Branch: refs/heads/master
Home: https://github.com/OpenSIPS/opensips
Commit: dac6ef8a98af1aad01c0e68ea9d3254cb0c7c2b3
https://github.com/OpenSIPS/opensips/commit/dac6ef8a98af1aad01c0e68ea9d3254cb0c7c2b3
Author: Liviu Chircu <li...@opensips.org>
Date: 2019-04-02 (Tue, 02 Apr 2019)
Changed paths:
M doc/build-contrib.sh
Log Message:
-----------
Update contributors file
Commit: 976431ec904afd7fe7ba45e1025eb46ae069ca9d
https://github.com/OpenSIPS/opensips/commit/976431ec904afd7fe7ba45e1025eb46ae069ca9d
Author: Liviu Chircu <li...@opensips.org>
Date: 2019-04-02 (Tue, 02 Apr 2019)
Changed paths:
M mem/common.c
M mem/mem.c
M mem/shm_mem.c
Log Message:
-----------
dynamic allocators: Add some useful startup hints
Commit: 8b19dab594ef7be7690ee3111438ffcae953e531
https://github.com/OpenSIPS/opensips/commit/8b19dab594ef7be7690ee3111438ffcae953e531
Author: Liviu Chircu <li...@opensips.org>
Date: 2019-04-02 (Tue, 02 Apr 2019)
Changed paths:
M mem/rpm_mem.c
M mem/shm_mem.c
Log Message:
-----------
Fix HP_MALLOC statistics with multi allocators
With multiple allocators, the SHM stats table will declare a bunch of
function-based statistics. While this works fine for F_MALLOC and
Q_MALLOC, HP_MALLOC requires atomic counter statistics, due to its
parallel processing nature. So, for HP_MALLOC, we now re-declare its
stats during the init phase to be atomic counters instead of functions
-- not the most elegant solution, but it's quick and not by far dirty.
Thanks to Razvan Crainea for the report! <3
Commit: 8e61c95424a0bd1762b778b3237389f295c8dfd2
https://github.com/OpenSIPS/opensips/commit/8e61c95424a0bd1762b778b3237389f295c8dfd2
Author: Liviu Chircu <li...@opensips.org>
Date: 2019-04-02 (Tue, 02 Apr 2019)
Changed paths:
M mem/f_malloc.c
M mem/f_malloc.h
M mem/hp_malloc.c
M mem/hp_malloc.h
M mem/hp_malloc_stats.c
M mem/hp_malloc_stats.h
M mem/q_malloc.c
M mem/q_malloc.h
M mem/rpm_mem.c
Log Message:
-----------
HP_MALLOC: Make HP_MALLOC_FAST_STATS great again
To reiterate the idea behind HP_MALLOC_FAST_STATS:
* core problem: atomic SHM stat counters are nice, but they do
include locking!
* with this flag, there are almost no stat computations at runtime
(allocs, frees, reallocs will only +/- the numbers of fragments)
* every time user asks for, say, shmem:real_used_size, the stat
will be fully re-computed, at most once every second.
Compare:
https://github.com/OpenSIPS/opensips/compare/39febee49e2f...8e61c95424a0
_______________________________________________
Devel mailing list
Devel@lists.opensips.org
http://lists.opensips.org/cgi-bin/mailman/listinfo/devel
