Using factory method maked it easier to grasp the lifetime of all object
that get created and destroyed. Instead of spanning this thoughout whole
source file we have it nicely encapsulated in this a little horrendeous
_parseDevicePath that is of course to improve more.
Otherwise we're going to leak memory without any need.
Before this fix we've created ConsoleDevice 4 times in case -A switch hadn't been supplied,
but we hadn't ever deleted those. Now there's no memory leak here.
Minor changes to follow coding style and improve readability:
- sort headers
- move struct definition to compilation unit rather than header
- Add braces to if, for, etc
../../libraries/AP_HAL_Linux/Perf.cpp: In member function ‘void Linux::Perf::_debug_counters()’:
../../libraries/AP_HAL_Linux/Perf.cpp:85:36: warning: format ‘%llu’ expects argument of type ‘long long unsigned int’, but argument 4 has type ‘uint64_t {aka long unsigned int}’ [-Wformat=]
c.name, c.count);
^
Test code for integration with another thread to pull data from internal
perf counters. Since we are using the timer thread here, there's no
retry mechanism and we only print that data can be corrupted.
Instead of creating a new object Perf_Lttng copying the necessaries
fields, just make a tighter integration with the internal perf counters
and re-use the same fields.
The idea is to leave the internal perf enabled all the time, like it is
in PX4, and then allow the integration with lttng on top. Next step
would be to runtime enable/disable only the perf counters we are
interested in.
This also changes the structure so it's easy to allow another thread to
pull data from the Perf object. A rw lock protects from addition of new
counters and an atomic unsigned int allows other threads to do a
lockless copy of the data.
In order for this to work the allocation was changed to use a single
memory pool instead of returning a calloc'ed data for each perf counter.
Since most of our counters are of ' elapsed' type, don't bother using a
smaller struct for the 'count' type
If the RPi version detection fails, the standard version is now RPi2/3 instead of RPi1.
I think this is useful, because the RPi1 is not really supported (performance reasons).
Like others, use HAVE_ prefix and name it HAVE_LTTNG_UST to be the same
name as exported by pkg-config While at it remove wrong comment with
_HELLO_TP_H.
Implementation of alloca() is very much architecture and compiler
dependent. Avoid the case in which it could return a non-aligned
pointer, which would mean Thread::_poison_stack() would do the wrong
thing.
../../libraries/AP_HAL_Linux/Thread.cpp: In member function ‘void Linux::Thread::_poison_stack()’:
../../libraries/AP_HAL_Linux/Thread.cpp:63:20: warning: declaration of ‘start’ shadows a member of 'this' [-Wshadow]
uint8_t *end, *start, *curr;
^
Sort include alphabetically and make them in order:
Main header
system headers
library headers
local headers
While reordering, change a include of endian.h to our sparse-endian.h
which is more reliant to toolchain changes.
Running the vehicles we check the stack size doesn't grow too much by
enabling the DEBUG_STACK in the scheduler. Even on 64bit boards the
stack is consistent around 4k. Just to be a little conservative, let it
be a little bit more that that: 256kB.
Since we have RT prio and we call mlock(), the memory for the stack of
each thread is locked in memory. This means we are effectively taking
that much memory. The default stack size varies per distro, but it's
common to have 8MB for 64 bit boards and 4MB for 32 bit boards. Here is
the output of ps -L -o 'comm,rtprio,rss $(pidof arducopter-quad)', showing the
RSS of arducopter-quad before and after this change:
Before:
COMMAND RTPRIO RSS
arducopter-quad 12 46960
sched-timer 15 46960
sched-uart 14 46960
sched-rcin 13 46960
sched-tonealarm 11 46960
sched-io 10 46960
After:
COMMAND RTPRIO RSS
arducopter-quad 12 7320
sched-timer 15 7320
sched-uart 14 7320
sched-rcin 13 7320
sched-tonealarm 11 7320
sched-io 10 7320
Andrew Tridgell
Staroselskii Georgii
Lucas De Marchi
Luiz Ywata
Gustavo Jose de Sousa
dgrat
Leandro Pereira
Aaron Wang Shi
Ricardo de Almeida Gonzaga