Re: [e-users] How to make enlightenment more battery friendly?
ok some numbers - this is intel driver, gl, with partial rendering enabled and 1 window with a spinny busy animation: http://www.enlightenment.org/ss/e-53d89ef1d9efe3.65777916.png 5% cpu with 1 window having a busy spinner. gl engine for compositor. perf says this is how cpu usage is distributed within the enlightenment process: http://www.enlightenment.org/ss/e-53d89f67eb5bb6.40053571.png now if we turn partial update off (i forced it on above, as opposed to default auto which detects buffer age extension automatically and uses the correct info from gl). http://www.enlightenment.org/ss/e-53d8a020545aa4.08083375.png 19% cpu now. 4x as much. with profile of: http://www.enlightenment.org/ss/e-53d8a07f6fc960.37155853.png note these numbers are about 15-20% higher (17 vs 20% cpu) because i'm runing a perf top at 5 hz - this seems to raise cpu usage by about 15-20% for processes. so drop back down by that for normal situations. either way - as you can see. partial rendering uses 1/4 or so the cpu of re-rendering everything. the optimization is already there and done - if the driver supports it, it's automatically used without further ado. if not - you get full redraws every frame. this means evas walks a lot more canvas content, generates a lot more geometry etc. etc. we already have the optimizations there, but the extension support is not. even the people behind the extension we use don't implement it. their major gpu competitor does though. on glx anyway. you can try egl+gles and see if it's there. i'm not sure. but that will drop your cpu usage a lot. i am actually ASSUMING and expecting this will happen - or eventually will. you're getting a slow path atm due to drivers. the cpu use in canvas walking as before (the top 1 items) is a matter of a canvas render rewrite of 10 year old code. i have started with some skeleton code but getting time to do it is tough. a simpler theme will help little (especially with gl). if i get some time, i'll try find a way to add some numbers or debug to track # of triangles per frame and put an older e17 install in to collect them - better yet even to draw outlines of the triangles too and get a screenshot so you can see how we drive geometry differently. -- - Codito, ergo sum - I code, therefore I am -- The Rasterman (Carsten Haitzler)ras...@rasterman.com -- Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with a free copy of Black Duck Code Sight - the same software that powers the world's largest code search on Ohloh, the Black Duck Open Hub! Try it now. http://p.sf.net/sfu/bds ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Mon, Aug 04, 2014 at 07:29:59AM +0900, Carsten Haitzler wrote: ok some numbers - this is intel driver, gl, with partial rendering enabled and 1 window with a spinny busy animation: http://www.enlightenment.org/ss/e-53d89ef1d9efe3.65777916.png 5% cpu with 1 window having a busy spinner. gl engine for compositor. perf says this is how cpu usage is distributed within the enlightenment process: http://www.enlightenment.org/ss/e-53d89f67eb5bb6.40053571.png Interesting. Just curious, how do you turn on partial rendering? Is that in composite/rendering? Also, I looked in advanced/power managemnt How do you tell e which level it should be using? 19% cpu now. 4x as much. with profile of: http://www.enlightenment.org/ss/e-53d8a07f6fc960.37155853.png note these numbers are about 15-20% higher (17 vs 20% cpu) because i'm runing a perf top at 5 hz - this seems to raise cpu usage by about 15-20% for processes. so drop back down by that for normal situations. either way - as you can see. partial rendering uses 1/4 or so the cpu of re-rendering everything. the optimization is already there and done - if the Very interesting find, thanks for sharing. Marc -- A mouse is a device used to point at the xterm you want to type in - A.S.R. Microsoft is to operating systems what McDonalds is to gourmet cooking Home page: http://marc.merlins.org/ | PGP 1024R/763BE901 -- Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with a free copy of Black Duck Code Sight - the same software that powers the world's largest code search on Ohloh, the Black Duck Open Hub! Try it now. http://p.sf.net/sfu/bds ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Sun, 3 Aug 2014 16:01:21 -0700 Marc MERLIN marc_...@merlins.org said:
On Mon, Aug 04, 2014 at 07:29:59AM +0900, Carsten Haitzler wrote:
ok some numbers - this is intel driver, gl, with partial rendering enabled
and 1 window with a spinny busy animation:
http://www.enlightenment.org/ss/e-53d89ef1d9efe3.65777916.png
5% cpu with 1 window having a busy spinner. gl engine for compositor. perf
says this is how cpu usage is distributed within the enlightenment process:
http://www.enlightenment.org/ss/e-53d89f67eb5bb6.40053571.png
Interesting.
Just curious, how do you turn on partial rendering?
comp settings - in advanced under rendering - that's turning it on manually
ignoring what the driver actualy does. - you may end up with flickering and
bugs as the driver sometimes re-orders buffer and evas then has no idea. it's
automatic by default in evas if you have the right extensions. as i also
mentioned egl/gles may support it where glx does not on intel - so go rebuild
evas for gles and see. (mesa offers glx and egl/gles).
Is that in composite/rendering?
Also, I looked in advanced/power managemnt
How do you tell e which level it should be using?
leave that alone.
19% cpu now. 4x as much. with profile of:
http://www.enlightenment.org/ss/e-53d8a07f6fc960.37155853.png
note these numbers are about 15-20% higher (17 vs 20% cpu) because i'm
runing a perf top at 5 hz - this seems to raise cpu usage by about
15-20% for processes. so drop back down by that for normal situations.
either way - as you can see. partial rendering uses 1/4 or so the cpu of
re-rendering everything. the optimization is already there and done - if the
Very interesting find, thanks for sharing.
Marc
--
A mouse is a device used to point at the xterm you want to type in - A.S.R.
Microsoft is to operating systems
what McDonalds is to gourmet
cooking Home page: http://marc.merlins.org/ | PGP
1024R/763BE901
--
- Codito, ergo sum - I code, therefore I am --
The Rasterman (Carsten Haitzler)ras...@rasterman.com
--
Infragistics Professional
Build stunning WinForms apps today!
Reboot your WinForms applications with our WinForms controls.
Build a bridge from your legacy apps to the future.
http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk
___
enlightenment-users mailing list
enlightenment-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Thu, Jul 31, 2014 at 3:38 PM, Mick michaelkintz...@gmail.com wrote: I use opensource radeon drivers with various AMD cards from ancient to modern and never had any problems with them - although there is the odd kernel regression bug here and there. I came across an Intel graphics card and the performance was nowhere near as smooth as with the AMD. I have to agree with this - mostly. There was a nasty time back there - starting from the r300 days - and apparently ending some time I failed to notice - where both drivers ( ATI's and open-source ) were pretty bad, and Intel cards just worked ( albeit slowly ). I even swore off buying any more ATI graphics products ( and the lack of open-source nVidia drivers mean I'm not really interested in them ). I now have an i5 laptop and and AMD server ( A8-5600K ) and for me, the situation is well and truly reversed. The AMD setup is rock solid ( open-source - never tried the AMD drivers ). My son plays supertuxkart, stunt rally, and even a windows port of wizball running under wine for hours on end without an issue. I can't do that on my i5 laptop - I've tried playing stunt rally networked with him, and my i5 lasts about 5 minutes before locking up hard. And there's the obvious performance difference too - the AMD setup blows the Intel one out of the water. I guess people have different use cases and different experiences. But there is absolutely no way I'll buy another Intel graphics product if I can find an AMD one in a similar form ( which can be difficult with laptops, unfortunately ). -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Thu, 31 Jul 2014 06:38:18 +0100 Mick michaelkintz...@gmail.com said: On Thursday 31 Jul 2014 00:10:21 Carsten Haitzler wrote: the only gfx card i don't touch is anything from ati/amd - that's on my blacklist of evil. if someone has a bad day there, i would be totally unsurprised. i've been bitten before enough to now totally avoid them. I use opensource radeon drivers with various AMD cards from ancient to modern and never had any problems with them - although there is the odd kernel it has for me - when i had a 4650hd - radeon didn't support it (it simply didn't work). :) regression bug here and there. I came across an Intel graphics card and the performance was nowhere near as smooth as with the AMD. wonderful mem leak from a user that has radeon allocate all memory until exhausted: http://i.imgur.com/Qasy2Du.jpg only on radeon. nvidia, intel don't have it. :) -- - Codito, ergo sum - I code, therefore I am -- The Rasterman (Carsten Haitzler)ras...@rasterman.com -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Thu, 31 Jul 2014 18:09:16 +1000 Daniel Kasak d.j.kasak...@gmail.com wrote: I even swore off buying any more ATI graphics products ( and the lack of open-source nVidia drivers mean I'm not really interested in them ). There is an open-source driver for nVidia: nouveau. It's been around for some years now. It's not the fasttest thing in the world, but it works. --Dale -- Imagine if every Thursday your shoes exploded if you tied them the usual way. This happens to us all the time with computers, and nobody thinks of complaining. -- Jeff Raskin -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Wed, Jul 30, 2014 at 1:23 PM, Mick michaelkintz...@gmail.com wrote: On Wednesday 30 Jul 2014 01:37:13 P Purkayastha wrote: Installed versions: 10.0.4(12:16:49 AM 04/11/2014)(classic egl gallium nptl xvmc -bindist -debug -gbm -gles1 -gles2 -llvm -llvm-shared-libs -opencl -openvg -osmesa -pax_kernel -pic -r600-llvm-compiler -selinux -vdpau -wayland -xa ABI_MIPS=-n32 -n64 -o32 ABI_X86=64 -32 -x32 KERNEL=-FreeBSD VIDEO_CARDS=intel -freedreno -i915 -i965 -ilo -nouveau -r100 -r200 -r300 -r600 -radeon -radeonsi -vmware) Shouldn't you have i965 also set in VIDEO_CARDS? Excellent! Thank you! This fixes the problems I have been having on my system. It is really very weird why intel is present as is. There should have been something more descriptive put in its place. Hardware acceleration, glx, accelerated video, e17 had all been working fine till now so I never suspected my VIDEO_CARDS setting. I don't play games, otherwise I might have noticed this anomaly earlier. -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Thu, 31 Jul 2014 00:54:05 +0800 P Purkayastha ppu...@gmail.com said: On Wed, Jul 30, 2014 at 1:23 PM, Mick michaelkintz...@gmail.com wrote: On Wednesday 30 Jul 2014 01:37:13 P Purkayastha wrote: Installed versions: 10.0.4(12:16:49 AM 04/11/2014)(classic egl gallium nptl xvmc -bindist -debug -gbm -gles1 -gles2 -llvm -llvm-shared-libs -opencl -openvg -osmesa -pax_kernel -pic -r600-llvm-compiler -selinux -vdpau -wayland -xa ABI_MIPS=-n32 -n64 -o32 ABI_X86=64 -32 -x32 KERNEL=-FreeBSD VIDEO_CARDS=intel -freedreno -i915 -i965 -ilo -nouveau -r100 -r200 -r300 -r600 -radeon -radeonsi -vmware) Shouldn't you have i965 also set in VIDEO_CARDS? Excellent! Thank you! This fixes the problems I have been having on my system. It is really very weird why intel is present as is. There should have been something more descriptive put in its place. so you can stop blaming enlightenment/efl then eh? :) gr. funroll-loops. Hardware acceleration, glx, accelerated video, e17 had all been working fine till now so I never suspected my VIDEO_CARDS setting. I don't play games, otherwise I might have noticed this anomaly earlier. others of us use packages pre-built with all the proper accel there - like me. we see a smooth experience. we have no idea what you are speaking of. the only gfx card i don't touch is anything from ati/amd - that's on my blacklist of evil. if someone has a bad day there, i would be totally unsurprised. i've been bitten before enough to now totally avoid them. -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users -- - Codito, ergo sum - I code, therefore I am -- The Rasterman (Carsten Haitzler)ras...@rasterman.com -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Thursday 31 Jul 2014 00:10:21 Carsten Haitzler wrote: the only gfx card i don't touch is anything from ati/amd - that's on my blacklist of evil. if someone has a bad day there, i would be totally unsurprised. i've been bitten before enough to now totally avoid them. I use opensource radeon drivers with various AMD cards from ancient to modern and never had any problems with them - although there is the odd kernel regression bug here and there. I came across an Intel graphics card and the performance was nowhere near as smooth as with the AMD. -- Regards, Mick -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Mon, 28 Jul 2014 21:52:28 -0700 Marc MERLIN marc_...@merlins.org said: On Tue, Jul 29, 2014 at 08:53:05AM +0900, Carsten Haitzler wrote: I understand that, but would it make sense to have an e config option for less bling, more battery that one could switch to by sending a command from e from ACPI? i'm sorry - but no. it's like asking bmw could i have a small button in my car that converts my x6 suv into an i3 electric car? it'd be so much better to roll around town!. it's not so simple. Ah now, if you're putting facts in front of my dreams, we're not going to get anywhere :) e has no choice in this matter. it's a deep complex bi-product of many many many factors - partly canvas itself, its content, design choices and much more. a lot of it is a direct result of input from apps (more text in titlebars == more triangles to render (1 per char), and apps decide the text that will be there - they also decide how many windows you have and min/max size etc.), input from a user (choosing where their windows are, how to lay them out, what So I was thinking about a minimal eye candy mode, like I believe e17 had early on when you set it up. Isn't there a way to disable at least some of the work it does to make things look good? For instance, is it possible to run the main event loop less often? Disable all things that move if not needed like the 'window is not responding spinning stuff' which did improve my battery after I turned it off? I know you have no control over what the 3D drivers do, but you have control over what you feed them :) as well as how often e wakes up. if anything you are asking for evas to do things like cache rendering of objects (individual ones and entire object trees) into textures to avoid more geometry. this is not simple - it is in fact a pain in the butt given current evas internals. you can do this EXPLICITLY via 2 mechanisms in evas - map and proxy. these both have bugs in dark corners that do things like affect the connman. epluse and pkgkit popups where the clipping of content goes wrong as a result. this is basically impossible to fix because fixing this will almost 100% certainly just create a new bug to hunt. the solution is a rewrite of the render core to simplify it and put these concepts into there in stone. but i never have the time to do this, partly due to emails like this, patch reviews, other bug reports and a continual stream of other mentoring, and even more i can't mention. so until then - you have what you have That's fair obviously :) and you also know what's possible and what's not as well as how much some of those are, but somehow I was hoping that there would be low hanging fruits which yield degraded performance but save battery. 1) Aren't there any other such items like the one you just gave me (turn off the spinning circle?) 2) Can't e18 be made to sleep longer between event runs? If I'm not getting it, and just wishful thinking, tell me and I'll go away :) there is a framerate slider - u can slide that down. advanced - performance -- - Codito, ergo sum - I code, therefore I am -- The Rasterman (Carsten Haitzler)ras...@rasterman.com -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Tue, Jul 29, 2014 at 6:52 AM, Marc MERLIN marc_...@merlins.org wrote: On Tue, Jul 29, 2014 at 08:53:05AM +0900, Carsten Haitzler wrote: I understand that, but would it make sense to have an e config option for less bling, more battery that one could switch to by sending a command from e from ACPI? i'm sorry - but no. it's like asking bmw could i have a small button in my car that converts my x6 suv into an i3 electric car? it'd be so much better to roll around town!. it's not so simple. Ah now, if you're putting facts in front of my dreams, we're not going to get anywhere :) e has no choice in this matter. it's a deep complex bi-product of many many many factors - partly canvas itself, its content, design choices and much more. a lot of it is a direct result of input from apps (more text in titlebars == more triangles to render (1 per char), and apps decide the text that will be there - they also decide how many windows you have and min/max size etc.), input from a user (choosing where their windows are, how to lay them out, what So I was thinking about a minimal eye candy mode, like I believe e17 had early on when you set it up. Isn't there a way to disable at least some of the work it does to make things look good? For instance, is it possible to run the main event loop less often? Disable all things that move if not needed like the 'window is not responding spinning stuff' which did improve my battery after I turned it off? I have been thinking for some time that we should have a flat no effect theme, that we could use for this kind of use case. If we had a new kind of action, we could trigger theme change on a few scenario (battery unplug/plug or battery low/charging) and achieve what I think you are looking for. I know you have no control over what the 3D drivers do, but you have control over what you feed them :) as well as how often e wakes up. if anything you are asking for evas to do things like cache rendering of objects (individual ones and entire object trees) into textures to avoid more geometry. this is not simple - it is in fact a pain in the butt given current evas internals. you can do this EXPLICITLY via 2 mechanisms in evas - map and proxy. these both have bugs in dark corners that do things like affect the connman. epluse and pkgkit popups where the clipping of content goes wrong as a result. this is basically impossible to fix because fixing this will almost 100% certainly just create a new bug to hunt. the solution is a rewrite of the render core to simplify it and put these concepts into there in stone. but i never have the time to do this, partly due to emails like this, patch reviews, other bug reports and a continual stream of other mentoring, and even more i can't mention. so until then - you have what you have That's fair obviously :) and you also know what's possible and what's not as well as how much some of those are, but somehow I was hoping that there would be low hanging fruits which yield degraded performance but save battery. 1) Aren't there any other such items like the one you just gave me (turn off the spinning circle?) 2) Can't e18 be made to sleep longer between event runs? Did you try reducing the frame rate already ? Also there is noway to delay when select wake up without having some kind of more active loop. Also on my older i7, switching to CPU rendering was giving me a 10% battery increase. Did you try that ? -- Cedric BAIL -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
Hello, On Sat, Jul 26, 2014 at 3:49 PM, P Purkayastha ppu...@gmail.com wrote: On Sun, Jul 20, 2014 at 10:53 PM, Carsten Haitzler ras...@rasterman.com wrote: On Sun, 20 Jul 2014 14:02:52 +0800 P Purkayastha ppu...@gmail.com said: I have seen e get to high cpu when the spinning wheel starts, especially affects e19 even if the wheel is spinning in a different desktop than the currently visible one. even if not visible - it's waking up anbd changing object state - likely figuring out it's a rendering NOOP and then doing nothing after a canvas object pass. I think you should try out e19 on drivers that do not support the buffer egl extension. It not only takes more CPU when there is a single tiny rotating wheel on the desktop, the whole desktop stutters when changing desks. *Everything* is slower and choppy - desktop transitions, window movement, opening new windows, etc. My laptop is just over a year old (dell xps 13) that has the intel graphics and core i7 cpu just from the generation just preceding haswell. It is not as powerful as nvidia, but it should have plenty of processing power to run e19 without stuttering. The screen on the laptop is of 1920x1080 resolution (if it matters to e19). I have never seen e17 produce this kind of choppy desktop. I am wondering if there is not something else going on with your hardware. I have a 3 years old i7 with intel GPU and a 2 years old one also, both of them are running both GL and software backend pretty smoothly and battery usage was actually lower than other compositor (especially in software). I must say I haven't done any benchmark recently as I didn't felt any behavior change. So maybe there is something else going on with your setup. Did you try to get a valgrind, oprofile or perf trace to see what's going on ? -- Cedric BAIL -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Tue, Jul 29, 2014 at 5:44 PM, Cedric BAIL cedric.b...@free.fr wrote: Hello, On Sat, Jul 26, 2014 at 3:49 PM, P Purkayastha ppu...@gmail.com wrote: On Sun, Jul 20, 2014 at 10:53 PM, Carsten Haitzler ras...@rasterman.com wrote: On Sun, 20 Jul 2014 14:02:52 +0800 P Purkayastha ppu...@gmail.com said: I have seen e get to high cpu when the spinning wheel starts, especially affects e19 even if the wheel is spinning in a different desktop than the currently visible one. even if not visible - it's waking up anbd changing object state - likely figuring out it's a rendering NOOP and then doing nothing after a canvas object pass. I think you should try out e19 on drivers that do not support the buffer egl extension. It not only takes more CPU when there is a single tiny rotating wheel on the desktop, the whole desktop stutters when changing desks. *Everything* is slower and choppy - desktop transitions, window movement, opening new windows, etc. My laptop is just over a year old (dell xps 13) that has the intel graphics and core i7 cpu just from the generation just preceding haswell. It is not as powerful as nvidia, but it should have plenty of processing power to run e19 without stuttering. The screen on the laptop is of 1920x1080 resolution (if it matters to e19). I have never seen e17 produce this kind of choppy desktop. I am wondering if there is not something else going on with your hardware. I have a 3 years old i7 with intel GPU and a 2 years old one also, both of them are running both GL and software backend pretty smoothly and battery usage was actually lower than other compositor (especially in software). I must say I haven't done any benchmark recently as I didn't felt any behavior change. So maybe there is something else going on with your setup. Did you try to get a valgrind, oprofile or perf trace to see what's going on ? -- Cedric BAIL Haven't tried valgrind or other things. My system packages are all stable Gentoo packages (gcc-4.7.3, kernel-3.12.21, X-1.15.0, etc). Only things that are not stable are a few end user packages like firefox, chrome, libreoffice, e19, etc. -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Tue, Jul 29, 2014 at 12:15 PM, P Purkayastha ppu...@gmail.com wrote: On Tue, Jul 29, 2014 at 5:44 PM, Cedric BAIL cedric.b...@free.fr wrote: On Sat, Jul 26, 2014 at 3:49 PM, P Purkayastha ppu...@gmail.com wrote: On Sun, Jul 20, 2014 at 10:53 PM, Carsten Haitzler ras...@rasterman.com wrote: On Sun, 20 Jul 2014 14:02:52 +0800 P Purkayastha ppu...@gmail.com said: I have seen e get to high cpu when the spinning wheel starts, especially affects e19 even if the wheel is spinning in a different desktop than the currently visible one. even if not visible - it's waking up anbd changing object state - likely figuring out it's a rendering NOOP and then doing nothing after a canvas object pass. I think you should try out e19 on drivers that do not support the buffer egl extension. It not only takes more CPU when there is a single tiny rotating wheel on the desktop, the whole desktop stutters when changing desks. *Everything* is slower and choppy - desktop transitions, window movement, opening new windows, etc. My laptop is just over a year old (dell xps 13) that has the intel graphics and core i7 cpu just from the generation just preceding haswell. It is not as powerful as nvidia, but it should have plenty of processing power to run e19 without stuttering. The screen on the laptop is of 1920x1080 resolution (if it matters to e19). I have never seen e17 produce this kind of choppy desktop. I am wondering if there is not something else going on with your hardware. I have a 3 years old i7 with intel GPU and a 2 years old one also, both of them are running both GL and software backend pretty smoothly and battery usage was actually lower than other compositor (especially in software). I must say I haven't done any benchmark recently as I didn't felt any behavior change. So maybe there is something else going on with your setup. Did you try to get a valgrind, oprofile or perf trace to see what's going on ? Haven't tried valgrind or other things. My system packages are all stable Gentoo packages (gcc-4.7.3, kernel-3.12.21, X-1.15.0, etc). Only things that are not stable are a few end user packages like firefox, chrome, libreoffice, e19, etc. What is the exact version of your CPU/GPU and memory information to ? What is your glxinfo information ? -- Cedric BAIL -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Tue, Jul 29, 2014 at 6:54 PM, Cedric BAIL cedric.b...@free.fr wrote:
On Tue, Jul 29, 2014 at 12:15 PM, P Purkayastha ppu...@gmail.com wrote:
Haven't tried valgrind or other things. My system packages are all
stable Gentoo packages (gcc-4.7.3, kernel-3.12.21, X-1.15.0, etc).
Only things that are not stable are a few end user packages like
firefox, chrome, libreoffice, e19, etc.
What is the exact version of your CPU/GPU and memory information to ?
What is your glxinfo information ?
cpu: i7-3537U (/proc/cpuinfo attached)
gpu: intel HD 4000 (see below)
ram: 8G
glxinfo: attached.
Some other info:
00:02.0 VGA compatible controller: Intel Corporation 3rd Gen Core
processor Graphics Controller (rev 09) (prog-if 00 [VG
A controller])
Subsystem: Dell Device 058b
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-
ParErr- Stepping- SERR- FastB2B- DisINTx+
Status: Cap+ 66MHz- UDF- FastB2B+ ParErr- DEVSEL=fast TAbort-
TAbort- MAbort- SERR- PERR- INTx-
Latency: 0
Interrupt: pin A routed to IRQ 40
Region 0: Memory at d000 (64-bit, non-prefetchable) [size=4M]
Region 2: Memory at c000 (64-bit, prefetchable) [size=256M]
Region 4: I/O ports at 2000 [size=64]
Expansion ROM at unassigned [disabled]
Capabilities: [90] MSI: Enable+ Count=1/1 Maskable- 64bit-
Address: fee0f00c Data: 4171
Capabilities: [d0] Power Management version 2
Flags: PMEClk- DSI+ D1- D2- AuxCurrent=0mA
PME(D0-,D1-,D2-,D3hot-,D3cold-)
Status: D0 NoSoftRst- PME-Enable- DSel=0 DScale=0 PME-
Capabilities: [a4] PCI Advanced Features
AFCap: TP+ FLR+
AFCtrl: FLR-
AFStatus: TP-
Kernel driver in use: i915
/tmp» free -m
total used free sharedbuffers cached
Mem: 7835 7206629878 84 5027
-/+ buffers/cache: 2094 5741
Swap: 1999 18 1981
/tmp» uname -r
3.12.21-gentoo-r1
/tmp» eix -e mesa-progs
[I] x11-apps/mesa-progs
Available versions: 8.0.1 8.1.0 {egl gles1 gles2}
Installed versions: 8.1.0(01:54:12 PM 07/20/2014)(-egl -gles1 -gles2)
Homepage:http://mesa3d.sourceforge.net/
Description: Mesa's OpenGL utility and demo programs
(glxgears and glxinfo)
/tmp» eix -I intel
[I] x11-drivers/xf86-video-intel
Available versions: ~*2.9.1 2.19.0 2.20.13 2.21.15 ~2.99.903
~2.99.905-r1 ~2.99.906 ~2.99.907-r1 ~2.99.909 ~2.99.910 ~2.99.911-r1
~2.99.912 ~2.99.912-r1 ~2.99.914 {debug dri glamor (+)sna +udev uxa
xvmc}
Installed versions: 2.21.15(06:55:14 PM 05/28/2014)(dri sna udev
uxa xvmc -glamor)
Homepage:http://xorg.freedesktop.org/
Description: X.Org driver for Intel cards
[I] x11-libs/libva-intel-driver
Available versions: 1.0.17 ~1.0.19 ~1.0.20 ~1.0.20-r1 ~1.2.0
~1.2.1 ~1.2.2 ~1.2.2-r1 1.3.0 ~1.3.2 ** {X +drm wayland
ABI_MIPS=n32 n64 o32 ABI_PPC=32 64 ABI_S390=32 64 ABI_X86=32 64
x32}
Installed versions: 1.3.0(09:04:16 PM 07/19/2014)(X drm -wayland
ABI_MIPS=-n32 -n64 -o32 ABI_PPC=-32 -64 ABI_S390=-32 -64
ABI_X86=64 -32 -x32)
Homepage:http://www.freedesktop.org/wiki/Software/vaapi
Description: HW video decode support for Intel integrated graphics
name of display: :0.0
display: :0 screen: 0
direct rendering: Yes
server glx vendor string: SGI
server glx version string: 1.4
server glx extensions:
GLX_ARB_create_context, GLX_ARB_create_context_profile,
GLX_ARB_multisample, GLX_EXT_create_context_es2_profile,
GLX_EXT_import_context, GLX_EXT_texture_from_pixmap, GLX_EXT_visual_info,
GLX_EXT_visual_rating, GLX_INTEL_swap_event, GLX_MESA_copy_sub_buffer,
GLX_OML_swap_method, GLX_SGIS_multisample, GLX_SGIX_fbconfig,
GLX_SGIX_pbuffer, GLX_SGIX_visual_select_group, GLX_SGI_swap_control
client glx vendor string: Mesa Project and SGI
client glx version string: 1.4
client glx extensions:
GLX_ARB_create_context, GLX_ARB_create_context_profile,
GLX_ARB_create_context_robustness, GLX_ARB_fbconfig_float,
GLX_ARB_framebuffer_sRGB, GLX_ARB_get_proc_address, GLX_ARB_multisample,
GLX_EXT_create_context_es2_profile, GLX_EXT_fbconfig_packed_float,
GLX_EXT_framebuffer_sRGB, GLX_EXT_import_context,
GLX_EXT_texture_from_pixmap, GLX_EXT_visual_info, GLX_EXT_visual_rating,
GLX_INTEL_swap_event, GLX_MESA_copy_sub_buffer,
GLX_MESA_multithread_makecurrent, GLX_MESA_query_renderer,
GLX_MESA_swap_control, GLX_OML_swap_method, GLX_OML_sync_control,
GLX_SGIS_multisample, GLX_SGIX_fbconfig, GLX_SGIX_pbuffer,
GLX_SGIX_visual_select_group, GLX_SGI_make_current_read,
GLX_SGI_swap_control, GLX_SGI_video_sync
GLX version: 1.4
GLX extensions:
GLX_ARB_create_context, GLX_ARB_create_context_profile,
GLX_ARB_get_proc_address, GLX_ARB_multisample,
Re: [e-users] How to make enlightenment more battery friendly?
On Wed, Jul 30, 2014 at 8:35 AM, P Purkayastha ppu...@gmail.com wrote:
On Tue, Jul 29, 2014 at 6:54 PM, Cedric BAIL cedric.b...@free.fr wrote:
On Tue, Jul 29, 2014 at 12:15 PM, P Purkayastha ppu...@gmail.com wrote:
Haven't tried valgrind or other things. My system packages are all
stable Gentoo packages (gcc-4.7.3, kernel-3.12.21, X-1.15.0, etc).
Only things that are not stable are a few end user packages like
firefox, chrome, libreoffice, e19, etc.
What is the exact version of your CPU/GPU and memory information to ?
What is your glxinfo information ?
cpu: i7-3537U (/proc/cpuinfo attached)
gpu: intel HD 4000 (see below)
ram: 8G
glxinfo: attached.
Some other info:
00:02.0 VGA compatible controller: Intel Corporation 3rd Gen Core
processor Graphics Controller (rev 09) (prog-if 00 [VG
A controller])
Subsystem: Dell Device 058b
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-
ParErr- Stepping- SERR- FastB2B- DisINTx+
Status: Cap+ 66MHz- UDF- FastB2B+ ParErr- DEVSEL=fast TAbort-
TAbort- MAbort- SERR- PERR- INTx-
Latency: 0
Interrupt: pin A routed to IRQ 40
Region 0: Memory at d000 (64-bit, non-prefetchable) [size=4M]
Region 2: Memory at c000 (64-bit, prefetchable) [size=256M]
Region 4: I/O ports at 2000 [size=64]
Expansion ROM at unassigned [disabled]
Capabilities: [90] MSI: Enable+ Count=1/1 Maskable- 64bit-
Address: fee0f00c Data: 4171
Capabilities: [d0] Power Management version 2
Flags: PMEClk- DSI+ D1- D2- AuxCurrent=0mA
PME(D0-,D1-,D2-,D3hot-,D3cold-)
Status: D0 NoSoftRst- PME-Enable- DSel=0 DScale=0 PME-
Capabilities: [a4] PCI Advanced Features
AFCap: TP+ FLR+
AFCtrl: FLR-
AFStatus: TP-
Kernel driver in use: i915
/tmp» free -m
total used free sharedbuffers cached
Mem: 7835 7206629878 84 5027
-/+ buffers/cache: 2094 5741
Swap: 1999 18 1981
/tmp» uname -r
3.12.21-gentoo-r1
/tmp» eix -e mesa-progs
[I] x11-apps/mesa-progs
Available versions: 8.0.1 8.1.0 {egl gles1 gles2}
Installed versions: 8.1.0(01:54:12 PM 07/20/2014)(-egl -gles1 -gles2)
Homepage:http://mesa3d.sourceforge.net/
Description: Mesa's OpenGL utility and demo programs
(glxgears and glxinfo)
/tmp» eix -I intel
[I] x11-drivers/xf86-video-intel
Available versions: ~*2.9.1 2.19.0 2.20.13 2.21.15 ~2.99.903
~2.99.905-r1 ~2.99.906 ~2.99.907-r1 ~2.99.909 ~2.99.910 ~2.99.911-r1
~2.99.912 ~2.99.912-r1 ~2.99.914 {debug dri glamor (+)sna +udev uxa
xvmc}
Installed versions: 2.21.15(06:55:14 PM 05/28/2014)(dri sna udev
uxa xvmc -glamor)
Homepage:http://xorg.freedesktop.org/
Description: X.Org driver for Intel cards
[I] x11-libs/libva-intel-driver
Available versions: 1.0.17 ~1.0.19 ~1.0.20 ~1.0.20-r1 ~1.2.0
~1.2.1 ~1.2.2 ~1.2.2-r1 1.3.0 ~1.3.2 ** {X +drm wayland
ABI_MIPS=n32 n64 o32 ABI_PPC=32 64 ABI_S390=32 64 ABI_X86=32 64
x32}
Installed versions: 1.3.0(09:04:16 PM 07/19/2014)(X drm -wayland
ABI_MIPS=-n32 -n64 -o32 ABI_PPC=-32 -64 ABI_S390=-32 -64
ABI_X86=64 -32 -x32)
Homepage:http://www.freedesktop.org/wiki/Software/vaapi
Description: HW video decode support for Intel integrated
graphics
Sorry, mesa version installed is this:
[I] media-libs/mesa
Available versions: [M]7.10.3 [M]7.11.2 [M]8.0.4-r1 [M]~9.0.3
9.1.6 ~9.2.5-r1 10.0.4 ~10.1.0 ~10.1.1 ~10.1.3 ~10.1.4 ~10.1.6 ~10.2.1
~10.2.2 ~10.2.4 {bindist +classic debug +dri3 +egl g3dvl +gallium gbm
gles gles1 gles2 hardened (+)llvm (+)llvm-shared-libs motif +nptl
opencl openmax openvg osmesa pax_kernel pic r600-llvm-compiler selinux
shared-dricore +shared-glapi vdpau wayland xa xorg xvmc ABI_MIPS=n32
n64 o32 ABI_PPC=32 64 ABI_S390=32 64 ABI_X86=32 64 x32
KERNEL=FreeBSD PYTHON_SINGLE_TARGET=python2_7
PYTHON_TARGETS=python2_7 VIDEO_CARDS=freedreno i915 i965 ilo intel
mach64 mga nouveau r100 r128 r200 r300 r600 radeon radeonsi savage sis
tdfx via vmware}
Installed versions: 10.0.4(12:16:49 AM 04/11/2014)(classic egl
gallium nptl xvmc -bindist -debug -gbm -gles1 -gles2 -llvm
-llvm-shared-libs -opencl -openvg -osmesa -pax_kernel -pic
-r600-llvm-compiler -selinux -vdpau -wayland -xa ABI_MIPS=-n32 -n64
-o32 ABI_X86=64 -32 -x32 KERNEL=-FreeBSD VIDEO_CARDS=intel
-freedreno -i915 -i965 -ilo -nouveau -r100 -r200 -r300 -r600 -radeon
-radeonsi -vmware)
Homepage:http://mesa3d.sourceforge.net/
Description: OpenGL-like graphic library for Linux
--
Infragistics Professional
Build stunning WinForms apps today!
Reboot your WinForms applications with our WinForms controls.
Re: [e-users] How to make enlightenment more battery friendly?
On Tue, Jul 29, 2014 at 04:43:50PM +0900, Carsten Haitzler wrote: If I'm not getting it, and just wishful thinking, tell me and I'll go away :) there is a framerate slider - u can slide that down. advanced - performance Sorry, which menu is that in? I went to Composite settings, and can't find it there. However I found that composite/rendering was set to Software somehow, so I just switched it to Opengl. On Tue, Jul 29, 2014 at 11:40:40AM +0200, Cedric BAIL wrote: I have been thinking for some time that we should have a flat no effect theme, that we could use for this kind of use case. If we had a new kind of action, we could trigger theme change on a few scenario (battery unplug/plug or battery low/charging) and achieve what I think you are looking for. That would indeed be a great option for low end hardware, and battery debugging. Did you try reducing the frame rate already ? Also there is noway to delay when select wake up without having some kind of more active loop. Where is that? Also on my older i7, switching to CPU rendering was giving me a 10% battery increase. Did you try that ? Actually I was on software and just switched to hardware. Good news is that I'm now showing much better numbers: 700 mW 12.4 ms/s 61.1Process/usr/bin/enlightenment 1W and 61 wakeups is definitely better. On Tue, Jul 29, 2014 at 12:54:37PM +0200, Cedric BAIL wrote: What is the exact version of your CPU/GPU and memory information to ? What is your glxinfo information ? model name : Intel(R) Core(TM) i7-4800MQ CPU @ 2.70GHz name of display: :0.0 display: :0 screen: 0 direct rendering: Yes server glx vendor string: SGI server glx version string: 1.4 server glx extensions: GLX_ARB_create_context, GLX_ARB_create_context_profile, GLX_ARB_multisample, GLX_EXT_create_context_es2_profile, GLX_EXT_import_context, GLX_EXT_texture_from_pixmap, GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_INTEL_swap_event, GLX_MESA_copy_sub_buffer, GLX_OML_swap_method, GLX_SGIS_multisample, GLX_SGIX_fbconfig, GLX_SGIX_pbuffer, GLX_SGIX_visual_select_group, GLX_SGI_swap_control client glx vendor string: Mesa Project and SGI client glx version string: 1.4 client glx extensions: GLX_ARB_create_context, GLX_ARB_create_context_profile, GLX_ARB_create_context_robustness, GLX_ARB_fbconfig_float, GLX_ARB_framebuffer_sRGB, GLX_ARB_get_proc_address, GLX_ARB_multisample, GLX_EXT_create_context_es2_profile, GLX_EXT_fbconfig_packed_float, GLX_EXT_framebuffer_sRGB, GLX_EXT_import_context, GLX_EXT_texture_from_pixmap, GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_INTEL_swap_event, GLX_MESA_copy_sub_buffer, GLX_MESA_multithread_makecurrent, GLX_MESA_swap_control, GLX_OML_swap_method, GLX_OML_sync_control, GLX_SGIS_multisample, GLX_SGIX_fbconfig, GLX_SGIX_pbuffer, GLX_SGIX_visual_select_group, GLX_SGI_make_current_read, GLX_SGI_swap_control, GLX_SGI_video_sync GLX version: 1.4 GLX extensions: GLX_ARB_create_context, GLX_ARB_create_context_profile, GLX_ARB_get_proc_address, GLX_ARB_multisample, GLX_EXT_create_context_es2_profile, GLX_EXT_import_context, GLX_EXT_texture_from_pixmap, GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_INTEL_swap_event, GLX_MESA_copy_sub_buffer, GLX_MESA_multithread_makecurrent, GLX_MESA_swap_control, GLX_OML_swap_method, GLX_OML_sync_control, GLX_SGIS_multisample, GLX_SGIX_fbconfig, GLX_SGIX_pbuffer, GLX_SGIX_visual_select_group, GLX_SGI_make_current_read, GLX_SGI_swap_control, GLX_SGI_video_sync OpenGL vendor string: Intel Open Source Technology Center OpenGL renderer string: Mesa DRI Intel(R) Haswell Mobile OpenGL core profile version string: 3.1 (Core Profile) Mesa 9.2.2 OpenGL core profile shading language version string: 1.40 OpenGL core profile context flags: (none) OpenGL core profile extensions: GL_3DFX_texture_compression_FXT1, GL_AMD_draw_buffers_blend, GL_ANGLE_texture_compression_dxt3, GL_ANGLE_texture_compression_dxt5, GL_APPLE_object_purgeable, GL_ARB_ES2_compatibility, GL_ARB_ES3_compatibility, GL_ARB_base_instance, GL_ARB_blend_func_extended, GL_ARB_copy_buffer, GL_ARB_debug_output, GL_ARB_depth_buffer_float, GL_ARB_depth_clamp, GL_ARB_draw_buffers, GL_ARB_draw_buffers_blend, GL_ARB_draw_elements_base_vertex, GL_ARB_draw_instanced, GL_ARB_explicit_attrib_location, GL_ARB_fragment_coord_conventions, GL_ARB_fragment_shader, GL_ARB_framebuffer_object, GL_ARB_framebuffer_sRGB, GL_ARB_get_program_binary, GL_ARB_half_float_pixel, GL_ARB_half_float_vertex, GL_ARB_instanced_arrays, GL_ARB_internalformat_query, GL_ARB_invalidate_subdata, GL_ARB_map_buffer_range, GL_ARB_occlusion_query2, GL_ARB_pixel_buffer_object, GL_ARB_point_sprite, GL_ARB_provoking_vertex, GL_ARB_robustness, GL_ARB_sampler_objects, GL_ARB_seamless_cube_map, GL_ARB_shader_bit_encoding,
Re: [e-users] How to make enlightenment more battery friendly?
On Wednesday 30 Jul 2014 01:37:13 P Purkayastha wrote: Installed versions: 10.0.4(12:16:49 AM 04/11/2014)(classic egl gallium nptl xvmc -bindist -debug -gbm -gles1 -gles2 -llvm -llvm-shared-libs -opencl -openvg -osmesa -pax_kernel -pic -r600-llvm-compiler -selinux -vdpau -wayland -xa ABI_MIPS=-n32 -n64 -o32 ABI_X86=64 -32 -x32 KERNEL=-FreeBSD VIDEO_CARDS=intel -freedreno -i915 -i965 -ilo -nouveau -r100 -r200 -r300 -r600 -radeon -radeonsi -vmware) Shouldn't you have i965 also set in VIDEO_CARDS? -- Regards, Mick -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Tue, 29 Jul 2014 21:11:48 -0700 Marc MERLIN marc_...@merlins.org said: On Tue, Jul 29, 2014 at 04:43:50PM +0900, Carsten Haitzler wrote: If I'm not getting it, and just wishful thinking, tell me and I'll go away :) there is a framerate slider - u can slide that down. advanced - performance Sorry, which menu is that in? I went to Composite settings, and can't find it there. advanced - performance. not in comp settings. it's just global fps rate for the ecore animator - it covers everything that animates in e. note - i fixed our vsync support so as of efl 1.11 setting this fps rate will have no effect - it'll go off your screen refresh directly - whatever it happens to be. However I found that composite/rendering was set to Software somehow, so I just switched it to Opengl. well well.. THAT would explain your cpu usage! On Tue, Jul 29, 2014 at 11:40:40AM +0200, Cedric BAIL wrote: I have been thinking for some time that we should have a flat no effect theme, that we could use for this kind of use case. If we had a new kind of action, we could trigger theme change on a few scenario (battery unplug/plug or battery low/charging) and achieve what I think you are looking for. That would indeed be a great option for low end hardware, and battery debugging. it actually would make fairly little difference. trust me on this. it'd mostly be a stylistic thing rather than actual real difference in performance. we spend the largest chunk of our time walking objects in evas_render for the cpu used on evas's side, and this is part of what i was saying - needs a rewrite after 10 years of growing. we can't help the cpu used by the driver or driver perf. it affects software rendering more - but given how the theme is done, it'd be very little. with gl it's basically moot. Did you try reducing the frame rate already ? Also there is noway to delay when select wake up without having some kind of more active loop. Where is that? as above. Also on my older i7, switching to CPU rendering was giving me a 10% battery increase. Did you try that ? Actually I was on software and just switched to hardware. Good news is that I'm now showing much better numbers: 700 mW 12.4 ms/s 61.1Process/usr/bin/enlightenment 1W and 61 wakeups is definitely better. wakeups are really a result of framerate and input mostly. the sw rendering in evas uses a thread, and this causes the child thread to rendering, when done wake up the main thread to finalize (push to screen), then sleep again. this appears as more wakeups, but we are just shuffling from one thread to another so in this case it's not a helpful number. On Tue, Jul 29, 2014 at 12:54:37PM +0200, Cedric BAIL wrote: What is the exact version of your CPU/GPU and memory information to ? What is your glxinfo information ? model name: Intel(R) Core(TM) i7-4800MQ CPU @ 2.70GHz name of display: :0.0 display: :0 screen: 0 direct rendering: Yes server glx vendor string: SGI server glx version string: 1.4 server glx extensions: GLX_ARB_create_context, GLX_ARB_create_context_profile, GLX_ARB_multisample, GLX_EXT_create_context_es2_profile, GLX_EXT_import_context, GLX_EXT_texture_from_pixmap, GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_INTEL_swap_event, GLX_MESA_copy_sub_buffer, GLX_OML_swap_method, GLX_SGIS_multisample, GLX_SGIX_fbconfig, GLX_SGIX_pbuffer, GLX_SGIX_visual_select_group, GLX_SGI_swap_control client glx vendor string: Mesa Project and SGI client glx version string: 1.4 client glx extensions: GLX_ARB_create_context, GLX_ARB_create_context_profile, GLX_ARB_create_context_robustness, GLX_ARB_fbconfig_float, GLX_ARB_framebuffer_sRGB, GLX_ARB_get_proc_address, GLX_ARB_multisample, GLX_EXT_create_context_es2_profile, GLX_EXT_fbconfig_packed_float, GLX_EXT_framebuffer_sRGB, GLX_EXT_import_context, GLX_EXT_texture_from_pixmap, GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_INTEL_swap_event, GLX_MESA_copy_sub_buffer, GLX_MESA_multithread_makecurrent, GLX_MESA_swap_control, GLX_OML_swap_method, GLX_OML_sync_control, GLX_SGIS_multisample, GLX_SGIX_fbconfig, GLX_SGIX_pbuffer, GLX_SGIX_visual_select_group, GLX_SGI_make_current_read, GLX_SGI_swap_control, GLX_SGI_video_sync GLX version: 1.4 GLX extensions: GLX_ARB_create_context, GLX_ARB_create_context_profile, GLX_ARB_get_proc_address, GLX_ARB_multisample, GLX_EXT_create_context_es2_profile, GLX_EXT_import_context, GLX_EXT_texture_from_pixmap, GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_INTEL_swap_event, GLX_MESA_copy_sub_buffer, GLX_MESA_multithread_makecurrent, GLX_MESA_swap_control, GLX_OML_swap_method, GLX_OML_sync_control, GLX_SGIS_multisample, GLX_SGIX_fbconfig, GLX_SGIX_pbuffer, GLX_SGIX_visual_select_group, GLX_SGI_make_current_read, GLX_SGI_swap_control,
Re: [e-users] How to make enlightenment more battery friendly?
On Sat, Jul 26, 2014 at 11:55:27PM +0900, Carsten Haitzler wrote: e17 just thrw a few dumb large textures (big triangle pairs to make rects). e18 + throws much more geometry at gl. all the text in e (titlebars, menus, shelf) is renderd as text. it uses less memory and is actually far more flexible and powerful, and it works fine if you aren't re-rendering the bits you don't need to. i have an intel laptop, intel desktop, nvidia desktop, only accel-less laptop, another intel laptop, another nvidia laptop (nouveau)... i know of other drivers. i have far more than just one machine. if you don't have extensions to partial-render, then it full renders. full rendering is far more in e19 vs e17. that's not going to change any time soon. more complex gadgets will cost more. text with soft drop shadows or glows will cost a lot more. they don't get held in memory as a texture/pixmap - they get generated on the fly. I understand that, but would it make sense to have an e config option for less bling, more battery that one could switch to by sending a command from e from ACPI? Basically I'd want e18/e19 to become as basic CPU-wise as twm if I'm on batteries and not moving windows around. Would that be possible? :) Thanks, Marc -- A mouse is a device used to point at the xterm you want to type in - A.S.R. Microsoft is to operating systems what McDonalds is to gourmet cooking Home page: http://marc.merlins.org/ | PGP 1024R/763BE901 -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Mon, 28 Jul 2014 15:59:11 -0700 Marc MERLIN marc_...@merlins.org said:
On Sat, Jul 26, 2014 at 11:55:27PM +0900, Carsten Haitzler wrote:
e17 just thrw a few dumb large textures (big triangle pairs to make rects).
e18
+ throws much more geometry at gl. all the text in e (titlebars, menus,
shelf) is renderd as text. it uses less memory and is actually far more
flexible and powerful, and it works fine if you aren't re-rendering the
bits you don't need to.
i have an intel laptop, intel desktop, nvidia desktop, only accel-less
laptop, another intel laptop, another nvidia laptop (nouveau)... i know of
other drivers. i have far more than just one machine.
if you don't have extensions to partial-render, then it full renders. full
rendering is far more in e19 vs e17. that's not going to change any time
soon. more complex gadgets will cost more. text with soft drop shadows or
glows will cost a lot more. they don't get held in memory as a
texture/pixmap - they get generated on the fly.
I understand that, but would it make sense to have an e config option
for less bling, more battery that one could switch to by sending a
command from e from ACPI?
i'm sorry - but no. it's like asking bmw could i have a small button in my car
that converts my x6 suv into an i3 electric car? it'd be so much better to roll
around town!. it's not so simple.
Basically I'd want e18/e19 to become as basic CPU-wise as twm if I'm on
batteries and not moving windows around.
e has no choice in this matter. it's a deep complex bi-product of many many
many factors - partly canvas itself, its content, design choices and much more.
a lot of it is a direct result of input from apps (more text in titlebars ==
more triangles to render (1 per char), and apps decide the text that will be
there - they also decide how many windows you have and min/max size etc.),
input from a user (choosing where their windows are, how to lay them out, what
gadgets and icons they want in their shelf etc. etc.), and some styling... and
much more. in addition drivers create a big part of the problem. some (all the
egl/gles2 drivers in tizen support this extension and even more - also support
partial swap which affects gl clients). i don't know if the intel egl/gles2
drivers support buffer_age in x11 - they may so try that. you can try force
triple buffer mode in in advanced settings in compositor settings but beware
this may screw up if your driver decided to quad buffer all of a sudden, or
stop working in a sequence which seems common enough on. nvidia support
buffer age which allows redraw minimization in glx. this affects things a lot.
also there simply also is internal driver cost - how much work does your driver
do in terms of copying, spinning to sleep etc.. all of this is unable to be
controlled. some drivers work great, some so-so, some badly.
why don't you poke the intel driver devs to support buffer-age in glx/ that'd
get you most of your mileage and they (intel) were the ones who initiated the
buffer_age extension in khronos.
https://www.opengl.org/registry/specs/EXT/glx_buffer_age.txt
and yet (first 2 guys are definitely intel - or last i knew) THEY don't
support it, but nvidia do?
Would that be possible? :)
no. as above. it's not so simple. and it certainly is not something to put into
e.
if anything you are asking for evas to do things like cache rendering of
objects (individual ones and entire object trees) into textures to avoid more
geometry. this is not simple - it is in fact a pain in the butt given current
evas internals. you can do this EXPLICITLY via 2 mechanisms in evas - map and
proxy. these both have bugs in dark corners that do things like affect the
connman. epluse and pkgkit popups where the clipping of content goes wrong as a
result. this is basically impossible to fix because fixing this will almost
100% certainly just create a new bug to hunt. the solution is a rewrite of the
render core to simplify it and put these concepts into there in stone. but i
never have the time to do this, partly due to emails like this, patch reviews,
other bug reports and a continual stream of other mentoring, and even more i
can't mention. so until then - you have what you have
Thanks,
Marc
--
A mouse is a device used to point at the xterm you want to type in - A.S.R.
Microsoft is to operating systems
what McDonalds is to gourmet
cooking Home page: http://marc.merlins.org/ | PGP
1024R/763BE901
--
- Codito, ergo sum - I code, therefore I am --
The Rasterman (Carsten Haitzler)ras...@rasterman.com
--
Infragistics Professional
Build stunning WinForms apps today!
Reboot your WinForms applications with our WinForms controls.
Build a bridge from your legacy apps to the future.
Re: [e-users] How to make enlightenment more battery friendly?
On Tue, Jul 29, 2014 at 08:53:05AM +0900, Carsten Haitzler wrote: I understand that, but would it make sense to have an e config option for less bling, more battery that one could switch to by sending a command from e from ACPI? i'm sorry - but no. it's like asking bmw could i have a small button in my car that converts my x6 suv into an i3 electric car? it'd be so much better to roll around town!. it's not so simple. Ah now, if you're putting facts in front of my dreams, we're not going to get anywhere :) e has no choice in this matter. it's a deep complex bi-product of many many many factors - partly canvas itself, its content, design choices and much more. a lot of it is a direct result of input from apps (more text in titlebars == more triangles to render (1 per char), and apps decide the text that will be there - they also decide how many windows you have and min/max size etc.), input from a user (choosing where their windows are, how to lay them out, what So I was thinking about a minimal eye candy mode, like I believe e17 had early on when you set it up. Isn't there a way to disable at least some of the work it does to make things look good? For instance, is it possible to run the main event loop less often? Disable all things that move if not needed like the 'window is not responding spinning stuff' which did improve my battery after I turned it off? I know you have no control over what the 3D drivers do, but you have control over what you feed them :) as well as how often e wakes up. if anything you are asking for evas to do things like cache rendering of objects (individual ones and entire object trees) into textures to avoid more geometry. this is not simple - it is in fact a pain in the butt given current evas internals. you can do this EXPLICITLY via 2 mechanisms in evas - map and proxy. these both have bugs in dark corners that do things like affect the connman. epluse and pkgkit popups where the clipping of content goes wrong as a result. this is basically impossible to fix because fixing this will almost 100% certainly just create a new bug to hunt. the solution is a rewrite of the render core to simplify it and put these concepts into there in stone. but i never have the time to do this, partly due to emails like this, patch reviews, other bug reports and a continual stream of other mentoring, and even more i can't mention. so until then - you have what you have That's fair obviously :) and you also know what's possible and what's not as well as how much some of those are, but somehow I was hoping that there would be low hanging fruits which yield degraded performance but save battery. 1) Aren't there any other such items like the one you just gave me (turn off the spinning circle?) 2) Can't e18 be made to sleep longer between event runs? If I'm not getting it, and just wishful thinking, tell me and I'll go away :) Marc -- A mouse is a device used to point at the xterm you want to type in - A.S.R. Microsoft is to operating systems what McDonalds is to gourmet cooking Home page: http://marc.merlins.org/ | PGP 1024R/763BE901 -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Tue, Jul 29, 2014 at 12:52 PM, Marc MERLIN marc_...@merlins.org wrote: So I was thinking about a minimal eye candy mode, like I believe e17 had early on when you set it up. Isn't there a way to disable at least some of the work it does to make things look good? For instance, is it possible to run the main event loop less often? Disable all things that move if not needed like the 'window is not responding spinning stuff' which did improve my battery after I turned it off? You can reduce/disable some of the composite effects. That could help reduce some of the cpu/battery usage. In the Composite settings, just check either or both Enable fast composite effects Disable composite effects restart e, and see if it helps. It disables some animations, but it won't disable the full composited desktop. It probably won't help much in the case of the spinning wheel. What I have definitely noticed is that it is a ***spinning wheel*** that takes up a lot of cpu. Other potential events that should cause frequent refreshes of the screen, such as a movie playing or a progress bar does not take so much cpu. The desktop transitions, window animations, etc all remain smooth only in the case it is not a spinning wheel. -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Tue, 29 Jul 2014 13:24:35 +0800 P Purkayastha ppu...@gmail.com wrote: On Tue, Jul 29, 2014 at 12:52 PM, Marc MERLIN marc_...@merlins.org wrote: So I was thinking about a minimal eye candy mode, like I believe e17 had early on when you set it up. Isn't there a way to disable at least some of the work it does to make things look good? For instance, is it possible to run the main event loop less often? Disable all things that move if not needed like the 'window is not responding spinning stuff' which did improve my battery after I turned it off? You can reduce/disable some of the composite effects. That could help reduce some of the cpu/battery usage. In the Composite settings, just check either or both Enable fast composite effects Disable composite effects restart e, and see if it helps. It disables some animations, but it won't disable the full composited desktop. It probably won't help much in the case of the spinning wheel. What I have definitely noticed is that it is a ***spinning wheel*** that takes up a lot of cpu. Other potential events that should cause frequent refreshes of the screen, such as a movie playing or a progress bar does not take so much cpu. The desktop transitions, window animations, etc all remain smooth only in the case it is not a spinning wheel. Depending on how things are setup, and the capabilities of your hardware, playing videos might be hardware accelerated. So likely those wont take much CPU. That wont apply to progress bars and the other stuff though. -- A big old stinking pile of genius that no one wants coz there are too many silver coated monkeys in the world. -- Infragistics Professional Build stunning WinForms apps today! Reboot your WinForms applications with our WinForms controls. Build a bridge from your legacy apps to the future. http://pubads.g.doubleclick.net/gampad/clk?id=153845071iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Sun, Jul 20, 2014 at 10:53 PM, Carsten Haitzler ras...@rasterman.com wrote: On Sun, 20 Jul 2014 14:02:52 +0800 P Purkayastha ppu...@gmail.com said: I have seen e get to high cpu when the spinning wheel starts, especially affects e19 even if the wheel is spinning in a different desktop than the currently visible one. even if not visible - it's waking up anbd changing object state - likely figuring out it's a rendering NOOP and then doing nothing after a canvas object pass. I think you should try out e19 on drivers that do not support the buffer egl extension. It not only takes more CPU when there is a single tiny rotating wheel on the desktop, the whole desktop stutters when changing desks. *Everything* is slower and choppy - desktop transitions, window movement, opening new windows, etc. My laptop is just over a year old (dell xps 13) that has the intel graphics and core i7 cpu just from the generation just preceding haswell. It is not as powerful as nvidia, but it should have plenty of processing power to run e19 without stuttering. The screen on the laptop is of 1920x1080 resolution (if it matters to e19). I have never seen e17 produce this kind of choppy desktop. -- Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with a free copy of Black Duck Code Sight - the same software that powers the world's largest code search on Ohloh, the Black Duck Open Hub! Try it now. http://p.sf.net/sfu/bds ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Sat, 26 Jul 2014 21:49:37 +0800 P Purkayastha ppu...@gmail.com said: On Sun, Jul 20, 2014 at 10:53 PM, Carsten Haitzler ras...@rasterman.com wrote: On Sun, 20 Jul 2014 14:02:52 +0800 P Purkayastha ppu...@gmail.com said: I have seen e get to high cpu when the spinning wheel starts, especially affects e19 even if the wheel is spinning in a different desktop than the currently visible one. even if not visible - it's waking up anbd changing object state - likely figuring out it's a rendering NOOP and then doing nothing after a canvas object pass. I think you should try out e19 on drivers that do not support the buffer egl extension. It not only takes more CPU when there is a single tiny rotating wheel on the desktop, the whole desktop stutters when changing desks. *Everything* is slower and choppy - desktop transitions, window movement, opening new windows, etc. My laptop is just over a year old (dell xps 13) that has the intel graphics and core i7 cpu just from the generation just preceding haswell. It is not as powerful as nvidia, but it should have plenty of processing power to run e19 without stuttering. The screen on the laptop is of 1920x1080 resolution (if it matters to e19). I have never seen e17 produce this kind of choppy desktop. e17 just thrw a few dumb large textures (big triangle pairs to make rects). e18 + throws much more geometry at gl. all the text in e (titlebars, menus, shelf) is renderd as text. it uses less memory and is actually far more flexible and powerful, and it works fine if you aren't re-rendering the bits you don't need to. i have an intel laptop, intel desktop, nvidia desktop, only accel-less laptop, another intel laptop, another nvidia laptop (nouveau)... i know of other drivers. i have far more than just one machine. if you don't have extensions to partial-render, then it full renders. full rendering is far more in e19 vs e17. that's not going to change any time soon. more complex gadgets will cost more. text with soft drop shadows or glows will cost a lot more. they don't get held in memory as a texture/pixmap - they get generated on the fly. -- - Codito, ergo sum - I code, therefore I am -- The Rasterman (Carsten Haitzler)ras...@rasterman.com -- Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with a free copy of Black Duck Code Sight - the same software that powers the world's largest code search on Ohloh, the Black Duck Open Hub! Try it now. http://p.sf.net/sfu/bds ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
this days internet is full with chrome battery killer, no idea why
from one discussion i saw:
https://code.google.com/p/chromium/issues/detail?id=153139
On Mon, Jul 21, 2014 at 4:43 AM, Marc MERLIN marc_...@merlins.org wrote:
On Sun, Jul 20, 2014 at 11:58:00PM +0900, Carsten Haitzler wrote:
I have built in haswell intel drivers, I expect they're basic.
If I were running nvidia, then my battery life would be much worse.
fyi - nvidia support buffer_age in GLX, last i looked intel don't, so
nvidia
are actually far nicer here at trynig to cut down power usage and
drawing than
intel. :)
Sorry, but I'll call bull on this. Intel video drivers built in my CPU
with full power management, even if the drivers are a bit dumb, uses
less power than an additional nvidia chip that has to be powered
separately and talked to remotely over a bus, no matter how smart its
drivers are.
change theme. :) or simple check power when anim is not on - it takes a
while
to kick in. or go to windows - window process management - and disable
ping
clients.
So I disabled that, the spinning circle is gone, but enlightenment is
definitely doing better.
2.08 W 17.2 ms/s 88.9Process
/usr/bin/enlightenment
But it still spikes randomly to
3.30 W 22.7 ms/s 141.1Process
/usr/bin/enlightenment
3.64 W 19.7 ms/s 161.2Process
/usr/bin/enlightenment
4.30 W 44.7 ms/s 203.8Process
/usr/bin/enlightenment
Is that acceptable/expected for e17?
Thanks,
Marc
--
A mouse is a device used to point at the xterm you want to type in -
A.S.R.
Microsoft is to operating systems
what McDonalds is to gourmet
cooking
Home page: http://marc.merlins.org/ | PGP
1024R/763BE901
--
Want fast and easy access to all the code in your enterprise? Index and
search up to 200,000 lines of code with a free copy of Black Duck
Code Sight - the same software that powers the world's largest code
search on Ohloh, the Black Duck Open Hub! Try it now.
http://p.sf.net/sfu/bds
___
enlightenment-users mailing list
enlightenment-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/enlightenment-users
--
Want fast and easy access to all the code in your enterprise? Index and
search up to 200,000 lines of code with a free copy of Black Duck
Code Sight - the same software that powers the world's largest code
search on Ohloh, the Black Duck Open Hub! Try it now.
http://p.sf.net/sfu/bds
___
enlightenment-users mailing list
enlightenment-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Sun, 20 Jul 2014 18:43:21 -0700 Marc MERLIN marc_...@merlins.org said:
On Sun, Jul 20, 2014 at 11:58:00PM +0900, Carsten Haitzler wrote:
I have built in haswell intel drivers, I expect they're basic.
If I were running nvidia, then my battery life would be much worse.
fyi - nvidia support buffer_age in GLX, last i looked intel don't, so nvidia
are actually far nicer here at trynig to cut down power usage and drawing
than intel. :)
Sorry, but I'll call bull on this. Intel video drivers built in my CPU
with full power management, even if the drivers are a bit dumb, uses
less power than an additional nvidia chip that has to be powered
separately and talked to remotely over a bus, no matter how smart its
drivers are.
not bull. talking here. this extension. it can save drawing a 4480*1440
frameubuffer every frame to just drawing maybe 32x32 pixels per frames. that's
over 6 THOUSAND times as many pixels drawn (and likely textels read/accessed
too) than without the extension. that's the example of my home desktop with an
nvidia card in it.
without the extension you draw the whole lot, not just the updates (this is an
opengl requirement due to the defintion of the backbuffer state after a swap).
thats a LOT Of DRAWING ... that you don't need, and if the extn is not
supported, that's going to impact power a lot.
here is the topic of this extension. nvidia have supported this for a long
while. looking here - on very new intel drivers (which given an update about 2
weeks back lost their vsync ability - yay! tearing now), do not support this
extension, so intel is down to drawing the whole buffer, nvidia is down to
drawing just the updates. so a spinning busy icon or a blinking cursor
nvidia supports it, intel currently does not. not on glx.
change theme. :) or simple check power when anim is not on - it takes a
while to kick in. or go to windows - window process management - and
disable ping clients.
So I disabled that, the spinning circle is gone, but enlightenment is
definitely doing better.
2.08 W 17.2 ms/s 88.9Process/usr/bin/enlightenment
But it still spikes randomly to
3.30 W 22.7 ms/s 141.1Process/usr/bin/enlightenment
3.64 W 19.7 ms/s 161.2Process/usr/bin/enlightenment
4.30 W 44.7 ms/s 203.8Process/usr/bin/enlightenment
Is that acceptable/expected for e17?
it's animating, doing something, responding to any number of clients that are
drawing - any number of mouse events, key events, and any number of x client
messages and x events coming in - if you have a lot of clients then e may very
well every 10-20 seconds or so wake up and send a ping to each and every window
you have and get N replies - one for every window. if you have 50 windows that
could mean 50+ wakeups in a row (depending on luck). a cursor in terminology
will be animating almost all the time if blinking (not just on and off - it
glows up and down, fades etc.). if some app does a redraw, e gets damage events
and has to respond. in order to ensure e rendering at a given framerate it may
wake up for events, do something, then go to sleep again, only to wake up when
the right frame tick comes along and begin the render then - i have code in
ecore_evas to do exactly this. it simply is splitting its waking time in 2 and
sleeping in the middle. so it's possible. it depends what is going on.
Thanks,
Marc
--
A mouse is a device used to point at the xterm you want to type in - A.S.R.
Microsoft is to operating systems
what McDonalds is to gourmet
cooking Home page: http://marc.merlins.org/ | PGP
1024R/763BE901
--
- Codito, ergo sum - I code, therefore I am --
The Rasterman (Carsten Haitzler)ras...@rasterman.com
--
Want fast and easy access to all the code in your enterprise? Index and
search up to 200,000 lines of code with a free copy of Black Duck
Code Sight - the same software that powers the world's largest code
search on Ohloh, the Black Duck Open Hub! Try it now.
http://p.sf.net/sfu/bds
___
enlightenment-users mailing list
enlightenment-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
I have seen e get to high cpu when the spinning wheel starts, especially affects e19 even if the wheel is spinning in a different desktop than the currently visible one. Do the Intel drivers have support for this extension? I can't find information online. Only thing I did find was that it was added to Mesa in beginning of 2013 (EGL_BUFFER_EXT). On Jul 20, 2014 11:53 AM, Carsten Haitzler ras...@rasterman.com wrote: On Sat, 19 Jul 2014 09:53:54 -0700 Marc MERLIN marc_...@merlins.org said: I had this problem with e17, and now same thing with e18 Here's powertop when I'm on batteries: 5.66 W 11.1 ms/s 269.5 Process/usr/local/opt/google/chrome-beta/chrome --user-data-dir=/home/me 4.71 W 96.8 ms/s 224.9 Process/usr/local/opt/google/chrome-beta/chrome --type=renderer --lang=e 4.10 W 6.6 ms/s 196.3Interrupt PS/2 Touchpad / Keyboard / Mouse 1.93 W 27.7 ms/s 92.8 Process/usr/bin/enlightenment 1.21 W 1.0 ms/s 58.1 Timer hrtimer_wakeup (ignore the watt values, but look at the wakeups). So this isn't too bad for enlightenemnt, right? But then I killall -STOP chrome because it's pigging my battery and I'm not using it. The moment I do this, enlightenment gets crazy and starts looping and killing my battery. Powertop now says: 8.35 W 91.2 ms/s 383.0Process /usr/bin/enlightenment 4.52 W 5.1 ms/s 206.6Interrupt PS/2 Touchpad / Keyboard / Mouse 903 mW 1.3 ms/s 41.3Interrupt [42] i915@pci::00:02.0 Why does e18 go from 92 wakeups to 383? i believe that is number of wakeups over the 5s poll... do you have a little spinning thing next to the icon on the top-left in the window border of chrome? http://www.enlightenment.org/ss/e-53cb3b206dea92.59532872.png ? becuase that spinner == application hung,. not responding to pings. chrome sa is talks the netwm pin protocol: WM_PROTOCOLS(ATOM): protocols WM_DELETE_WINDOW, _NET_WM_PING this means that if something sends u a ping client message, you will respond. enlightenment sees this and pings any client every 10 sec or so if it says it speaks this. if your app doesn't respond to this within about 10 sec, it's marked as hung and you get the little spinner. 10 sec is a very very very slow and conservative ping interval to save power, but this unfortunately means it takes a while to detect idleness. it's that animation that is causing redraws and thus wakeups. if your gl drivers support buffer age extension, then the redraws can be limited to just the area that changes, otherwise the whole screen is redrawn (if you have 1 screen that means all screens - this is a requirement of opengl unless you have extensions). Because e is started under gdb, or somesuch, I can't strace it to see what it's doing wrong, but clearly it's spinning silly. Can a developer try to run google-chrome, switch to batteries, start powertop, run killall -STOP chrome, and observe? (it will likely be faster than me relaying this via Email anyway) well not going to run google chrome - but chromium will do. :) Thanks, Marc -- A mouse is a device used to point at the xterm you want to type in - A.S.R. Microsoft is to operating systems what McDonalds is to gourmet cooking Home page: http://marc.merlins.org/ | PGP 1024R/763BE901 -- Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with a free copy of Black Duck Code Sight - the same software that powers the world's largest code search on Ohloh, the Black Duck Open Hub! Try it now. http://p.sf.net/sfu/bds ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users -- - Codito, ergo sum - I code, therefore I am -- The Rasterman (Carsten Haitzler)ras...@rasterman.com -- Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with a free copy of Black Duck Code Sight - the same software that powers the world's largest code search on Ohloh, the Black Duck Open Hub! Try it now. http://p.sf.net/sfu/bds ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users -- Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with
Re: [e-users] How to make enlightenment more battery friendly?
On Sun, Jul 20, 2014 at 12:43:05PM +0900, Carsten Haitzler wrote: i believe that is number of wakeups over the 5s poll... do you have a little spinning thing next to the icon on the top-left in the window border of chrome? Yes, I do, and as you explain, this is expected. Is that tiny thing the reason for so much CPU burn and watts lost? it's that animation that is causing redraws and thus wakeups. if your gl drivers support buffer age extension, then the redraws can be limited to just the area that changes, otherwise the whole screen is redrawn (if you have 1 screen that means all screens - this is a requirement of opengl unless you have extensions). I have built in haswell intel drivers, I expect they're basic. If I were running nvidia, then my battery life would be much worse. Can a developer try to run google-chrome, switch to batteries, start powertop, run killall -STOP chrome, and observe? (it will likely be faster than me relaying this via Email anyway) well not going to run google chrome - but chromium will do. :) chromium is fine indeed, run that for this test :) Anyway, can I tell e to stop making that circle animation to see if it's the reason for my battery problem? Thanks, Marc -- A mouse is a device used to point at the xterm you want to type in - A.S.R. Microsoft is to operating systems what McDonalds is to gourmet cooking Home page: http://marc.merlins.org/ | PGP 1024R/763BE901 -- Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with a free copy of Black Duck Code Sight - the same software that powers the world's largest code search on Ohloh, the Black Duck Open Hub! Try it now. http://p.sf.net/sfu/bds ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Sun, 20 Jul 2014 14:02:52 +0800 P Purkayastha ppu...@gmail.com said: I have seen e get to high cpu when the spinning wheel starts, especially affects e19 even if the wheel is spinning in a different desktop than the currently visible one. even if not visible - it's waking up anbd changing object state - likely figuring out it's a rendering NOOP and then doing nothing after a canvas object pass. Do the Intel drivers have support for this extension? I can't find information online. Only thing I did find was that it was added to Mesa in beginning of 2013 (EGL_BUFFER_EXT). for GLX you need GLX_EXT_buffer_age - for EGL you need EGL_EXT_buffer_age in your extension strings. if you run an efl app with the gl engine and do: export EVAS_GL_INFO=1 expedite -e gl | grep _buffer_age you should see (grep should show the line with that in it). fyi nvidia drivers have supported this for quite a while now on GLX. On Jul 20, 2014 11:53 AM, Carsten Haitzler ras...@rasterman.com wrote: On Sat, 19 Jul 2014 09:53:54 -0700 Marc MERLIN marc_...@merlins.org said: I had this problem with e17, and now same thing with e18 Here's powertop when I'm on batteries: 5.66 W 11.1 ms/s 269.5 Process/usr/local/opt/google/chrome-beta/chrome --user-data-dir=/home/me 4.71 W 96.8 ms/s 224.9 Process/usr/local/opt/google/chrome-beta/chrome --type=renderer --lang=e 4.10 W 6.6 ms/s 196.3Interrupt PS/2 Touchpad / Keyboard / Mouse 1.93 W 27.7 ms/s 92.8 Process/usr/bin/enlightenment 1.21 W 1.0 ms/s 58.1 Timer hrtimer_wakeup (ignore the watt values, but look at the wakeups). So this isn't too bad for enlightenemnt, right? But then I killall -STOP chrome because it's pigging my battery and I'm not using it. The moment I do this, enlightenment gets crazy and starts looping and killing my battery. Powertop now says: 8.35 W 91.2 ms/s 383.0Process /usr/bin/enlightenment 4.52 W 5.1 ms/s 206.6Interrupt PS/2 Touchpad / Keyboard / Mouse 903 mW 1.3 ms/s 41.3Interrupt [42] i915@pci::00:02.0 Why does e18 go from 92 wakeups to 383? i believe that is number of wakeups over the 5s poll... do you have a little spinning thing next to the icon on the top-left in the window border of chrome? http://www.enlightenment.org/ss/e-53cb3b206dea92.59532872.png ? becuase that spinner == application hung,. not responding to pings. chrome sa is talks the netwm pin protocol: WM_PROTOCOLS(ATOM): protocols WM_DELETE_WINDOW, _NET_WM_PING this means that if something sends u a ping client message, you will respond. enlightenment sees this and pings any client every 10 sec or so if it says it speaks this. if your app doesn't respond to this within about 10 sec, it's marked as hung and you get the little spinner. 10 sec is a very very very slow and conservative ping interval to save power, but this unfortunately means it takes a while to detect idleness. it's that animation that is causing redraws and thus wakeups. if your gl drivers support buffer age extension, then the redraws can be limited to just the area that changes, otherwise the whole screen is redrawn (if you have 1 screen that means all screens - this is a requirement of opengl unless you have extensions). Because e is started under gdb, or somesuch, I can't strace it to see what it's doing wrong, but clearly it's spinning silly. Can a developer try to run google-chrome, switch to batteries, start powertop, run killall -STOP chrome, and observe? (it will likely be faster than me relaying this via Email anyway) well not going to run google chrome - but chromium will do. :) Thanks, Marc -- A mouse is a device used to point at the xterm you want to type in - A.S.R. Microsoft is to operating systems what McDonalds is to gourmet cooking Home page: http://marc.merlins.org/ | PGP 1024R/763BE901 -- Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with a free copy of Black Duck Code Sight - the same software that powers the world's largest code search on Ohloh, the Black Duck Open Hub! Try it now. http://p.sf.net/sfu/bds ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users -- - Codito, ergo sum - I code, therefore I am -- The Rasterman (Carsten Haitzler)ras...@rasterman.com
Re: [e-users] How to make enlightenment more battery friendly?
On Sat, 19 Jul 2014 23:10:29 -0700 Marc MERLIN marc_...@merlins.org said:
On Sun, Jul 20, 2014 at 12:43:05PM +0900, Carsten Haitzler wrote:
i believe that is number of wakeups over the 5s poll... do you have a little
spinning thing next to the icon on the top-left in the window border of
chrome?
Yes, I do, and as you explain, this is expected.
Is that tiny thing the reason for so much CPU burn and watts lost?
if everything is rendering - then yes, even if not everything, then it still is
ticking over and down draw elimination and as i have found after some perf
work... it spends a lot of time inside stuff inside gl drivers. at least if we
are spinning and drawing little bits i notise intel dri/drm things being some
of the highest consumers of cpu time - try:
sudo perf top
and find out. :)
it's that animation that is causing redraws and thus wakeups. if your gl
drivers support buffer age extension, then the redraws can be limited to
just the area that changes, otherwise the whole screen is redrawn (if you
have 1 screen that means all screens - this is a requirement of opengl
unless you have extensions).
I have built in haswell intel drivers, I expect they're basic.
If I were running nvidia, then my battery life would be much worse.
fyi - nvidia support buffer_age in GLX, last i looked intel don't, so nvidia
are actually far nicer here at trynig to cut down power usage and drawing than
intel. :)
Can a developer try to run google-chrome, switch to batteries, start
powertop, run killall -STOP chrome, and observe?
(it will likely be faster than me relaying this via Email anyway)
well not going to run google chrome - but chromium will do. :)
chromium is fine indeed, run that for this test :)
Anyway, can I tell e to stop making that circle animation to see if it's
the reason for my battery problem?
change theme. :) or simple check power when anim is not on - it takes a while
to kick in. or go to windows - window process management - and disable ping
clients.
Thanks,
Marc
--
A mouse is a device used to point at the xterm you want to type in - A.S.R.
Microsoft is to operating systems
what McDonalds is to gourmet
cooking Home page: http://marc.merlins.org/ | PGP
1024R/763BE901
--
- Codito, ergo sum - I code, therefore I am --
The Rasterman (Carsten Haitzler)ras...@rasterman.com
--
Want fast and easy access to all the code in your enterprise? Index and
search up to 200,000 lines of code with a free copy of Black Duck
Code Sight - the same software that powers the world's largest code
search on Ohloh, the Black Duck Open Hub! Try it now.
http://p.sf.net/sfu/bds
___
enlightenment-users mailing list
enlightenment-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Sun, Jul 20, 2014 at 11:58:00PM +0900, Carsten Haitzler wrote: I have built in haswell intel drivers, I expect they're basic. If I were running nvidia, then my battery life would be much worse. fyi - nvidia support buffer_age in GLX, last i looked intel don't, so nvidia are actually far nicer here at trynig to cut down power usage and drawing than intel. :) Sorry, but I'll call bull on this. Intel video drivers built in my CPU with full power management, even if the drivers are a bit dumb, uses less power than an additional nvidia chip that has to be powered separately and talked to remotely over a bus, no matter how smart its drivers are. change theme. :) or simple check power when anim is not on - it takes a while to kick in. or go to windows - window process management - and disable ping clients. So I disabled that, the spinning circle is gone, but enlightenment is definitely doing better. 2.08 W 17.2 ms/s 88.9Process/usr/bin/enlightenment But it still spikes randomly to 3.30 W 22.7 ms/s 141.1Process/usr/bin/enlightenment 3.64 W 19.7 ms/s 161.2Process/usr/bin/enlightenment 4.30 W 44.7 ms/s 203.8Process/usr/bin/enlightenment Is that acceptable/expected for e17? Thanks, Marc -- A mouse is a device used to point at the xterm you want to type in - A.S.R. Microsoft is to operating systems what McDonalds is to gourmet cooking Home page: http://marc.merlins.org/ | PGP 1024R/763BE901 -- Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with a free copy of Black Duck Code Sight - the same software that powers the world's largest code search on Ohloh, the Black Duck Open Hub! Try it now. http://p.sf.net/sfu/bds ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] How to make enlightenment more battery friendly?
On Sat, 19 Jul 2014 09:53:54 -0700 Marc MERLIN marc_...@merlins.org said:
I had this problem with e17, and now same thing with e18
Here's powertop when I'm on batteries:
5.66 W 11.1 ms/s 269.5
Process/usr/local/opt/google/chrome-beta/chrome
--user-data-dir=/home/me 4.71 W 96.8 ms/s 224.9
Process/usr/local/opt/google/chrome-beta/chrome --type=renderer
--lang=e 4.10 W 6.6 ms/s 196.3Interrupt PS/2 Touchpad /
Keyboard / Mouse 1.93 W 27.7 ms/s 92.8
Process/usr/bin/enlightenment 1.21 W 1.0 ms/s 58.1
Timer hrtimer_wakeup
(ignore the watt values, but look at the wakeups).
So this isn't too bad for enlightenemnt, right?
But then I killall -STOP chrome because it's pigging my battery and I'm
not using it.
The moment I do this, enlightenment gets crazy and starts looping and
killing my battery.
Powertop now says:
8.35 W 91.2 ms/s 383.0Process/usr/bin/enlightenment
4.52 W 5.1 ms/s 206.6Interrupt PS/2 Touchpad /
Keyboard / Mouse 903 mW 1.3 ms/s 41.3Interrupt [42]
i915@pci::00:02.0
Why does e18 go from 92 wakeups to 383?
i believe that is number of wakeups over the 5s poll... do you have a little
spinning thing next to the icon on the top-left in the window border of chrome?
http://www.enlightenment.org/ss/e-53cb3b206dea92.59532872.png
? becuase that spinner == application hung,. not responding to pings. chrome
sa is talks the netwm pin protocol:
WM_PROTOCOLS(ATOM): protocols WM_DELETE_WINDOW, _NET_WM_PING
this means that if something sends u a ping client message, you will respond.
enlightenment sees this and pings any client every 10 sec or so if it says it
speaks this. if your app doesn't respond to this within about 10 sec, it's
marked as hung and you get the little spinner. 10 sec is a very very very
slow and conservative ping interval to save power, but this unfortunately means
it takes a while to detect idleness.
it's that animation that is causing redraws and thus wakeups. if your gl
drivers support buffer age extension, then the redraws can be limited to just
the area that changes, otherwise the whole screen is redrawn (if you have 1
screen that means all screens - this is a requirement of opengl unless you have
extensions).
Because e is started under gdb, or somesuch, I can't strace it to see
what it's doing wrong, but clearly it's spinning silly.
Can a developer try to run google-chrome, switch to batteries, start
powertop, run killall -STOP chrome, and observe?
(it will likely be faster than me relaying this via Email anyway)
well not going to run google chrome - but chromium will do. :)
Thanks,
Marc
--
A mouse is a device used to point at the xterm you want to type in - A.S.R.
Microsoft is to operating systems
what McDonalds is to gourmet
cooking Home page: http://marc.merlins.org/ | PGP
1024R/763BE901
--
Want fast and easy access to all the code in your enterprise? Index and
search up to 200,000 lines of code with a free copy of Black Duck
Code Sight - the same software that powers the world's largest code
search on Ohloh, the Black Duck Open Hub! Try it now.
http://p.sf.net/sfu/bds
___
enlightenment-users mailing list
enlightenment-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/enlightenment-users
--
- Codito, ergo sum - I code, therefore I am --
The Rasterman (Carsten Haitzler)ras...@rasterman.com
--
Want fast and easy access to all the code in your enterprise? Index and
search up to 200,000 lines of code with a free copy of Black Duck
Code Sight - the same software that powers the world's largest code
search on Ohloh, the Black Duck Open Hub! Try it now.
http://p.sf.net/sfu/bds
___
enlightenment-users mailing list
enlightenment-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/enlightenment-users
