enlightenment/efl
enlightenment
/
efl
8
7
Fork 14
Code Issues 13 Pull Requests 3 Projects Releases Wiki Activity
efl/legacy/evas/src/modules/engines/software_x11/evas_xlib_outbuf.c

1097 lines
32 KiB
C
Raw Normal View History

Evas: Don't #define _GNU_SOURCE Instead use AC_GNU_SOURCE macro in configure.ac and #include <config.h> properly SVN revision: 53159
2010-10-07 12:52:47 -07:00
#ifdef HAVE_CONFIG_H
# include <config.h>
#endif
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
#include <sys/time.h>
#include <sys/utsname.h>
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
#include "evas_common.h"
#include "evas_macros.h"
#include "evas_xlib_outbuf.h"
#include "evas_xlib_buffer.h"
#include "evas_xlib_color.h"
typedef struct _Outbuf_Region Outbuf_Region;
struct _Outbuf_Region
{
X_Output_Buffer *xob;
X_Output_Buffer *mxob;
int x;
int y;
int w;
int h;
};
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
static Eina_List *shmpool = NULL;
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
static int shmsize = 0;
cedric's eet and edje patches to use eet dictionaries for strings. SVN revision: 33873
2008-02-29 22:38:09 -08:00
static int shmmemlimit = 10 * 1024 * 1024;
cleanup: simple clean of "comparison between signed and unsigned errors" not much to see here, please review but these are simple changes. SVN revision: 52426
2010-09-18 16:07:31 -07:00
static const unsigned int shmcountlimit = 32;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
static LK(lock_shmpool);
#define SHMPOOL_LOCK() LKL(lock_shmpool)
#define SHMPOOL_UNLOCK() LKU(lock_shmpool)
#else
#define SHMPOOL_LOCK()
#define SHMPOOL_UNLOCK()
#endif
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
static X_Output_Buffer *
_find_xob(Display *d, Visual *v, int depth, int w, int h, int shm, void *data)
{
* evas: break engine API !!! Improvements: Now evas rendering loop is the one responsible to initialize the surface to 0 correctly (taking into account surface alpha and object opacity). This will reduce the number of memset we do. Note: Current software_x11 (xlib and xcb) are buggy. They are copying too much data when the surface use a mask. That's why two memset are left in their code. They could be removed, but we should fix the surface we copy on change (look at mxob user and evas_software_xlib_x_output_buffer_paste). SVN revision: 41206
2009-06-26 06:26:52 -07:00
Eina_List *l, *xl = NULL;
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
X_Output_Buffer *xob = NULL;
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
X_Output_Buffer *xob2;
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
int fitness = 0x7fffffff;
int sz, lbytes, bpp;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
// return evas_software_xlib_x_output_buffer_new(d, v, depth, w, h, shm, data);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
if (!shm)
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
return evas_software_xlib_x_output_buffer_new(d, v, depth, w, h, shm, data);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
if (depth > 1)
{
bpp = depth / 8;
if (bpp == 3) bpp = 4;
lbytes = (((w * bpp) + 3) / 4) * 4;
}
else
lbytes = ((w + 31) / 32) * 4;
sz = lbytes * h;
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
SHMPOOL_LOCK();
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
EINA_LIST_FOREACH(shmpool, l, xob2)
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
{
int szdif;
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
if ((xob2->xim->depth != depth) || (xob2->visual != v) ||
(xob2->display != d))
continue;
szdif = xob2->psize - sz;
if (szdif < 0) continue;
if (szdif == 0)
{
xob = xob2;
xl = l;
goto have_xob;
}
if (szdif < fitness)
{
fitness = szdif;
xob = xob2;
xl = l;
}
}
if ((fitness > (100 * 100)) || (!xob))
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
{
SHMPOOL_UNLOCK();
xob = evas_software_xlib_x_output_buffer_new(d, v, depth, w, h, shm, data);
return xob;
}
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
have_xob:
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
shmpool = eina_list_remove_list(shmpool, xl);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
xob->w = w;
xob->h = h;
xob->bpl = lbytes;
xob->xim->width = xob->w;
xob->xim->height = xob->h;
xob->xim->bytes_per_line = xob->bpl;
shmsize -= xob->psize * (xob->xim->depth / 8);
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
SHMPOOL_UNLOCK();
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
return xob;
}
static void
_unfind_xob(X_Output_Buffer *xob, int sync)
{
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
// evas_software_xlib_x_output_buffer_free(xob, sync); return;
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
if (xob->shm_info)
{
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
SHMPOOL_LOCK();
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
shmpool = eina_list_prepend(shmpool, xob);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
shmsize += xob->psize * xob->xim->depth / 8;
cedric's eet and edje patches to use eet dictionaries for strings. SVN revision: 33873
2008-02-29 22:38:09 -08:00
while ((shmsize > (shmmemlimit)) ||
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
(eina_list_count(shmpool) > shmcountlimit))
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
{
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
Eina_List *xl;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
xl = eina_list_last(shmpool);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
if (!xl)
{
shmsize = 0;
break;
}
xob = xl->data;
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
shmpool = eina_list_remove_list(shmpool, xl);
shmsize -= xob->psize * xob->xim->depth / 8;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_free(xob, sync);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
}
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
SHMPOOL_UNLOCK();
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
else
evas_software_xlib_x_output_buffer_free(xob, sync);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
}
static void
_clear_xob(int sync)
{
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
SHMPOOL_LOCK();
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
while (shmpool)
{
X_Output_Buffer *xob;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
xob = shmpool->data;
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
shmpool = eina_list_remove_list(shmpool, shmpool);
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_free(xob, sync);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
}
shmsize = 0;
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
SHMPOOL_UNLOCK();
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
}
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
void
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_init(void)
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
{
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
LKI(lock_shmpool);
#endif
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
}
void
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_free(Outbuf *buf)
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
{
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
LKL(buf->priv.lock);
#endif
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
while (buf->priv.pending_writes)
{
RGBA_Image *im;
Outbuf_Region *obr;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
im = buf->priv.pending_writes->data;
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
buf->priv.pending_writes = eina_list_remove_list(buf->priv.pending_writes, buf->priv.pending_writes);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
obr = im->extended_info;
* Make RGBA_IMAGE_HAS_ALPHA and RGBA_IMAGE_ALPHA_SPARSE part of Image_Entry flag structure. This fix a bug with 16 bpp software engine. * Change image loader module API to take any Image_Entry. Same goes for evas_common_image_premul and evas_common_image_set_alpha_sparse. * Use new eet API: eet_data_image_read_to_surface. SVN revision: 34728
2008-06-03 02:09:39 -07:00
evas_cache_image_drop(&im->cache_entry);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
if (obr->xob) _unfind_xob(obr->xob, 0);
if (obr->mxob) _unfind_xob(obr->mxob, 0);
free(obr);
}
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
LKU(buf->priv.lock);
#endif
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_idle_flush(buf);
evas_software_xlib_outbuf_flush(buf);
if (buf->priv.x11.xlib.gc)
XFreeGC(buf->priv.x11.xlib.disp, buf->priv.x11.xlib.gc);
if (buf->priv.x11.xlib.gcm)
XFreeGC(buf->priv.x11.xlib.disp, buf->priv.x11.xlib.gcm);
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
if (buf->priv.pal)
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_color_deallocate(buf->priv.x11.xlib.disp, buf->priv.x11.xlib.cmap,
buf->priv.x11.xlib.vis, buf->priv.pal);
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
LKD(buf->priv.lock);
#endif
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
free(buf);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
_clear_xob(0);
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
Outbuf *
evas_software_xlib_outbuf_setup_x(int w, int h, int rot, Outbuf_Depth depth,
formatting SVN revision: 31929
2007-10-01 20:41:19 -07:00
Display *disp, Drawable draw, Visual *vis,
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
Colormap cmap, int x_depth,
add a dst alpha flag SVN revision: 21104
2006-03-08 07:35:05 -08:00
int grayscale, int max_colors, Pixmap mask,
int shape_dither, int destination_alpha)
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
{
Outbuf *buf;
buf = calloc(1, sizeof(Outbuf));
if (!buf)
do not free buf if no available buf SVN revision: 25694
2006-09-10 09:06:51 -07:00
return NULL;
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
buf->w = w;
buf->h = h;
buf->depth = depth;
buf->rot = rot;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
buf->priv.x11.xlib.disp = disp;
buf->priv.x11.xlib.vis = vis;
buf->priv.x11.xlib.cmap = cmap;
buf->priv.x11.xlib.depth = x_depth;
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
buf->priv.mask_dither = shape_dither;
add a dst alpha flag SVN revision: 21104
2006-03-08 07:35:05 -08:00
buf->priv.destination_alpha = destination_alpha;
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
{
Gfx_Func_Convert conv_func;
X_Output_Buffer *xob;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
buf->priv.x11.xlib.shm = evas_software_xlib_x_can_do_shm(buf->priv.x11.xlib.disp);
xob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
buf->priv.x11.xlib.depth,
1, 1, buf->priv.x11.xlib.shm, NULL);
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
conv_func = NULL;
if (xob)
{
#ifdef WORDS_BIGENDIAN
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if (evas_software_xlib_x_output_buffer_byte_order(xob) == LSBFirst)
buf->priv.x11.xlib.swap = 1;
if (evas_software_xlib_x_output_buffer_bit_order(xob) == MSBFirst)
buf->priv.x11.xlib.bit_swap = 1;
whitespace SVN revision: 14889
2005-05-21 19:49:50 -07:00
#else
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if (evas_software_xlib_x_output_buffer_byte_order(xob) == MSBFirst)
buf->priv.x11.xlib.swap = 1;
if (evas_software_xlib_x_output_buffer_bit_order(xob) == MSBFirst)
buf->priv.x11.xlib.bit_swap = 1;
whitespace SVN revision: 14889
2005-05-21 19:49:50 -07:00
#endif
being an anally retentive bastard.. i checked if e worked in 8bpp under vnc. GASP! it didn't. evas's color allocator was screwed. fixed! :) SVN revision: 24686
2006-08-14 06:50:07 -07:00
if (((vis->class == TrueColor) || (vis->class == DirectColor)) &&
(x_depth > 8))
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
{
buf->priv.mask.r = (DATA32) vis->red_mask;
buf->priv.mask.g = (DATA32) vis->green_mask;
buf->priv.mask.b = (DATA32) vis->blue_mask;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if (buf->priv.x11.xlib.swap)
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
{
SWAP32(buf->priv.mask.r);
SWAP32(buf->priv.mask.g);
SWAP32(buf->priv.mask.b);
}
}
else if ((vis->class == PseudoColor) ||
(vis->class == StaticColor) ||
being an anally retentive bastard.. i checked if e worked in 8bpp under vnc. GASP! it didn't. evas's color allocator was screwed. fixed! :) SVN revision: 24686
2006-08-14 06:50:07 -07:00
(vis->class == GrayScale) ||
(vis->class == StaticGray) ||
(x_depth <= 8))
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
{
Convert_Pal_Mode pm = PAL_MODE_RGB332;
if ((vis->class == GrayScale) || (vis->class == StaticGray))
grayscale = 1;
if (grayscale)
{
if (max_colors >= 256)
pm = PAL_MODE_GRAY256;
else if (max_colors >= 64)
pm = PAL_MODE_GRAY64;
else if (max_colors >= 16)
pm = PAL_MODE_GRAY16;
else if (max_colors >= 4)
pm = PAL_MODE_GRAY4;
else
pm = PAL_MODE_MONO;
}
else
{
if (max_colors >= 256)
pm = PAL_MODE_RGB332;
else if (max_colors >= 216)
pm = PAL_MODE_RGB666;
else if (max_colors >= 128)
pm = PAL_MODE_RGB232;
else if (max_colors >= 64)
pm = PAL_MODE_RGB222;
else if (max_colors >= 32)
pm = PAL_MODE_RGB221;
else if (max_colors >= 16)
pm = PAL_MODE_RGB121;
else if (max_colors >= 8)
pm = PAL_MODE_RGB111;
else if (max_colors >= 4)
pm = PAL_MODE_GRAY4;
else
pm = PAL_MODE_MONO;
}
been doing instrumentation. i'm cuttong down on startup time - a little here and there - noticing lots of excessive io and stat and what not junk. trimming out. SVN revision: 26729
2006-10-21 03:28:01 -07:00
/* FIXME: only alloc once per display+cmap */
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
buf->priv.pal = evas_software_xlib_x_color_allocate(disp, cmap, vis,
use pm instead of PAL_MODE_RGB666 when allocating colors SVN revision: 39188
2009-02-24 23:24:53 -08:00
pm);
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
if (!buf->priv.pal)
{
free(buf);
return NULL;
}
}
if (buf->priv.pal)
{
rotation patches for 180 degrees - thankyoooo! :) SVN revision: 28527
2007-03-02 06:51:17 -08:00
if (buf->rot == 0 || buf->rot == 180)
add a dst alpha flag SVN revision: 21104
2006-03-08 07:35:05 -08:00
conv_func = evas_common_convert_func_get(0, buf->w, buf->h,
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_depth
add a dst alpha flag SVN revision: 21104
2006-03-08 07:35:05 -08:00
(xob), buf->priv.mask.r,
buf->priv.mask.g,
buf->priv.mask.b,
buf->priv.pal->colors,
buf->rot);
rotation patches for 180 degrees - thankyoooo! :) SVN revision: 28527
2007-03-02 06:51:17 -08:00
else if (buf->rot == 90 || buf->rot == 270)
add a dst alpha flag SVN revision: 21104
2006-03-08 07:35:05 -08:00
conv_func = evas_common_convert_func_get(0, buf->h, buf->w,
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_depth
add a dst alpha flag SVN revision: 21104
2006-03-08 07:35:05 -08:00
(xob), buf->priv.mask.r,
buf->priv.mask.g,
buf->priv.mask.b,
buf->priv.pal->colors,
buf->rot);
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
}
else
{
rotation patches for 180 degrees - thankyoooo! :) SVN revision: 28527
2007-03-02 06:51:17 -08:00
if (buf->rot == 0 || buf->rot == 180)
add a dst alpha flag SVN revision: 21104
2006-03-08 07:35:05 -08:00
conv_func = evas_common_convert_func_get(0, buf->w, buf->h,
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_depth
add a dst alpha flag SVN revision: 21104
2006-03-08 07:35:05 -08:00
(xob), buf->priv.mask.r,
buf->priv.mask.g,
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
buf->priv.mask.b, PAL_MODE_NONE,
add a dst alpha flag SVN revision: 21104
2006-03-08 07:35:05 -08:00
buf->rot);
rotation patches for 180 degrees - thankyoooo! :) SVN revision: 28527
2007-03-02 06:51:17 -08:00
else if (buf->rot == 90 || buf->rot == 270)
add a dst alpha flag SVN revision: 21104
2006-03-08 07:35:05 -08:00
conv_func = evas_common_convert_func_get(0, buf->h, buf->w,
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_depth
add a dst alpha flag SVN revision: 21104
2006-03-08 07:35:05 -08:00
(xob), buf->priv.mask.r,
buf->priv.mask.g,
buf->priv.mask.b, PAL_MODE_NONE,
buf->rot);
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_free(xob, 1);
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
if (!conv_func)
{
better logging. * log domains in lower-case only please. let's make it a standard so we don't have to look at the code everytime to figure out the name... * logs do NOT require trailing newline (\n), it's automatic! * do NOT add newline inside log messages! * add gl_common logging. NOTE: I tried to compile all modules, but there are clear broken modules such as cairo and qtopia. Other modules like gl_sdl are broken as they were not updated to new gl_common api (resize method AFAIR). SVN revision: 53174
2010-10-07 16:46:42 -07:00
ERR("At depth: %i, RGB format mask: %08x %08x %08x, "
"Palette mode: %i. "
"Not supported by compiled in converters!",
buf->priv.x11.xlib.depth,
buf->priv.mask.r,
buf->priv.mask.g,
buf->priv.mask.b,
buf->priv.pal ? (int)buf->priv.pal->colors : -1);
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
}
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_drawable_set(buf, draw);
evas_software_xlib_outbuf_mask_set(buf, mask);
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
}
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
LKI(buf->priv.lock);
#endif
code move SVN revision: 6445
2002-11-08 00:02:15 -08:00
return buf;
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
RGBA_Image *
evas_software_xlib_outbuf_new_region_for_update(Outbuf *buf, int x, int y, int w, int h, int *cx, int *cy, int *cw, int *ch)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
RGBA_Image *im;
Outbuf_Region *obr;
int bpl = 0;
int use_shm = 1;
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
int alpha;
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if ((buf->onebuf) && (buf->priv.x11.xlib.shm))
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
{
Evas_Rectangle are gone. We now use Eina_Rectangle and memory pool is shared accross the application. SVN revision: 41170
2009-06-23 06:57:27 -07:00
Eina_Rectangle *rect;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
RECTS_CLIP_TO_RECT(x, y, w, h, 0, 0, buf->w, buf->h);
Evas_Rectangle are gone. We now use Eina_Rectangle and memory pool is shared accross the application. SVN revision: 41170
2009-06-23 06:57:27 -07:00
rect = eina_rectangle_new(x, y, w, h);
if (!rect) return NULL;
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
buf->priv.onebuf_regions = eina_list_append(buf->priv.onebuf_regions, rect);
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
if (buf->priv.onebuf)
{
*cx = x;
*cy = y;
*cw = w;
*ch = h;
if (!buf->priv.synced)
{
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XSync(buf->priv.x11.xlib.disp, False);
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
buf->priv.synced = 1;
}
return buf->priv.onebuf;
}
obr = calloc(1, sizeof(Outbuf_Region));
obr->x = 0;
obr->y = 0;
obr->w = buf->w;
obr->h = buf->h;
*cx = x;
*cy = y;
*cw = w;
*ch = h;
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
alpha = ((buf->priv.x11.xlib.mask) || (buf->priv.destination_alpha));
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
use_shm = buf->priv.x11.xlib.shm;
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
if ((buf->rot == 0) &&
(buf->priv.mask.r == 0xff0000) &&
(buf->priv.mask.g == 0x00ff00) &&
(buf->priv.mask.b == 0x0000ff))
{
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
obr->xob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
buf->priv.x11.xlib.depth,
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
buf->w, buf->h,
use_shm,
NULL);
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
im = (RGBA_Image *) evas_cache_image_data(evas_common_image_cache_get(),
buf->w, buf->h,
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
(DATA32 *) evas_software_xlib_x_output_buffer_data(obr->xob, &bpl),
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
alpha, EVAS_COLORSPACE_ARGB8888);
im->extended_info = obr;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if (buf->priv.x11.xlib.mask)
obr->mxob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
1,
formatting SVN revision: 31929
2007-10-01 20:41:19 -07:00
buf->w, buf->h,
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
use_shm,
NULL);
}
else
{
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
im = (RGBA_Image *) evas_cache_image_empty(evas_common_image_cache_get());
* Make RGBA_IMAGE_HAS_ALPHA and RGBA_IMAGE_ALPHA_SPARSE part of Image_Entry flag structure. This fix a bug with 16 bpp software engine. * Change image loader module API to take any Image_Entry. Same goes for evas_common_image_premul and evas_common_image_set_alpha_sparse. * Use new eet API: eet_data_image_read_to_surface. SVN revision: 34728
2008-06-03 02:09:39 -07:00
im->cache_entry.flags.alpha |= alpha ? 1 : 0;
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
evas_cache_image_surface_alloc(&im->cache_entry, buf->w, buf->h);
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
im->extended_info = obr;
if ((buf->rot == 0) || (buf->rot == 180))
fix softwared x11 xlib engine to generate masks when rotated too SVN revision: 40523
2009-05-06 05:35:38 -07:00
{
obr->xob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
buf->priv.x11.xlib.depth,
buf->w, buf->h,
use_shm,
NULL);
if (buf->priv.x11.xlib.mask)
obr->mxob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
1, buf->w, buf->h,
use_shm,
NULL);
}
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
else if ((buf->rot == 90) || (buf->rot == 270))
fix softwared x11 xlib engine to generate masks when rotated too SVN revision: 40523
2009-05-06 05:35:38 -07:00
{
obr->xob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
buf->priv.x11.xlib.depth,
buf->h, buf->w,
use_shm,
NULL);
if (buf->priv.x11.xlib.mask)
obr->mxob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
1, buf->h, buf->w,
use_shm,
NULL);
}
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
}
* evas: break engine API !!! Improvements: Now evas rendering loop is the one responsible to initialize the surface to 0 correctly (taking into account surface alpha and object opacity). This will reduce the number of memset we do. Note: Current software_x11 (xlib and xcb) are buggy. They are copying too much data when the surface use a mask. That's why two memset are left in their code. They could be removed, but we should fix the surface we copy on change (look at mxob user and evas_software_xlib_x_output_buffer_paste). SVN revision: 41206
2009-06-26 06:26:52 -07:00
/* FIXME: We should be able to remove this memset, but somewhere in the process
we copy too much to the destination surface and some area are not cleaned before copy. */
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
if (alpha)
/* FIXME: faster memset! */
memset(im->image.data, 0, w * h * sizeof(DATA32));
buf->priv.onebuf = im;
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
return im;
}
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
obr = calloc(1, sizeof(Outbuf_Region));
obr->x = x;
obr->y = y;
obr->w = w;
obr->h = h;
*cx = 0;
*cy = 0;
*cw = w;
*ch = h;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
use_shm = buf->priv.x11.xlib.shm;
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
/* FIXME: magic - i found if shm regions are smaller than 200x200 its
* faster to use ximages over unix sockets - trial and error
*/
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
// use_shm = 0; /* 630 -> 1006 fps */
// if ((w * h) < (200 * 200)) use_shm = 0; /* 630 -> 962 fps */
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
alpha = ((buf->priv.x11.xlib.mask) || (buf->priv.destination_alpha));
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
if ((buf->rot == 0) &&
(buf->priv.mask.r == 0xff0000) &&
(buf->priv.mask.g == 0x00ff00) &&
(buf->priv.mask.b == 0x0000ff))
{
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
obr->xob = _find_xob(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
buf->priv.x11.xlib.depth,
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
w, h,
use_shm,
NULL);
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
/* obr->xob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp, */
/* buf->priv.x11.xlib.vis, */
/* buf->priv.x11.xlib.depth, */
/* w, h, */
/* use_shm, */
/* NULL); */
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
im = (RGBA_Image *) evas_cache_image_data(evas_common_image_cache_get(),
w, h,
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
(DATA32 *) evas_software_xlib_x_output_buffer_data(obr->xob, &bpl),
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
alpha, EVAS_COLORSPACE_ARGB8888);
im->extended_info = obr;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if (buf->priv.x11.xlib.mask)
obr->mxob = _find_xob(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
1, w, h,
use_shm,
NULL);
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
/*
obr->mxob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
whitespace SVN revision: 14889
2005-05-21 19:49:50 -07:00
1, w, h,
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
use_shm,
NULL);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
*/
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
else
{
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
im = (RGBA_Image *) evas_cache_image_empty(evas_common_image_cache_get());
* Make RGBA_IMAGE_HAS_ALPHA and RGBA_IMAGE_ALPHA_SPARSE part of Image_Entry flag structure. This fix a bug with 16 bpp software engine. * Change image loader module API to take any Image_Entry. Same goes for evas_common_image_premul and evas_common_image_set_alpha_sparse. * Use new eet API: eet_data_image_read_to_surface. SVN revision: 34728
2008-06-03 02:09:39 -07:00
im->cache_entry.flags.alpha |= alpha ? 1 : 0;
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
evas_cache_image_surface_alloc(&im->cache_entry, w, h);
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
im->extended_info = obr;
if ((buf->rot == 0) || (buf->rot == 180))
fix softwared x11 xlib engine to generate masks when rotated too SVN revision: 40523
2009-05-06 05:35:38 -07:00
{
obr->xob = _find_xob(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
buf->priv.x11.xlib.depth,
w, h,
use_shm,
NULL);
if (buf->priv.x11.xlib.mask)
obr->mxob = _find_xob(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
1, w, h,
use_shm,
NULL);
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
/*
obr->xob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
buf->priv.x11.xlib.depth,
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
w, h,
use_shm,
NULL);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
*/
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
else if ((buf->rot == 90) || (buf->rot == 270))
fix softwared x11 xlib engine to generate masks when rotated too SVN revision: 40523
2009-05-06 05:35:38 -07:00
{
obr->xob = _find_xob(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
buf->priv.x11.xlib.depth,
h, w,
use_shm,
NULL);
if (buf->priv.x11.xlib.mask)
obr->mxob = _find_xob(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
1, h, w,
use_shm,
NULL);
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
/*
obr->xob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
buf->priv.x11.xlib.depth,
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
h, w,
use_shm,
NULL);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
*/
fix softwared x11 xlib engine to generate masks when rotated too SVN revision: 40523
2009-05-06 05:35:38 -07:00
/*
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if (buf->priv.x11.xlib.mask)
obr->mxob = _find_xob(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
1, w, h,
use_shm,
NULL);
fix softwared x11 xlib engine to generate masks when rotated too SVN revision: 40523
2009-05-06 05:35:38 -07:00
*/
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
/*
obr->mxob = evas_software_xlib_x_output_buffer_new(buf->priv.x11.xlib.disp,
buf->priv.x11.xlib.vis,
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
1, w, h,
use_shm,
NULL);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
*/
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
* evas: break engine API !!! Improvements: Now evas rendering loop is the one responsible to initialize the surface to 0 correctly (taking into account surface alpha and object opacity). This will reduce the number of memset we do. Note: Current software_x11 (xlib and xcb) are buggy. They are copying too much data when the surface use a mask. That's why two memset are left in their code. They could be removed, but we should fix the surface we copy on change (look at mxob user and evas_software_xlib_x_output_buffer_paste). SVN revision: 41206
2009-06-26 06:26:52 -07:00
/* FIXME: We should be able to remove this memset, but somewhere in the process
we copy too much to the destination surface and some area are not cleaned before copy. */
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if ((buf->priv.x11.xlib.mask) || (buf->priv.destination_alpha))
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
/* FIXME: faster memset! */
memset(im->image.data, 0, w * h * sizeof(DATA32));
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
if (!evas_common_frameq_enabled())
#endif
buf->priv.pending_writes = eina_list_append(buf->priv.pending_writes, im);
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
return im;
}
void
* remove 'unused parameter' warnings * i didn't touch glitz engine as it will certainly be disabled later * i'll update windows engines later SVN revision: 39274
2009-02-28 02:08:45 -08:00
evas_software_xlib_outbuf_free_region_for_update(Outbuf *buf __UNUSED__, RGBA_Image *update __UNUSED__)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
/* no need to do anything - they are cleaned up on flush */
}
void
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_flush(Outbuf *buf)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
Eina_List *l;
RGBA_Image *im;
Outbuf_Region *obr;
whitespace SVN revision: 14889
2005-05-21 19:49:50 -07:00
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
if ((buf->priv.onebuf) && (buf->priv.onebuf_regions))
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
Region tmpr;
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
im = buf->priv.onebuf;
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
obr = im->extended_info;
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
tmpr = XCreateRegion();
while (buf->priv.onebuf_regions)
{
Evas_Rectangle are gone. We now use Eina_Rectangle and memory pool is shared accross the application. SVN revision: 41170
2009-06-23 06:57:27 -07:00
Eina_Rectangle *rect;
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
XRectangle xr;
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
rect = buf->priv.onebuf_regions->data;
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
buf->priv.onebuf_regions = eina_list_remove_list(buf->priv.onebuf_regions, buf->priv.onebuf_regions);
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
xr.x = rect->x;
xr.y = rect->y;
xr.width = rect->w;
xr.height = rect->h;
XUnionRectWithRegion(&xr, tmpr, tmpr);
if (buf->priv.debug)
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_debug_show(buf, buf->priv.x11.xlib.win,
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
rect->x, rect->y, rect->w, rect->h);
Evas_Rectangle are gone. We now use Eina_Rectangle and memory pool is shared accross the application. SVN revision: 41170
2009-06-23 06:57:27 -07:00
eina_rectangle_free(rect);
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XSetRegion(buf->priv.x11.xlib.disp, buf->priv.x11.xlib.gc, tmpr);
evas_software_xlib_x_output_buffer_paste(obr->xob, buf->priv.x11.xlib.win,
buf->priv.x11.xlib.gc,
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
0, 0, 0);
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
if (obr->mxob)
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
{
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XSetRegion(buf->priv.x11.xlib.disp, buf->priv.x11.xlib.gcm, tmpr);
evas_software_xlib_x_output_buffer_paste(obr->mxob,
buf->priv.x11.xlib.mask,
buf->priv.x11.xlib.gcm,
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
0, 0, 0);
}
XDestroyRegion(tmpr);
buf->priv.synced = 0;
}
else
{
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
#if 1
XSync(buf->priv.x11.xlib.disp, False);
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
EINA_LIST_FOREACH(buf->priv.pending_writes, l, im)
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
{
obr = im->extended_info;
if (buf->priv.debug)
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_debug_show(buf, buf->priv.x11.xlib.win,
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
obr->x, obr->y, obr->w, obr->h);
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_paste(obr->xob, buf->priv.x11.xlib.win,
buf->priv.x11.xlib.gc,
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
obr->x, obr->y, 0);
if (obr->mxob)
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_paste(obr->mxob,
buf->priv.x11.xlib.mask,
buf->priv.x11.xlib.gcm,
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
obr->x, obr->y, 0);
}
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
LKL(buf->priv.lock);
#endif
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
while (buf->priv.prev_pending_writes)
{
im = buf->priv.prev_pending_writes->data;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
buf->priv.prev_pending_writes =
eina_list_remove_list(buf->priv.prev_pending_writes,
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
buf->priv.prev_pending_writes);
obr = im->extended_info;
* Make RGBA_IMAGE_HAS_ALPHA and RGBA_IMAGE_ALPHA_SPARSE part of Image_Entry flag structure. This fix a bug with 16 bpp software engine. * Change image loader module API to take any Image_Entry. Same goes for evas_common_image_premul and evas_common_image_set_alpha_sparse. * Use new eet API: eet_data_image_read_to_surface. SVN revision: 34728
2008-06-03 02:09:39 -07:00
evas_cache_image_drop(&im->cache_entry);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
if (obr->xob) _unfind_xob(obr->xob, 0);
if (obr->mxob) _unfind_xob(obr->mxob, 0);
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
/*
if (obr->xob) evas_software_xlib_x_output_buffer_free(obr->xob, 0);
if (obr->mxob) evas_software_xlib_x_output_buffer_free(obr->mxob, 0);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
*/
free(obr);
}
buf->priv.prev_pending_writes = buf->priv.pending_writes;
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
LKU(buf->priv.lock);
#endif
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
buf->priv.pending_writes = NULL;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XFlush(buf->priv.x11.xlib.disp);
#else
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
/* XX async push - disable */
/*
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
EINA_LIST_FOREACH(buf->priv.pending_writes, l, im)
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
{
obr = im->extended_info;
if (buf->priv.debug)
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_debug_show(buf, buf->priv.x11.xlib.win,
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
obr->x, obr->y, obr->w, obr->h);
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_paste(obr->xob, buf->priv.x11.xlib.win,
buf->priv.x11.xlib.gc,
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
obr->x, obr->y, 0);
if (obr->mxob)
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_paste(obr->mxob,
buf->priv.x11.xlib.mask,
buf->priv.x11.xlib.gcm,
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
obr->x, obr->y, 0);
}
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
*/
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XSync(buf->priv.x11.xlib.disp, False);
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
while (buf->priv.pending_writes)
{
RGBA_Image *im;
Outbuf_Region *obr;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
Remove Evas_List from Evas. SVN revision: 36942
2008-10-21 09:31:05 -07:00
im = eina_list_data_get(buf->priv.pending_writes);
buf->priv.pending_writes = eina_list_remove_list(buf->priv.pending_writes, buf->priv.pending_writes);
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
obr = im->extended_info;
* Make RGBA_IMAGE_HAS_ALPHA and RGBA_IMAGE_ALPHA_SPARSE part of Image_Entry flag structure. This fix a bug with 16 bpp software engine. * Change image loader module API to take any Image_Entry. Same goes for evas_common_image_premul and evas_common_image_set_alpha_sparse. * Use new eet API: eet_data_image_read_to_surface. SVN revision: 34728
2008-06-03 02:09:39 -07:00
evas_cache_image_drop(&im->cache_entry);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
if (obr->xob) _unfind_xob(obr->xob, 0);
if (obr->mxob) _unfind_xob(obr->mxob, 0);
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
/*
if (obr->xob) evas_software_xlib_x_output_buffer_free(obr->xob, 0);
if (obr->mxob) evas_software_xlib_x_output_buffer_free(obr->mxob, 0);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
*/
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
free(obr);
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
evas_cache_image_drop(&im->cache_entry);
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
#endif
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
evas_common_cpu_end_opt();
}
void
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_idle_flush(Outbuf *buf)
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
{
if (buf->priv.onebuf)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
RGBA_Image *im;
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
Outbuf_Region *obr;
whitespace SVN revision: 14889
2005-05-21 19:49:50 -07:00
working on optimising software-x11 with the one-buffer persistence idea that software-16 uses. it works and in some cases gets massive speedups (70%+) but in a few its slowdowns (30% down) in expedite tests - why, i don't know. it should be the same or better in all tests. disabled for now - also not complete. < 32bpp wont' work and not sure rotation works and masks don't work either. SVN revision: 31928
2007-10-01 20:40:14 -07:00
im = buf->priv.onebuf;
buf->priv.onebuf = NULL;
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
obr = im->extended_info;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if (obr->xob) evas_software_xlib_x_output_buffer_free(obr->xob, 0);
if (obr->mxob) evas_software_xlib_x_output_buffer_free(obr->mxob, 0);
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
free(obr);
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
evas_cache_image_drop(&im->cache_entry);
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
else
{
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
LKL(buf->priv.lock);
#endif
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if (buf->priv.prev_pending_writes) XSync(buf->priv.x11.xlib.disp, False);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
while (buf->priv.prev_pending_writes)
{
RGBA_Image *im;
Outbuf_Region *obr;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
im = buf->priv.prev_pending_writes->data;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
buf->priv.prev_pending_writes =
eina_list_remove_list(buf->priv.prev_pending_writes,
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
buf->priv.prev_pending_writes);
obr = im->extended_info;
* Make RGBA_IMAGE_HAS_ALPHA and RGBA_IMAGE_ALPHA_SPARSE part of Image_Entry flag structure. This fix a bug with 16 bpp software engine. * Change image loader module API to take any Image_Entry. Same goes for evas_common_image_premul and evas_common_image_set_alpha_sparse. * Use new eet API: eet_data_image_read_to_surface. SVN revision: 34728
2008-06-03 02:09:39 -07:00
evas_cache_image_drop(&im->cache_entry);
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
if (obr->xob) _unfind_xob(obr->xob, 0);
if (obr->mxob) _unfind_xob(obr->mxob, 0);
free(obr);
}
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
LKU(buf->priv.lock);
#endif
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
_clear_xob(0);
}
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
void
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_push_updated_region(Outbuf *buf, RGBA_Image *update, int x, int y, int w, int h)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
Gfx_Func_Convert conv_func = NULL;
Outbuf_Region *obr;
DATA32 *src_data;
void *data;
masks should be correctly rotated in the xcb engine, now remove a few unused parameters and unused variables SVN revision: 41975
2009-08-25 09:21:09 -07:00
int bpl = 0, yy;
whitespace SVN revision: 14889
2005-05-21 19:49:50 -07:00
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
obr = update->extended_info;
if (buf->priv.pal)
{
if ((buf->rot == 0) || (buf->rot == 180))
conv_func = evas_common_convert_func_get(0, w, h,
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_depth
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
(obr->xob), buf->priv.mask.r,
buf->priv.mask.g, buf->priv.mask.b,
buf->priv.pal->colors, buf->rot);
else if ((buf->rot == 90) || (buf->rot == 270))
conv_func = evas_common_convert_func_get(0, h, w,
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_depth
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
(obr->xob), buf->priv.mask.r,
buf->priv.mask.g, buf->priv.mask.b,
buf->priv.pal->colors, buf->rot);
}
else
{
if ((buf->rot == 0) || (buf->rot == 180))
conv_func = evas_common_convert_func_get(0, w, h,
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_depth
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
(obr->xob), buf->priv.mask.r,
buf->priv.mask.g, buf->priv.mask.b,
PAL_MODE_NONE, buf->rot);
else if ((buf->rot == 90) || (buf->rot == 270))
conv_func = evas_common_convert_func_get(0, h, w,
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_depth
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
(obr->xob), buf->priv.mask.r,
buf->priv.mask.g, buf->priv.mask.b,
PAL_MODE_NONE, buf->rot);
}
if (!conv_func) return;
whitespace SVN revision: 14889
2005-05-21 19:49:50 -07:00
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
data = evas_software_xlib_x_output_buffer_data(obr->xob, &bpl);
cedric's cache changes patch SVN revision: 34253
2008-04-11 17:32:30 -07:00
src_data = update->image.data;
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
if (buf->rot == 0)
{
obr->x = x;
obr->y = y;
}
else if (buf->rot == 90)
{
obr->x = y;
obr->y = buf->w - x - w;
}
else if (buf->rot == 180)
{
obr->x = buf->w - x - w;
obr->y = buf->h - y - h;
}
else if (buf->rot == 270)
{
obr->x = buf->h - y - h;
obr->y = x;
}
if ((buf->rot == 0) || (buf->rot == 180))
{
obr->w = w;
obr->h = h;
}
else if ((buf->rot == 90) || (buf->rot == 270))
{
obr->w = h;
obr->h = w;
}
if (buf->priv.pal)
{
if (data != src_data)
conv_func(src_data, data,
0,
bpl /
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
((evas_software_xlib_x_output_buffer_depth(obr->xob) /
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
8)) - obr->w, obr->w, obr->h, x, y,
buf->priv.pal->lookup);
}
else
{
if (data != src_data)
conv_func(src_data, data,
0,
bpl /
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
((evas_software_xlib_x_output_buffer_depth(obr->xob) /
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
8)) - obr->w, obr->w, obr->h, x, y, NULL);
}
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
#if 1
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
#else
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
/* XX async push */
if (!((buf->priv.onebuf) && (buf->priv.onebuf_regions)))
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
{
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
if (buf->priv.debug)
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_debug_show(buf, buf->priv.x11.xlib.win,
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
obr->x, obr->y, obr->w, obr->h);
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_paste(obr->xob, buf->priv.x11.xlib.win,
buf->priv.x11.xlib.gc,
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
obr->x, obr->y, 0);
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
#endif
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
if (obr->mxob)
{
fix softwared x11 xlib engine to generate masks when rotated too SVN revision: 40523
2009-05-06 05:35:38 -07:00
if (buf->rot == 0)
{
for (yy = 0; yy < obr->h; yy++)
evas_software_xlib_x_write_mask_line(buf, obr->mxob,
src_data +
(yy * obr->w), obr->w, yy);
}
else if (buf->rot == 90)
{
for (yy = 0; yy < obr->h; yy++)
evas_software_xlib_x_write_mask_line_vert(buf, obr->mxob,
src_data + yy,
h, // h
obr->h - yy - 1, // ym
w); // w
}
else if (buf->rot == 180)
{
for (yy = 0; yy < obr->h; yy++)
{
evas_software_xlib_x_write_mask_line_rev(buf, obr->mxob,
src_data +
(yy * obr->w),
obr->w, obr->h - yy - 1);
}
}
else if (buf->rot == 270)
{
for (yy = 0; yy < obr->h; yy++)
evas_software_xlib_x_write_mask_line_vert_rev(buf, obr->mxob,
src_data + yy,
h, // h
yy, // ym
w); // w
}
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
#if 1
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
#else
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
/* XX async push */
if (!((buf->priv.onebuf) && (buf->priv.onebuf_regions)))
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_x_output_buffer_paste(obr->mxob,
buf->priv.x11.xlib.mask,
buf->priv.x11.xlib.gcm,
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
obr->x, obr->y, 0);
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
#endif
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
e: 1. configure/build changes to allow cross-compiling painlessly 2. pager module namespace changes - this was still dirty afdter the namespace cleanup, so clean it up 3. add a powersave subsystem - doesnt have an "automatic" way to turn on and off right now, this i think is best provided by modules (that do things like monitor acpi status's (eg close lid of laptop), AC power status etc. etc. this allows e to nicely defer "power" expensive actions to avoid disk spinups etc. 4. move to use the new ecore poller system - discussed long ago as part of power management/saving issues. now it exists 5. add a canvas idle flush call that helsp cope with the new shm greedy software x11 engine stuff 6. use the new powersave subsystem where appropriate 7. fix non-zeroed/initted memory access in e_fm_main 8. fix mem leak for e menus 9. remove ipc handlers for changed/removed config values 10. use animaotr not timer for menu scrolls - then menu scrolls obey the fps config 11. fix up timer/poll happienss of cursor idle stuff 12. remove avoid damage from popups for now - causing problems 13. change battery and temp readouts to b e shorter so they fit 14. pager can emit signals on focus change for mini-windows now 15. temperature module now uses a slave process and uses stdin/out to talk to it and get output - this makes e smoother as in my expereicne i found getting the temp on my laptop actually took like 200ms so e "hang" for 200ms while reading the acpi files - so now the subprocess does it and just writesa back to e when it gets it. ecore: 1. add ecore_pollers. see the documentation on them in doxygen comments :) 2. fix timers to only go off when they have to - bug there that made e's select time out a LOT more than it needed to. defensive coding hid the problem. now fixed. e should be much more power friendly now. 3. formatting/niceness in ecore_exe stuff 4. some comments on comments with SIGIO ideas vs. select 5. add call to be able to add an idle enterer at the start of the list of them, not just the end (as has been the default) 6. fix ecore_evas to support auto evas idler calls after 0.5 secs of idle in all canvases - and to do it right 7. if argb destination - set the shape EVENT shape (to mask out events in transparent regions much like shape does withotu translucency) 8. in ecore_x add support for the event shape evas: 1. fix cache to work properly and not just always fill up (as it seemed to like to think cahce useage dropped below 0 when it didnt and thus just over-fill) 2. software x11 engine now ONLY uses shm segments - no ximages over the socket. this ximage hack was there to avoid the 2 round trips involved in setting up an shm image - now i mitigated that wih an shm image cache pool. it keeps shm images around and repurposes them for new update regions if appropriate. this means many fewer shm creates (about 1/100th the number) and since we recycle the memory less 0 memory page filling by the kernel - in the end, i recorded about a 10-20% speedup over the old software x11 engine. simple tests i have seen up to 120% speedups. idle flush now does something - it frees all the cached shm segments. it has a hard-coded limit of 4mb worth of shm segments (or 32 segments - whichever comes first) to keep around. once can never complain much about speedups methinks :). also evas will defer sync until the NEXT frame is written - this means evas can calculate the next frame of data while x dma's/copies the images to the screen at the same time (if you hve a dual core or multi-cpu machnike or your xserver is able to use DMA to copy image data to the screen/video ram then this should see a decent speedup). SVN revision: 33448
2008-01-10 23:33:57 -08:00
#if 1
#else
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XFlush(buf->priv.x11.xlib.disp);
#endif
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
void
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_reconfigure(Outbuf * buf, int w, int h, int rot,
ARGB targets for sw engine supported SVN revision: 21119
2006-03-08 19:51:27 -08:00
Outbuf_Depth depth)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
whitespace SVN revision: 14889
2005-05-21 19:49:50 -07:00
if ((w == buf->w) &&
(h == buf->h) &&
(rot == buf->rot) &&
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
(depth == buf->depth)) return;
buf->w = w;
buf->h = h;
buf->rot = rot;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_idle_flush(buf);
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
int
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_get_width(Outbuf * buf)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
return buf->w;
}
int
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_get_height(Outbuf * buf)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
return buf->h;
}
Outbuf_Depth
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_get_depth(Outbuf * buf)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
return buf->depth;
}
int
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_get_rot(Outbuf * buf)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
return buf->rot;
}
void
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_drawable_set(Outbuf * buf, Drawable draw)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
XGCValues gcv;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if (buf->priv.x11.xlib.win == draw) return;
if (buf->priv.x11.xlib.gc)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XFreeGC(buf->priv.x11.xlib.disp, buf->priv.x11.xlib.gc);
buf->priv.x11.xlib.gc = NULL;
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
buf->priv.x11.xlib.win = draw;
buf->priv.x11.xlib.gc = XCreateGC(buf->priv.x11.xlib.disp, buf->priv.x11.xlib.win, 0, &gcv);
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
void
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_mask_set(Outbuf * buf, Pixmap mask)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
XGCValues gcv;
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
if (buf->priv.x11.xlib.mask == mask) return;
if (buf->priv.x11.xlib.gcm)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XFreeGC(buf->priv.x11.xlib.disp, buf->priv.x11.xlib.gcm);
buf->priv.x11.xlib.gcm = NULL;
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
buf->priv.x11.xlib.mask = mask;
if (buf->priv.x11.xlib.mask)
buf->priv.x11.xlib.gcm = XCreateGC(buf->priv.x11.xlib.disp, buf->priv.x11.xlib.mask, 0, &gcv);
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
void
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_debug_set(Outbuf * buf, int debug)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
buf->priv.debug = debug;
}
void
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
evas_software_xlib_outbuf_debug_show(Outbuf * buf, Drawable draw, int x, int y, int w,
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
int h)
{
int i;
int screen_num = 0;
whitespace SVN revision: 14889
2005-05-21 19:49:50 -07:00
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
int wx, wy;
unsigned int ww, wh, bd, dp;
Window wdum, root;
XWindowAttributes wattr;
whitespace SVN revision: 14889
2005-05-21 19:49:50 -07:00
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XGetGeometry(buf->priv.x11.xlib.disp, draw, &root, &wx, &wy, &ww, &wh, &bd, &dp);
XGetGeometry(buf->priv.x11.xlib.disp, root, &wdum, &wx, &wy, &ww, &wh, &bd, &dp);
XGetWindowAttributes(buf->priv.x11.xlib.disp, root, &wattr);
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
screen_num = XScreenNumberOfScreen(wattr.screen);
}
1. gustavo';s fuzzy rect implementation for updates. fixed so del's work, tuned for best performance on my core2 duo desktop - for now. will check more. also make the yuv colorspace code be a bit more robust and fix leak in gl engine with shaders. SVN revision: 30192
2007-06-02 08:30:59 -07:00
for (i = 0; i < 20; i++)
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
{
Comment out unused variable. Seems the functions that were supposed to use it are commented out, but the variable was left. Fixed. SVN revision: 33659
2008-01-31 17:50:56 -08:00
// XImage *xim;
whitespace SVN revision: 14889
2005-05-21 19:49:50 -07:00
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XSetForeground(buf->priv.x11.xlib.disp, buf->priv.x11.xlib.gc,
BlackPixel(buf->priv.x11.xlib.disp, screen_num));
XFillRectangle(buf->priv.x11.xlib.disp, draw, buf->priv.x11.xlib.gc, x, y, w, h);
XSync(buf->priv.x11.xlib.disp, False);
1. gustavo';s fuzzy rect implementation for updates. fixed so del's work, tuned for best performance on my core2 duo desktop - for now. will check more. also make the yuv colorspace code be a bit more robust and fix leak in gl engine with shaders. SVN revision: 30192
2007-06-02 08:30:59 -07:00
// xim =
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
// XGetImage(buf->priv.x11.xlib.disp, draw, x, y, w, h, 0xffffffff, ZPixmap);
1. gustavo';s fuzzy rect implementation for updates. fixed so del's work, tuned for best performance on my core2 duo desktop - for now. will check more. also make the yuv colorspace code be a bit more robust and fix leak in gl engine with shaders. SVN revision: 30192
2007-06-02 08:30:59 -07:00
// if (xim)
// XDestroyImage(xim);
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XSync(buf->priv.x11.xlib.disp, False);
XSetForeground(buf->priv.x11.xlib.disp, buf->priv.x11.xlib.gc,
WhitePixel(buf->priv.x11.xlib.disp, screen_num));
XFillRectangle(buf->priv.x11.xlib.disp, draw, buf->priv.x11.xlib.gc, x, y, w, h);
XSync(buf->priv.x11.xlib.disp, False);
1. gustavo';s fuzzy rect implementation for updates. fixed so del's work, tuned for best performance on my core2 duo desktop - for now. will check more. also make the yuv colorspace code be a bit more robust and fix leak in gl engine with shaders. SVN revision: 30192
2007-06-02 08:30:59 -07:00
// xim =
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
// XGetImage(buf->priv.x11.xlib.disp, draw, x, y, w, h, 0xffffffff, ZPixmap);
1. gustavo';s fuzzy rect implementation for updates. fixed so del's work, tuned for best performance on my core2 duo desktop - for now. will check more. also make the yuv colorspace code be a bit more robust and fix leak in gl engine with shaders. SVN revision: 30192
2007-06-02 08:30:59 -07:00
// if (xim)
// XDestroyImage(xim);
vtorri's merge of xcb and x11 software engines. SVN revision: 38173
2008-12-17 00:13:49 -08:00
XSync(buf->priv.x11.xlib.disp, False);
clean up software x11 engine a little. buffer converted pixels and copy to screen all at once (less "tearing" on updates), minimise syncs, round trips and remove yet another layer of memcopies if in 24/32bpp and your gfx card thinks in ARGB. per DWORD. :) SVN revision: 14704
2005-05-11 01:52:46 -07:00
}
}
* evas: break engine API !!! Improvements: Now evas rendering loop is the one responsible to initialize the surface to 0 correctly (taking into account surface alpha and object opacity). This will reduce the number of memset we do. Note: Current software_x11 (xlib and xcb) are buggy. They are copying too much data when the surface use a mask. That's why two memset are left in their code. They could be removed, but we should fix the surface we copy on change (look at mxob user and evas_software_xlib_x_output_buffer_paste). SVN revision: 41206
2009-06-26 06:26:52 -07:00
Eina_Bool
evas_software_xlib_outbuf_alpha_get(Outbuf *buf)
{
return buf->priv.x11.xlib.mask;
}
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
#ifdef EVAS_FRAME_QUEUING
void
fix warnings in async render code.. and well.. warn about it - it crashes for me in expedite with 4 cores on x86 at random points. looks like it's bitrotting. though it was relatively tentative to begin with. SVN revision: 53856
2010-10-25 04:27:17 -07:00
evas_software_xlib_outbuf_set_priv(Outbuf *buf, void *cur, void *prev __UNUSED__)
big patch from Samsung SAIT (Advanced research group) for async multi-frame rendering. to turn on: 1. configure with --enable-async-render 2. export EVAS_RENDER_MODE=non-blocking presto. necessitates some api swizzling (thus the expedite. ecore etc. changes) the kind of results you get on a desktop: http://www.rasterman.com/files/evas-async-vs-none.html SVN revision: 49087
2010-05-21 00:10:45 -07:00
{
buf->priv.pending_writes = (Eina_List *)cur;
}
#endif