This has not been used for a while and is not even buildable after our
switch to meson. It was a niche to start with given that it needed the
PS3 OS to run on. I asked for any remaining users at EDD and on the list
but heard nothing. Time to remove.
Signed-off-by: Stefan Schmidt <s.schmidt@samsung.com>
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D10778
Revert "eina: remove no longer used function _eina_thread_queue_msg_block_real_free"
This reverts commit 695b44526c.
Revert "eina/threadqueue: use mempool_del for hash free function"
This reverts commit b0cb3b935a.
Revert "eina_thread_queue: use normal mempools for block allocation"
This reverts commit 14ae3e3dec.
Why? Threadqueue is a highly performance sensitive API.
_eina_thread_queue_msg_block_new() may be called quite often. Doing a
hash lookup to then find a mempool handle to then allocate from was
not the same as what was there and was going to be far more costly.
This would have actual performance impact as we have to compute a hash
and rummage through a hash, hunt for an environment var too. The
original code looked at a spare block pool where blocks *MAY* be of
different sizes (not always the same size so using a mempool is
actually wrong and will stop threadqueue from being able to send
larger messages at all). If you send large messages, larger blocks would
have been allocated and put in this pool. In almost all cases the first
item in the pool would be big enough so we don't hunt and the find pulls
out the first memory, resets the fields that are needed and returns that
block. If it needs a bigger one, it does hunt. This is going to be
rare that such big blocks are needed so I never tried to optimize this
(but it could be done with an array of sizes to make a walk to find
the right sized element cheap if the need arises).
Performance dropped quite a lot. On aarch64 The above mempool usage
dropped message rate from 1037251 msg/sec to 610316. On x86 it was even
worse. It dropped from 2815775 msg/sec to 378653.
So backing this out sees the message rate is 7.4 times faster and on
aarch64 it's 1.7 times faster.
So moving to a mempool was actually just wrong (size is not always the
same). Also this ended up with a mempool of 64k for thread queue blocks even
if we only sent messages sporadically, as opposed to a single 4kb
block. So backing this out saves memory by only having 1 or 2 4k blocks
around most of the time, not a 64k mempool.
So the above patch then follow-on patches were done without accounting
for the performance implications. There were good reasons to do what I
did - because this code was highly tuned even to the point where I
used atomics instead of locks specifically to cut down some contention
overhead. Beware when you change something that there may be steep
performance implications. 7.4 times faster to go back to what was
there is a great example.
close all fd's starting at a given fd and then leving out an exception
list specially passed, if any. useful for fork+exec. this uses it in
efl's fork+exec paths.
@feat
Summary:
cos function is much much more accurate than cosf.
this patch replaces cosf by cos to gain more accuracy.
Reviewers: cedric, jsuya, vtorri
Subscribers: vtorri, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D10695
The local cos and sin functions differ from
the math header cos and sin functions by result values
The 4th decimal place is different.
Computing large numbers can cause errors.
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D10467
Summary:
in task T8288 we concluded that a few APIs need to be adjusted in order
to stabelize animation classes at some point. This also adds a new macro
to eina in order to create EINA_VECTOR2 values more easily.
ref T8288
Reviewers: Jaehyun_Cho, segfaultxavi, zmike
Reviewed By: zmike
Subscribers: cedric, #reviewers, #committers
Tags: #efl
Maniphest Tasks: T8288
Differential Revision: https://phab.enlightenment.org/D10350
This avoid comparison with potentially uninitialized byte.
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Differential Revision: https://phab.enlightenment.org/D10479
The idea is to improve memory locality and hopefully get better cache hit
in general.
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Differential Revision: https://phab.enlightenment.org/D10536
Summary:
I'm tired of typing all this out. it's exhausting.
also add a couple usages internally to verify that this works as expected
@feature
Reviewers: cedric, bu5hm4n, devilhorns
Reviewed By: devilhorns
Subscribers: devilhorns, bu5hm4n, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D10557
This is never actually used since meson generates eina_config.h
from scratch. Apart from the byteswap checks done in the previous
commit, all of the other things already seem to be done by meson,
so just remove it.
For one, eina_config.h.in is never used by anything since meson,
so doing any checks in it and expecting them to work is wrong.
Byteswaps are one place where this is the case, so move the
checks back in their appropriate place.
Summary:
there is not any way to determine that this function has failed other
than to check this pointer, so ensure that it always has a value which
is meaningful in failure cases
Reviewers: cedric
Reviewed By: cedric
Subscribers: #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D10445
fix warning when used in C90 code.
Reviewed-by: Stefan Schmidt <stefan@datenfreihafen.org>
Differential Revision: https://phab.enlightenment.org/D10077
EFL_RUN_IN_TREE is set for all binaries that are executed during build
time, and test suites. With this commit we are ensuring that we are not
sending messages during that time to journald, otherwise we would waste
a lot of time + we would create a lot of unneccessary log messages.
(eina test suite runtime with journald hook: 11.21s without journald
hook: 0.57s)
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Differential Revision: https://phab.enlightenment.org/D10076
INFO is a way too verbose log domain, this lead to the fact that we
spammed journald before the first frame with ~200-300 messages
(depending on the setup) which is quite a lot. Additionally, this also
lead to the fact that we often TIMEOUT our own tests.
Reviewed-by: Stefan Schmidt <stefan@datenfreihafen.org>
Differential Revision: https://phab.enlightenment.org/D10075
eina_value_copy assume the given Eina_Value as target is not initialized
when it copy the source in it. In eina_value_reference_copy we were initialising
before calling eina_value_copy which would then override and leak the internal
pointer no fault of the users.
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Differential Revision: https://phab.enlightenment.org/D9943
Introduced in 9b294d6284 the new eina_file_statgen_* API has a typo in
its disable function. Fixing this before it gets out into a release.
Thanks goes to the https://abi-laboratory.pro service for offering their
great API/ABI reports to the open source communities to help finding such
things!
Differential Revision: https://phab.enlightenment.org/D9962
Summary:
there is a patch that add vpath_resolve() call before eina_file_open()
because `path` arg could be NULL, the patch kills some applications.
this patch add null checking logic on vpath_resolve api
Reviewers: bu5hm4n, cedric, Hermet
Reviewed By: Hermet
Subscribers: #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9748
Summary:
data should be checked regardless of EINA_MAGIC define,
because it is dereferenced in the following logic.
Reviewers: cedric, bu5hm4n, Hermet
Reviewed By: Hermet
Subscribers: #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9750
this should bring our list nodes back to 32bytes (or 16 on 32bit)
which is the rounded up libc allocation size anyway expected. less mem
usage but less safety. i haven't seen any magic complainets to do with
eina list for a while now... so let's see.
Summary:
integrate mman.h to make Evil private to the EFL, as mman.h does not exist on Windows. After a discussion with raster, i include sys/mman.h only on non Windows platform.
One issue, though, is that src/modules/emotion/generic/Emotion_Generic_Plugin.h has inlined functions using mmap()
Test Plan: compilation on Windows
Reviewers: cedric, raster, zmike
Subscribers: #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9542
coverity doesnt like inconsistent behavior of code, so make it
consistent even if pointless as this is during init when we wont have
threads .... yet.
CID 1403903
Summary:
Check if hash is empty before computing the hash key and look inside the
hash to find data.
Note: could have called the eina_hash_population() api but didn't
because of extra function call.
Reviewers: Hermet
Reviewed By: Hermet
Subscribers: cedric, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9531
do it once and remember the result from the first one. drops overhead
for sure by a chunk i actually could see in perf reports like about 1-2%
of cpu...
this is a performance optimization. it brings in a "stat generation".
for now it's disabled by default so we retain previous behavior. this
stops eina file from opening and stating a file every time you open
... it only does it if stat generation is off, or, if the generation
changed since the last time it opened that file. this makes cache hits
not have a 3 syscall cost (open+fstat+close). this optimizes that
lower end of things path. but .. it comes at a cost. if the file
changes before generation ticks over (which this forces to tick over
every time the loop exits idle by default).
now here is something to ask.
1. should we have this on by default and accept the "inexactness"
since you can eina_file_statgen_next() before any call that would do
i/o to force it to look at the real file stat info...
2. should we tick over every idle enter OR every N idle enters or
every frame we render instead? ... i want to avoid getting a timestamp
or having a timer interrupt often... so what should we do?
at least this introduces the idea, some api's and an env var to turn
this on. it definitely cuts down syscalls during things like creation
of widdgets or objects in large batches etc.
Summary: fix compilation on systems where backtrace() in execinfo.h is not available
Test Plan: compilation on Windows
Reviewers: zmike, cedric, raster
Reviewed By: cedric, raster
Subscribers: #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9392
on linux open supports() O_CLOEXEC. add test case for this in meson
build and ifdef. this rolls 3 syscalls into 1 as we were doing open,
then fnctl to get and fcntl to set flags.
less syscalls is a good thing as syscalls are not cheap on some
architectures or systems. I've seen a syscall on 1 system take 2-3x
as long as another and another syscall in the same 2 system
comparison take 10x as long. depending on the syscall you may only
have a budget of something like 5000 syscalls "per frame" (60fps)
before you spend all of your frame time just in syscalls not
doing any processing, so we should keep these down if possible
and that is what this does.
it doesn't matter if this doesn't work with eina_debug since this isn't
meant to be used for that kind of debugging
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9359
On Windows, one must be in kernel mode to obtain informations of other users
Reviewed-by: Carsten Haitzler (Rasterman) <rasterman.com>
Reviewed-by: Marcel Hollerbach <mail@marcel-hollerbach.de>
Differential Revision: https://phab.enlightenment.org/D9339
This is a work around compiler/linker limit on some system as reported
by Romain Naour.
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Differential Revision: https://phab.enlightenment.org/D9348
Summary:
this simplifies the process of manipulating lists inside hashes with a
non-copied key
@feature
Depends on D9205
Reviewers: cedric
Reviewed By: cedric
Subscribers: #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9206
Summary:
when install is not set, meson take the value of install_dir. So when
this is removed, things will continue working correctly.
Reviewers: zmike, stefan_schmidt, cedric, segfaultxavi
Reviewed By: zmike
Subscribers: #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9065
Summary:
this function is commonly used to detect the existence of files using
its return value. for this purpose, printing warnings any time the file
cannot be opened is spammy and not very helpful in the context of detecting
actual errors.
Depends on D8950
Reviewers: devilhorns
Reviewed By: devilhorns
Subscribers: cedric, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D8951
Summary: this means we don't need to handle this in the buildtool at all.
Reviewers: vtorri, zmike
Reviewed By: zmike
Subscribers: cedric, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D8553
This enable vpath to recognize also ${} as a variable. It does mimic
what Efl.ViewModel provide with Efl.ViewModel.PropertyText and various
other language.
Reviewed-by: Marcel Hollerbach <mail@marcel-hollerbach.de>
Differential Revision: https://phab.enlightenment.org/D8883