uses the same mutex macros used by the mutex on font objects, so it makes it
a bit simpler. old code is commented out for reference.
SVN revision: 39458
* mainly unused parameters
* in src/lib/imaging/evas_imaging.c, set font to NULL
* in src/lib/canvas/evas_object_gradient.c, add unititialized member
there are a *lot* of reported warnings by llvm, i'll fix them later
there are also *lots* of unused parameters (compile evas with -W). I'll
fix them later too
SVN revision: 39172
2. make gl engine able to use cutouts - in some cases its faster, some
slower. it's a mixed bag. not sure what to make of it. it's #defined to be
disabled atm.
SVN revision: 39114
* evas/src/lib/engines/common/evas_font.h,
* evas/src/lib/engines/common/evas_font_draw.c,
* evas/src/lib/engines/common/evas_font_load.c,
* evas/src/lib/engines/common/evas_font_query.c: Add cache for font kerning.
This patch give something around 2% for all tests around text in expedite,
except for Textblock Intl where it give a 3 times boost.
Regarding text rendering speed, something is strange when used by evas_pipe.
All tests using Styles are around 40% faster without evas_pipe. 30% faster
for Text Change. But Text Basic 7% slower. So it should be possible to have
faster rendering when using evas_pipe for font rendering.
SVN revision: 38993
For some unknown reason evas was informing EVAS_CALLBACK_* even if the
original call did not changed the internal object state, that is, new
value is already equal to current value.
This is specially costly since Edje, Box, Table and possible other
layout engines will call evas_object_resize(), move(), show(), hide()
even if the state has not changed, assuming evas will ignore the call
(as it does). The real overhead might come if there are listeners
attached to these events, that in turn might do lots of other stuff,
leading to a torrent of useless calls.
I marked it for removal, please test it and uncomment '#define
CALLBACK_NOOPS' to get the old behavior back. It does seems to work
with e17 and edje_editor. If problems appear, let's try to fix the
real problem instead of getting this code back, it's a performance
penalty.
SVN revision: 38955
If one want to load image at a given height or width and the other
dimension should be large enough to make it possible, give -1 as the
other coordinate and this will happen.
SVN revision: 38845
we should just remove entries pending preload from the cache being
shutdown, not all of them.
this is untested as it is hard to force this situation, but should be
more correct than the previous.
SVN revision: 38747
before, when no more images were to be preloaded asynchronously, the
thread exited, but were not collected. This leads to a huge leak if
the process is doing aggressive use of image preloading (ie: photo
wall).
collecting dead threads in a proper way (read: without race
conditions) is a bit harder than keeping just one thread alive,
forever. As we do that for evas_pipe (the renderer), let's do the same
with preload and save code.
SVN revision: 38746
users of buffer engine (ie: e_thumb_main.c) were broken since when
they resize the canvas they would implicitly call engine->setup()
again, which would destroy output and create it again. However the
cache could be destroyed and images using it would be bogus.
This does not happen if the process have other cache users, but
e_thumb is just one canvas live at time.
By reordering, we have the cache reference to go to 2 and then back to
1, not destroying it.
SVN revision: 38739
1 - use inlist as regular list uses non-thread safe mempool;
2 - lock around image loading, so if main thread requests pixels right
before worker thread is loading them, you don't get ie->info.module
to NULL while it would be used (triggered from engines/common).
Maybe this should be handled by a global mutex elsewhere instead of
per-image mutex, but it has more granularity now.
3 - emit "preloaded" callback if it was canceled to be loaded from main
thread.
Please someone review these changes.
SVN revision: 38312
image preload will use modules from threads, there is a possibility to
crash due wrong reference counting.
actually much more can fail, we need to check modules don't keep that
needs exclusive access in globals or per-Evas_Module, but that's
another issue.
TODO: replace spinlocks with atomic operations.
SVN revision: 38309
We had some problems with preload and after running LLVM's CLang
Static Analyser we found out that current->target could be NULL after
loop.
Also fixed some GCC and CLang warnings, kudos to these wonderful tools
that "Saved The Day".
PS: we should put some CLang Static Analyser results so others can
help fix other parts of E.
SVN revision: 38293
evas_object_image_preload() should not use object as const because it
will mdofiy the object state (so it's semantic makes more sense).
if data was already loaded, then callback before ignored it (return).
SVN revision: 38246
fast direct3d engine written by Dmitriy Mazovka. You rock !
* m4/evas_check_engine.m:
* m4/evas_check_loader.m4:
use m4_popdef for each macro (otherwise, fail if aclocal is too old)
* src/lib/canvas/evas_font_dir.c:
include evas_common.h and evas_private.h after Eet.h and Evil.h
so that EAPI is correctly defined
SVN revision: 38244
if image was already preloaded, inform user.
regular use case is to have image hidden, ask for preload and then
show image on callback, if there is no callback, image is never shown.
SVN revision: 38236
Wow, this was tricky to find since it is hard to trigger, thanks to
Canola complex edje files we could spot it!
In some cases we end with object being marked as dirty while
calculating its state (ie: edje), then we need to run smart calculate
again.
This has a drawback however: we cannot check for need_recalculate()
inside smart calculate anymore, we must assume it is only called if
the flag is set. To avoid that we could mark a shadow member and use
that or use a counter, that has the problem of using more data.
SVN revision: 38108
Eina hash api must get non NULL pointer allocated with
eina_hash_new(), but Evas hash started with NULL and would allocate
and destroy the hash as required by operations.
To do a proper wrapper we must ensure we don't call Eina hash API with
NULL, we must handle that outside Eina.
PLEASE do not remove this code again (it's the second time I add it),
this is the correct approach. Other than that is going after evas_hash
usage and converting directly to eina_hash.
SVN revision: 38091