, so will be
|
|
Packit |
577717 |
investigated separately.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-10-06
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi.h, papi_fwrappers.c: Add Fortran reference to doxygen
|
|
Packit |
577717 |
main page.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-10-05
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi.c, papi_internal.c, perf_events.c: There has been some
|
|
Packit |
577717 |
ongoing speculation about what would happen if you enabled
|
|
Packit |
577717 |
Multiplexing and Overflow at the same time.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
It turns out (at least on perf_events) that if you have kernel
|
|
Packit |
577717 |
multiplexing, the results are what you expect. You get
|
|
Packit |
577717 |
overflows, but less than in the non-multiplexing case because the
|
|
Packit |
577717 |
overflow counter isn't being run all the time.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
The results for software multiplexing involved a segfault. This
|
|
Packit |
577717 |
is because in the software multiplexing case the primary EventSet
|
|
Packit |
577717 |
is a fiction; a set of shadow EventSets are created behind the
|
|
Packit |
577717 |
scene, and these are the ones used. Therefore when you enable
|
|
Packit |
577717 |
overflow, the overflow event is attempted to be enabled on the
|
|
Packit |
577717 |
fictious main EventSet. There are no native events mapped for
|
|
Packit |
577717 |
it, so overflow tries to access native event array index "-1"
|
|
Packit |
577717 |
which causes bad things to happen.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This change avoids the issue by catching the "-1" case and
|
|
Packit |
577717 |
failing accordingly. We should probably decide if we want to
|
|
Packit |
577717 |
catch the oflo/mpx combination earlier and outright ban it.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
I also went through a lot of the code involved adding comments,
|
|
Packit |
577717 |
as it was really hard following what was going on. This involved
|
|
Packit |
577717 |
the infamously dense "_papi_hwi_remap_event_position()" function
|
|
Packit |
577717 |
too.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi.h: Moved cpu and inherit bits to end of structure for
|
|
Packit |
577717 |
compat across all 4.x lines. Found by Will Cohen.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
As it turns out, I ended up reviewing the CPU_ATTACH changes; I
|
|
Packit |
577717 |
had not done so before. This functionality actually belongs in
|
|
Packit |
577717 |
PAPI_set_granularity. A CPU is a natural unit of granularity of
|
|
Packit |
577717 |
counting, and that value was speced in papi.h a long time ago.
|
|
Packit |
577717 |
Right thing to do here is leave the current attach stuff but make
|
|
Packit |
577717 |
it work as part of set_granularity.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Consider that a TODO for 4.3.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-10-04
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* doc/: Doxyfile, Doxyfile-everything: Enable macro expansion in
|
|
Packit |
577717 |
the doxygen preprocessor step.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Doxygen was not creating docs for the fortran functions and I
|
|
Packit |
577717 |
believe it is because it was silently choking on our clever
|
|
Packit |
577717 |
preprocessor abuse; this fixes? that. However, its worth taking
|
|
Packit |
577717 |
a critical eye to the generated pages again.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi.c, papi_fwrappers.c, papi_hl.c: make "* #include" into
|
|
Packit |
577717 |
"* \#include" so doxygen doesn't treat it as a command.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_fwrappers.c: Added all doxygen stubs to the PAPIF group.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-10-03
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/ipc.c: My previous "fix" for the array bounds issue in
|
|
Packit |
577717 |
ipc.c had multiple embarassing bugs.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Thanks to Will Cohen for noticing. Things should be better now.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: Rules.perfctr-pfm, Rules.pfm_pe: Additionally remove the
|
|
Packit |
577717 |
now extraneous papi_libpfm_preset definition from the other Rules
|
|
Packit |
577717 |
files too.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: Makefile.inc, Rules.pfm4_pe: The change to make the preset
|
|
Packit |
577717 |
code generic accidentally ended up defining the build rules for
|
|
Packit |
577717 |
the file in duplicate places. This fixes that.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-30
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: linux-common.c, utils/decode.c: Fix two unused variable
|
|
Packit |
577717 |
warnings.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/second.c: We were allocating the "values" array but
|
|
Packit |
577717 |
never freeing it.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/: sdsc2.c, sdsc4.c: The SDSC tests could walk off the
|
|
Packit |
577717 |
end of an array.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/overflow_twoevents.c: We could potentially access
|
|
Packit |
577717 |
outside an array boundary in overflow_twoevents.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/ipc.c: ipc was also abusing array boundaries.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/flops.c: The flops.c ctest was abusing the notion of C
|
|
Packit |
577717 |
arrays, by writing INDEX*INDEX values to mresult[0][i], I suppose
|
|
Packit |
577717 |
"knowing" that this would fill in the whole array. Fix things to
|
|
Packit |
577717 |
use an additional iterator.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/byte_profile.c: The coverity checker rightly points
|
|
Packit |
577717 |
out that the last argument to strncat should be buffersize-1.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/: exeinfo.c, shlib.c: Coverity flagged that there were
|
|
Packit |
577717 |
some tests that had no effect. In particular the are tests that
|
|
Packit |
577717 |
the pointers are non-null. However, they are arrays rather than
|
|
Packit |
577717 |
pointers. This patch make it clear that arrays are being used in
|
|
Packit |
577717 |
the code.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Will Cohen at redhat
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/clockcore.c: This is a relatively minor patch that
|
|
Packit |
577717 |
ensures that all the allocated memory is initialized to zero
|
|
Packit |
577717 |
before it is used. Coverity might not be smart enough to
|
|
Packit |
577717 |
determine whether the test actually wrote into all the locations
|
|
Packit |
577717 |
because of the case statement. This is make it easier for
|
|
Packit |
577717 |
coverity to determine that the memory has been initialized.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Path from Will Cohen at redhat.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/multiplex.c: Coverity scan showed that MPX_cleanup() function
|
|
Packit |
577717 |
was blindly accessing a value through a pointer and then checking
|
|
Packit |
577717 |
to see that the pointer was null. This patch makes sure that the
|
|
Packit |
577717 |
pointer is checked before it is used.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Will Cohen at redhat.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/: pthrtough.c, pthrtough2.c: Coverity found that the
|
|
Packit |
577717 |
sizeof argument for pthrtough2.c and pthrtough.c was using
|
|
Packit |
577717 |
sizeof(pthread *) rather than sizeof(pthread). This patch fixes
|
|
Packit |
577717 |
that problem.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Will Cohen at redhat
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_internal.c: This change moves the setting for default
|
|
Packit |
577717 |
domain to be enforced at eventset add time, rather than eventset
|
|
Packit |
577717 |
creation time.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This fixes some problems seen when multiplexing.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
The patch was provided by Phil Mucci.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/pmapi-ppc64.h: One more file that is no longer needed.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in, perfctr.c, pmapi-ppc64_events.c,
|
|
Packit |
577717 |
ppc64_events.c: Clean up the now not-needed pmapi-ppc64_events.c
|
|
Packit |
577717 |
file.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: Makefile.inc, aix.c, aix.h, configure, configure.in,
|
|
Packit |
577717 |
papi_libpfm_presets.c: Finalize the merge of the preset code.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/aix.c: Fix a missing include.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: aix.c, configure, configure.in: Move more code to its
|
|
Packit |
577717 |
proper place.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: aix.c, configure, configure.in, pmapi-ppc64.c,
|
|
Packit |
577717 |
pmapi-ppc64_events.c, ppc64_events.c: Move the
|
|
Packit |
577717 |
ppc64_setup_native_table() routines out of the preset code.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This is complicated, as there are two very similar routines
|
|
Packit |
577717 |
setup_ppc64_native_table() used by AIX/pmapi and
|
|
Packit |
577717 |
ppc64_setup_native_table() used by perfctr
|
|
Packit |
577717 |
|
|
Packit |
577717 |
These could probably be merged too, but this is definitely not
|
|
Packit |
577717 |
the time.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: aix.c, papi_libpfm_presets.c, pmapi-ppc64_events.c: move
|
|
Packit |
577717 |
pmapi_find_full_event to be _aix_ntv_name_to_code() as it
|
|
Packit |
577717 |
probably always should have been.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_libpfm_presets.c, papi_setup_presets.h,
|
|
Packit |
577717 |
pmapi-ppc64_events.c: Make papi_libpfm_presets more generic by
|
|
Packit |
577717 |
calling _papi_hwi_native_name_to_code() rather than a
|
|
Packit |
577717 |
substrate-specific call.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: aix.c, papi_libpfm_presets.c, pmapi-ppc64_events.c: I was
|
|
Packit |
577717 |
mainly doing this to aid debugging, but now the
|
|
Packit |
577717 |
papi_libpfm_presets.c file and pmapi-ppc64_events.c file are
|
|
Packit |
577717 |
close enough to being identical I might try to merge them.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-29
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_libpfm_presets.c, pmapi-ppc64_events.c,
|
|
Packit |
577717 |
ppc64_events.h: The files are almost the same now.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_libpfm_presets.c, pmapi-ppc64_events.c: More making
|
|
Packit |
577717 |
these files the same, including some memory leak fixes that made
|
|
Packit |
577717 |
it to the former but not the latter.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_libpfm_presets.c, pmapi-ppc64_events.c: Tracking down
|
|
Packit |
577717 |
problems on AIX can be a bit of a pain because
|
|
Packit |
577717 |
papi_libpfm_presets.c and pmapi-ppc64_events.c are almost (but
|
|
Packit |
577717 |
not quite) the same. This change makes the files more similar,
|
|
Packit |
577717 |
mostly by cleaning up whitespace and normalizing comments and
|
|
Packit |
577717 |
debugging statements between the two.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/pmapi-ppc64_events.c: Ugh, obvious typo in that last commit.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/pmapi-ppc64_events.c: In ppc64_setup_gps() the current code
|
|
Packit |
577717 |
sometimes walks off the end of the group array and trashes
|
|
Packit |
577717 |
unrelated memory.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Until we work out the proper fix, this prints an error message
|
|
Packit |
577717 |
and stops the loop before memory is corrupted.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_data.h: No one seems to remember the last time this file
|
|
Packit |
577717 |
was used, so let's remove it.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-28
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/Makefile.inc: Remove the "u" option to the "ar" command that
|
|
Packit |
577717 |
links libpapi.a, as it was breaking the build on MIPS.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This *shouldn't* break anything, but messing around with "ar"
|
|
Packit |
577717 |
options can be potentially dangerous. I'll double-check the
|
|
Packit |
577717 |
non-Linux builds.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/libpfm4/lib/: Makefile, pfmlib_mips_priv.h,
|
|
Packit |
577717 |
events/intel_nhm_events.h, events/intel_wsm_events.h: Fix up the
|
|
Packit |
577717 |
"collisions" from the libpfm4 import
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-26
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/Makefile.inc: We would like to use parallel make on packages
|
|
Packit |
577717 |
to speed things up. However, when this was tried with papi the
|
|
Packit |
577717 |
"make -j4" failed
|
|
Packit |
577717 |
(https://bugzilla.redhat.com/show_bug.cgi?id=740909). I took a
|
|
Packit |
577717 |
look through the code and found that some of dependencies were
|
|
Packit |
577717 |
not quite right. Turns out that $(papiLIBS) is substituted during
|
|
Packit |
577717 |
the configure, but it isn't available for the actual make.
|
|
Packit |
577717 |
Attached is the patch that ensures that the $(LIBS) are built
|
|
Packit |
577717 |
before utils and tests.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Will Cohen <wcohen at redhat.com>
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/run_tests.sh: Modify run_tests.sh so that you can set the
|
|
Packit |
577717 |
VALGRIND command externally via environment variable without
|
|
Packit |
577717 |
having to edit run_tests.sh itself.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Also adds Date and cpuinfo information to the beginning of
|
|
Packit |
577717 |
run_tests.sh results. This can help when run run_tests.sh output
|
|
Packit |
577717 |
is passed around when debugging a problem.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Phil Mucci
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: If we have no Fortran compiler
|
|
Packit |
577717 |
available, then our current build system tries to build the
|
|
Packit |
577717 |
Fortran examples with an empty compiler string which just
|
|
Packit |
577717 |
generates strange errors.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This patch changes F77 to be "echo" which at least avoids the
|
|
Packit |
577717 |
errors. The proper fix is probably just not to build the Fortran
|
|
Packit |
577717 |
samples if no compiler is available.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Phil Mucci
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_libpfm4_events.c: The build on power6 was warning in a
|
|
Packit |
577717 |
DEBUG statement because sizeof() returns an int rather than a
|
|
Packit |
577717 |
long. So use a cast to avoid this.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perf_events.c: The move to use pid_t for pid values caused
|
|
Packit |
577717 |
warnings on a --with-debug build due to the lack of a way to
|
|
Packit |
577717 |
print a pid_t value without a cast.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This fix adds the proper casts.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-23
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_libpfm4_events.c: Rename the "perfmon_idx" structure
|
|
Packit |
577717 |
field the more evocative "libpfm4_idx" value.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Phil Mucci
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/all_native_events.c: Fix problem where we were passing
|
|
Packit |
577717 |
a pointer to an EventSet rather than the actual EventSet number
|
|
Packit |
577717 |
to PAPI_cleanup_eventset().
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Also include some of the cleanups from Phil Mucci's MIPS tree.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: perf_events.c, perf_events.h: Make the perf_event ctl
|
|
Packit |
577717 |
structure have more explicit data types.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Philip Mucci
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: cycle.h, linux-common.c, linux-context.h, linux-lock.h,
|
|
Packit |
577717 |
linux-timer.c, mb.h, papi.h: Add bare minimal MIPS74k support,
|
|
Packit |
577717 |
enough to compile.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch from Philip Mucci
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_events.csv: Add MIPS 74k pre-defined events
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Patch by Philip Mucci
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-22
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/all_native_events.c: Heike's cleanup_eventset work
|
|
Packit |
577717 |
allows the calling of PAPI_cleanup_eventset with cuda, so
|
|
Packit |
577717 |
uncomment the eventset cleanup code in all_native_events.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi.h: Update papi.h to properly detect if being built with
|
|
Packit |
577717 |
a C99 compiler.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_events.csv: Update PAPI_FP_INS event name on amd_fam14h
|
|
Packit |
577717 |
as it was changed in the most recent libpfm4 merge
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/libpfm4/: README, config.mk, docs/Makefile,
|
|
Packit |
577717 |
docs/man3/pfm_get_event_info.3, examples/Makefile,
|
|
Packit |
577717 |
examples/showevtinfo.c, include/Makefile,
|
|
Packit |
577717 |
include/perfmon/perf_event.h, lib/Makefile, lib/pfmlib_common.c,
|
|
Packit |
577717 |
lib/pfmlib_gen_mips64_priv.h, lib/pfmlib_mips.c,
|
|
Packit |
577717 |
lib/pfmlib_mips_74k.c, lib/pfmlib_mips_perf_event.c,
|
|
Packit |
577717 |
lib/pfmlib_mips_priv.h, lib/pfmlib_perf_event_pmu.c,
|
|
Packit |
577717 |
lib/pfmlib_priv.h, lib/events/intel_atom_events.h,
|
|
Packit |
577717 |
lib/events/intel_core_events.h, lib/events/intel_nhm_events.h,
|
|
Packit |
577717 |
lib/events/intel_snb_events.h, lib/events/intel_wsm_events.h: Fix
|
|
Packit |
577717 |
the "conflicts" from the libpfm4 git import
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/libpfm4/: docs/man3/libpfm_mips_74k.3, tests/validate_arm.c,
|
|
Packit |
577717 |
tests/validate_mips.c: Initial revision
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-21
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/multiplex.c: Fix problem where we were freeing a
|
|
Packit |
577717 |
singly-linked list in a for loop, possibly free()ing the
|
|
Packit |
577717 |
allocation before dereferencing ->next
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Problem reported by coverity tool, via Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/utils/cost.c: Fixed uninitialized data problem in papi_cost
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Problem reported by coverity tool, via Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_internal.c: Fix problem where we were copying around
|
|
Packit |
577717 |
chunks of memory that were not initialized yet.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Problem reported by coverity tool, via Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/multiplex.c: Fix two cases where we were dereferencing a
|
|
Packit |
577717 |
pointer without checking for NULL.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Problem reported by coverity tool, via Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/linux-memory.c: We were opening files but not properly
|
|
Packit |
577717 |
closing them if we returned early with an error condition.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Problem reported by coverity tool, via Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/linux-common.c: The coverity tool noticed that we allocate
|
|
Packit |
577717 |
and populate a cpu node info structure, but we never pass any
|
|
Packit |
577717 |
info on this structure outside of the cpu detection routine, in
|
|
Packit |
577717 |
effect leaking the allocation.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
For now just comment out this code as it is not used by anyone.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Problem reported by coverity tool, via Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi.c, papi_libpfm3_events.c, perfctr-x86.c: The coverity
|
|
Packit |
577717 |
checker was reporting we forgot to fclose() /proc/cpuinfo in
|
|
Packit |
577717 |
papi.c
|
|
Packit |
577717 |
|
|
Packit |
577717 |
The bigger question, is why were we unconditionally trying to
|
|
Packit |
577717 |
open /proc/cpuinfo in generic code in papi.c anyway?
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Turns out it was to set the event masks properly for itanium and
|
|
Packit |
577717 |
p4.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
The platform code sets CPU vendor and family for us though, so if
|
|
Packit |
577717 |
we just make the event mask code use those values then we don't
|
|
Packit |
577717 |
have to open cpuinfo. This also means that non-Linux users with
|
|
Packit |
577717 |
the misfortune of running on a P4 might actually work too.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_internal.c, papi_libpfm_presets.c: In various places
|
|
Packit |
577717 |
we were using MAX_COUNTER_TERMS (defined by substrate) rather
|
|
Packit |
577717 |
than PAPI_MAX_COUNTER_TERMS (a papi predefined event define).
|
|
Packit |
577717 |
This could cause buffer overruns.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This fixes things, though really we shouldn't have such similar
|
|
Packit |
577717 |
names for different defines.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Problem reported by coverity tool, via Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/multiplex.c: Avoid case where we could have been
|
|
Packit |
577717 |
dereferencing a NULL pointer in MPX_stop()
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Reported by coverity tool, via Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi.c: Fix problem where thread and cpu could be
|
|
Packit |
577717 |
dereferenced as NULL in PAPI_start()
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Reported by coverity tool, via Will Cohen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_events.csv: Update the AMD Family 14h (Bobcat)
|
|
Packit |
577717 |
pre-defined events.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
It turns out they are different enough from 10h that they need
|
|
Packit |
577717 |
their own category.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
In going through the Fam14h BKDG it turns out that Bobcat has a
|
|
Packit |
577717 |
really nice set of events available, especially for
|
|
Packit |
577717 |
Floating-Point/SSE but also memory bandwidth.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
With this change, all of the ctests pass on a Bobcat machine.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Recent Ubuntu versions use the ld
|
|
Packit |
577717 |
flag --as-needed by default.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This breaks the PAPI configure step for the libdl check, as the
|
|
Packit |
577717 |
--as-needed flag enforces the rule that libraries (in this case
|
|
Packit |
577717 |
-ldl) must come after the object files on the command line, not
|
|
Packit |
577717 |
before.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
The fix for this is easy, the libdl check was wrongly sticking
|
|
Packit |
577717 |
-ldl in LDFLAGS rather than in LIBS. Putting it in LIBS makes
|
|
Packit |
577717 |
things work as expected.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
You can see here:
|
|
Packit |
577717 |
http://www.gentoo.org/proj/en/qa/asneeded.xml
|
|
Packit |
577717 |
|
|
Packit |
577717 |
For more info on this issue than you probably ever want to know.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-19
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: ctests/Makefile, ftests/Makefile, utils/Makefile: When
|
|
Packit |
577717 |
building testlib dependencies from ctests/ ftests/ and utils/
|
|
Packit |
577717 |
call $(MAKE) and not make, this should fix aix.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-14
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: aix.c, freebsd.c, linux-bgp.c, papi_vector.c,
|
|
Packit |
577717 |
perf_events.c, perfctr-ppc64.c, perfctr-x86.c, perfmon-ia64.c,
|
|
Packit |
577717 |
perfmon.c, solaris-niagara2.c, solaris-ultra.c,
|
|
Packit |
577717 |
components/acpi/linux-acpi.c,
|
|
Packit |
577717 |
components/coretemp/linux-coretemp.c,
|
|
Packit |
577717 |
components/coretemp_freebsd/coretemp_freebsd.c,
|
|
Packit |
577717 |
components/example/example.c,
|
|
Packit |
577717 |
components/infiniband/linux-infiniband.c,
|
|
Packit |
577717 |
components/lmsensors/linux-lmsensors.c,
|
|
Packit |
577717 |
components/lustre/linux-lustre.c, components/mx/linux-mx.c,
|
|
Packit |
577717 |
components/net/linux-net.c, win2k/substrate/win32.c,
|
|
Packit |
577717 |
win2k/substrate/winpmc-p3.c: Change initialization of function
|
|
Packit |
577717 |
pointer cleanup_eventset() from vec_int_dummy to vec_int_ok_dummy
|
|
Packit |
577717 |
so that it returns PAPI_OK by default. Roll back initialization
|
|
Packit |
577717 |
for every substrate. AGAIN, keep an eye on builtbot.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/libpfm4/lib/: pfmlib_mips.c, pfmlib_mips_74k.c,
|
|
Packit |
577717 |
pfmlib_mips_perf_event.c, pfmlib_mips_priv.h,
|
|
Packit |
577717 |
events/mips_74k_events.h: Merged with HEAD, still passing all
|
|
Packit |
577717 |
tests
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-13
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_libpfm4_events.c: The libpfm4 code was doing a full call
|
|
Packit |
577717 |
to pfm_get_os_event_encoding() during every call to
|
|
Packit |
577717 |
update_control_state().
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This is unnecessary, as we can call pfm_get_os_event_encoding()
|
|
Packit |
577717 |
once at event creation time and cache the results. There's no
|
|
Packit |
577717 |
need to call it each update_control_state(), as that is called
|
|
Packit |
577717 |
during PAPI_start() and thus relatively time critical.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/run_tests.sh: Missed a $
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: run_tests.sh, components/example/tests/HelloWorld.c: Update
|
|
Packit |
577717 |
run_tests.sh to run component tests, and update the example test
|
|
Packit |
577717 |
to act more like a ctest.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/example/example.c: Fix warnings generated by the
|
|
Packit |
577717 |
example component.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: Makefile.inc, components/Makefile_comp_tests,
|
|
Packit |
577717 |
ctests/Makefile, ctests/do_loops.c, ctests/dummy.c,
|
|
Packit |
577717 |
ctests/papi_test.h, ctests/test_utils.c, ctests/test_utils.h,
|
|
Packit |
577717 |
ftests/Makefile, testlib/Makefile, testlib/do_loops.c,
|
|
Packit |
577717 |
testlib/dummy.c, testlib/papi_test.h, testlib/test_utils.c,
|
|
Packit |
577717 |
testlib/test_utils.h, utils/Makefile: ctests, ftests, utils, and
|
|
Packit |
577717 |
the component tests were all using some files in ctests.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
These weren't being built when --with-no-cpu-counters was
|
|
Packit |
577717 |
enabled, so the PAPI build was breaking when that was enabled as
|
|
Packit |
577717 |
well as a component.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Move the shared files to their own directory, testlib Then update
|
|
Packit |
577717 |
all the users to look in the right place.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
After this commit you might need to do a "cvs -d update" to make
|
|
Packit |
577717 |
sure you get the new subdirectory.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: When compiling with
|
|
Packit |
577717 |
--with-no-cpu-counters configure would report the platform as
|
|
Packit |
577717 |
linux-perfctr-x86. This changes it to report as
|
|
Packit |
577717 |
linux-no-counters
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-12
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: aix.c, freebsd.c, linux-bgp.c, perf_events.c,
|
|
Packit |
577717 |
perfctr-ppc64.c, perfctr-x86.c, perfmon-ia64.c, perfmon.c,
|
|
Packit |
577717 |
solaris-niagara2.c, solaris-ultra.c,
|
|
Packit |
577717 |
components/acpi/linux-acpi.c,
|
|
Packit |
577717 |
components/coretemp/linux-coretemp.c,
|
|
Packit |
577717 |
components/coretemp_freebsd/coretemp_freebsd.c,
|
|
Packit |
577717 |
components/example/example.c,
|
|
Packit |
577717 |
components/infiniband/linux-infiniband.c,
|
|
Packit |
577717 |
components/lmsensors/linux-lmsensors.c,
|
|
Packit |
577717 |
components/lustre/linux-lustre.c, components/mx/linux-mx.c,
|
|
Packit |
577717 |
components/net/linux-net.c, win2k/substrate/win32.c,
|
|
Packit |
577717 |
win2k/substrate/winpmc-p3.c: Initialize new function pointer
|
|
Packit |
577717 |
cleanup_eventset() for every substrate. Keep an eye on builtbot.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/components/cuda/: linux-cuda.c, linux-cuda.h: Cannot override
|
|
Packit |
577717 |
void* definitions from PAPI framework layer (e.g.
|
|
Packit |
577717 |
hwd_control_state_t) with typedefs to conform to PAPI Component
|
|
Packit |
577717 |
layer code if this technique has already been used in another
|
|
Packit |
577717 |
substrate (e.g. perfctr-x86). Or short: #undef and typedef can't
|
|
Packit |
577717 |
be done twice.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perf_events.c: Fix bug caused by forgetting to drop the
|
|
Packit |
577717 |
stream name when converting a fprintf() into a SUBDBG()
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_libpfm_presets.c: Patch from William Cohen fixing a
|
|
Packit |
577717 |
potential problem found by a static analysis tool where we could
|
|
Packit |
577717 |
possibly pass a NULL pointer to free_notes().
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_libpfm_presets.c: Some memory leak fixes made to libpfm3
|
|
Packit |
577717 |
papi_pfm_events.c by Robert Richter were lost when the
|
|
Packit |
577717 |
libpfm4/libpfm4 presets merge was done.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This re-applies these fixes.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-10
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/run_tests.sh: Cleaned up old comment regarding CUDA pre-4.0
|
|
Packit |
577717 |
when it was not possible to access a GPU from multiple CPU
|
|
Packit |
577717 |
threads.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi.c, papi_protos.h, papi_vector.c, papi_vector.h,
|
|
Packit |
577717 |
components/README, components/cuda/linux-cuda.c,
|
|
Packit |
577717 |
components/cuda/linux-cuda.h: Deleted function pointer
|
|
Packit |
577717 |
destroy_eventset from the PAPI vector table, and added
|
|
Packit |
577717 |
cleanup_eventset instead. PAPI_destroy_eventset() requires an
|
|
Packit |
577717 |
empty EventSet. Hence, usually PAPI_cleanup_eventset() is called
|
|
Packit |
577717 |
before PAPI_destroy_eventset(); which also sets the CompIdx to
|
|
Packit |
577717 |
-1. This means, PAPI_destroy_eventset() won't have any knowledge
|
|
Packit |
577717 |
about components. However, in order to disable CUDA eventGroups
|
|
Packit |
577717 |
and to free perfmon hardware on the GPU, knowledge about the CUDA
|
|
Packit |
577717 |
component index is required. Hence, I replaced
|
|
Packit |
577717 |
CUDA_destroy_eventset() with CUDA_cleanup_eventset() in the CUDA
|
|
Packit |
577717 |
component. NOTE: Please make sure you call
|
|
Packit |
577717 |
PAPI_cleanup_eventset() before calling PAPI_shutdown().
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-09
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: papi_protos.h, papi_vector.c, papi_vector.h,
|
|
Packit |
577717 |
components/cuda/linux-cuda.c, components/cuda/linux-cuda.h: CUDA
|
|
Packit |
577717 |
component is now thread-safe. Starting in CUDA 4.0, multiple CPU
|
|
Packit |
577717 |
threads can access the same CUDA context. This is a much easier
|
|
Packit |
577717 |
programming model then pre-4.0 as threads - using the same CUDA
|
|
Packit |
577717 |
context - can share memory, data, etc. Note, it's possible to
|
|
Packit |
577717 |
create a different CUDA context for each thread, but then we are
|
|
Packit |
577717 |
likely running into a limitation that only one context can be
|
|
Packit |
577717 |
profiled at a time.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-07
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/ctests/: do_loops.c, test_utils.c: Apply fixes to problems
|
|
Packit |
577717 |
noticed by a static analysis tool.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Provided by William Cohen at RedHat
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_events.csv: Update SandyBridge preset events.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
These were provided by Michel Brown at Bull
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/libpfm4/lib/: pfmlib_gen_mips64.c, pfmlib_mips.c,
|
|
Packit |
577717 |
pfmlib_mips_74k.c, pfmlib_mips_perf_event.c, pfmlib_mips_priv.h,
|
|
Packit |
577717 |
events/gen_mips64_events.h, events/mips_74k_events.h: MIPS 74K
|
|
Packit |
577717 |
little endian perf event support, requires 3.0.3+ kernel
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-06
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perf_events.c: The warning I had print on nmi_watchdog being
|
|
Packit |
577717 |
found was a bit much, make it a SUBDBG() call instead.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
I do wish there were a way to notify the user more visibly,
|
|
Packit |
577717 |
because losing a counter (when you might only have 4 total to
|
|
Packit |
577717 |
begin with) is a big deal, and most Linux vendors are starting to
|
|
Packit |
577717 |
ship kernels with the nmi_watchdog enabled.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: linux-common.c, linux-common.h, perf_events.c: On newer
|
|
Packit |
577717 |
Linux kernels (2.6.34+) the nmi_watchdog counter can steal one
|
|
Packit |
577717 |
of the counters, reducing by one the total available.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
There's a bug in Linux where if you try to use the full number of
|
|
Packit |
577717 |
counters on such a system with a group leader, the
|
|
Packit |
577717 |
sys_perf_open() call will succeed only to fail at read time.
|
|
Packit |
577717 |
(instead of the proper error code at open time).
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This patch attempts to work around this issue by detecting if a
|
|
Packit |
577717 |
watchdog timer is being used, and in that case re-use the
|
|
Packit |
577717 |
existing KERNEL_CHECKS_SCHEDUABILITY_UPON_OPEN bugfix code.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_events.csv: We were missing a proper libpfm4 interlagos
|
|
Packit |
577717 |
CPU name in the papi_events.csv file
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-09-02
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/libpfm4/: include/perfmon/perf_event.h, lib/Makefile,
|
|
Packit |
577717 |
lib/pfmlib_intel_nhm_unc.c, lib/pfmlib_intel_x86.c,
|
|
Packit |
577717 |
lib/pfmlib_intel_x86_priv.h, lib/pfmlib_priv.h,
|
|
Packit |
577717 |
lib/events/amd64_events_fam10h.h, lib/events/amd64_events_k7.h,
|
|
Packit |
577717 |
lib/events/amd64_events_k8.h, lib/events/intel_atom_events.h,
|
|
Packit |
577717 |
lib/events/intel_core_events.h,
|
|
Packit |
577717 |
lib/events/intel_coreduo_events.h, lib/events/intel_nhm_events.h,
|
|
Packit |
577717 |
lib/events/intel_nhm_unc_events.h, lib/events/intel_p6_events.h,
|
|
Packit |
577717 |
lib/events/intel_snb_events.h, lib/events/intel_wsm_events.h,
|
|
Packit |
577717 |
lib/events/intel_wsm_unc_events.h,
|
|
Packit |
577717 |
lib/events/intel_x86_arch_events.h: Fix "conflicts" from the
|
|
Packit |
577717 |
libpfm4 import
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/papi_libpfm4_events.c: Explicitly set num_native_events to
|
|
Packit |
577717 |
zero at init time.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Somehow the value was surviving fork/exec and making the
|
|
Packit |
577717 |
fork/exec test cases fail on a recent Debian system.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/perf_events.c: Set FD_CLOEXEC on the overflow signal handler
|
|
Packit |
577717 |
fd.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Otherwise if we exec() with overflow enabled, the exec'd process
|
|
Packit |
577717 |
will quickly die due to lack of signal handler.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
This patch is needed due to a change in behavior in Linux 3.0.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Mark Krentel first noticed this problem.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: Rules.perfctr-pfm, Rules.pfm, Rules.pfm4_pe, Rules.pfm_pe:
|
|
Packit |
577717 |
Remove the "unexport CFLAGS" lines from the Rules files.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: multiplex.c, papi_internal.c, utils/component.c: Fix a few
|
|
Packit |
577717 |
warnings reported by gcc-4.6
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Override auto-detection of
|
|
Packit |
577717 |
substrate if the user specifies what they want to build with.
|
|
Packit |
577717 |
This allows building perfctr and perfmon2 PAPI on systems
|
|
Packit |
577717 |
auto-detected as having perf_event support.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Add a "--with-libpfm3" argument to
|
|
Packit |
577717 |
configure that lets us specify libpfm3 for testing purposes.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/solaris-niagara2.c: Fix solaris niagara2 build problems
|
|
Packit |
577717 |
reported by tigrage on the PAPI forum.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-08-30
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/configure: Regen
|
|
Packit |
577717 |
|
|
Packit |
577717 |
2011-08-29
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/configure.in: Check for a requested interface to tweak build
|
|
Packit |
577717 |
flags
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Last bit for cross compiling...
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Better double quotes
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: There can be only 1. (choice of
|
|
Packit |
577717 |
perfctr, perfmon or perf events)
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: configure, configure.in: Further refinement of the
|
|
Packit |
577717 |
combinations of --with-perfctr --with-perfmon and
|
|
Packit |
577717 |
--with-perf-events
|
|
Packit |
577717 |
|
|
Packit |
577717 |
True autotools cross not yet supported until we move to automake.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
I did trick it into doing a cross compile with... # ARCH=mips
|
|
Packit |
577717 |
CC=scgcc ./configure --with-arch=mips
|
|
Packit |
577717 |
--host=mips64el-gentoo-linux-gnu- --with-ffsll --with-libpfm4 --w
|
|
Packit |
577717 |
ith-perf-events --with-virtualtimer=times
|
|
Packit |
577717 |
--with-walltimer=gettimeofday --with-tls=__thread --with-CPU=mips
|
|
Packit |
577717 |
# cross compiling should work differently...
|
|
Packit |
577717 |
|
|
Packit |
577717 |
Wow, do I hate specifying mips in 3 places...
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/: config.h.in, configure, configure.in: Some fixes for cross
|
|
Packit |
577717 |
compiling and not including x86_cache_info.c when not ensured an
|
|
Packit |
577717 |
x86.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/Makefile.inc: Surround component tests and cleanup recipies
|
|
Packit |
577717 |
with a conditional, the version of sh that our aix machine has
|
|
Packit |
577717 |
does not handle for i in {Empty set};
|
|
Packit |
577717 |
|
|
Packit |
577717 |
treating it as a syntax error.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
NOTE: This requires gnu make, my shell-foo couldn't make sh
|
|
Packit |
577717 |
happy, so for now gnu conditionals!
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* ChangeLogP414.txt, RELEASENOTES.txt: Update Release Notes and add
|
|
Packit |
577717 |
ChangeLog for PAPI 4.1.4.
|
|
Packit |
577717 |
|
|
Packit |
577717 |
* src/configure: Rebuild from configure.in with version number bump
|
|
Packit |
577717 |
to 4.1.4 in advance of pending internal vendor release for Cray.
|
|
Packit |
577717 |
|