Active Topics

 


Reply
Thread Tools
Posts: 254 | Thanked: 509 times | Joined on Nov 2011 @ Canada
#21
Originally Posted by ivgalvez View Post
Or a repository that actually depends on CSS Thumb for applications that are built with newer toolchain. There's no need to set up an autobuilder, that requires a lot of work and maintenance, just a repo to store all those packages that are randomly thrown in the forums.
I think a Thumb repo would be great to have on maemo.org... currently Thumb compiled apps are all over the place. Autobuilder would be ideal of course, so that users have at least some assurance that the binary they are installing came from the source code that was uploaded. Failing that, just a simple repo with ability to upload via maemo.org/garage/package interface would be good.

Maybe a bit harder to get setup, but surely not impossible...

If thumb is the "future" for n900 software (and I think it probably is), then this needs to be done at some point.
 

The Following 4 Users Say Thank You to shawnjefferson For This Useful Post:
Estel's Avatar
Posts: 5,028 | Thanked: 8,613 times | Joined on Mar 2011
#22
Originally Posted by shawnjefferson View Post
If thumb is the "future" for n900 software (and I think it probably is), then this needs to be done at some point.
I agree with everything you've said - with a small addition, thumb applications are not "future" for N900 software, they're absolutely the present. Every self-respecting developer offers thumb version where applicable, and most of even barely knowledgeable devs and users are utilizing thumb on their devices.

Some people took personal effort to make their programs easily available thumb-compiled - for example, gdizzz host personal repo with thumb versions of all programs he is maintaining, so no more download-link hunting... In this case.

Sadly, those thing works only if individuals do it on their own - none of the (supposedly?) governing bodies took steps (actively seeking volunteers, etc) to update repository system. They're busy "elsewhere" <put a link to any random mail from Maemo's mailing list, where council and board is throwing apesh|t at each other> ...

/Estel
__________________
N900's aluminum backcover / body replacement
-
N900's HDMI-Out
-
Camera cover MOD
-
Measure battery's real capacity on-device
-
TrueCrypt 7.1 | ereswap | bnf
-
Hardware's mods research is costly. To support my work, please consider donating. Thank You!
 

The Following User Says Thank You to Estel For This Useful Post:
Copernicus's Avatar
Posts: 1,986 | Thanked: 7,698 times | Joined on Dec 2010 @ Dayton, Ohio
#23
Originally Posted by Estel View Post
Every self-respecting developer offers thumb version where applicable, and most of even barely knowledgeable devs and users are utilizing thumb on their devices.
Er, does that mean that I should be creating thumb versions of my various apps? I had kinda assumed that cpu-bound apps would benefit the most from thumb; but my apps are generally benign. (Even Orecchiette is mostly just a wrapper around calls to GStreamer, so it really depends on that library being recompiled for thumb.)

I guess I should try to put thumb onto one of my N900s too, then. I've never really pushed my N900s to the limit, so I've never really worried about it; I play podcasts, watch videos, and edit files in vim just fine without needing to mess with the OS.
 

The Following User Says Thank You to Copernicus For This Useful Post:
pichlo's Avatar
Posts: 6,445 | Thanked: 20,981 times | Joined on Sep 2012 @ UK
#24
Originally Posted by Copernicus View Post
I had kinda assumed that cpu-bound apps would benefit the most from thumb
If that's the case then I've been misunderstanding Thumb all this time. My understanding was that Thumb instructions are slightly less efficient than ARM. The real benefit of Thumb is that the compiled binaries are smaller and thus take less memory space and require less swapping. In other words, any given binary will not really benefit from being compiled as Thumb, especially if it is a CPU-intensive one. Thumb is more about being nice to others: I sacrifice a percent or two of my efficiency to make you running smoother (less swapping).

Please correct me if I got it wrong.
__________________
Русский военный корабль, иди нахуй!
 

The Following 3 Users Say Thank You to pichlo For This Useful Post:
Copernicus's Avatar
Posts: 1,986 | Thanked: 7,698 times | Joined on Dec 2010 @ Dayton, Ohio
#25
Sorry, I shouldn't post while sleepy, you are absolutely right about thumb. (Once upon a time, I had educated myself pretty thoroughly on this...)

And yeah, now that I've got an Android device, I guess can use it as my phone and set up both my N900s as test devices -- one for cssu/thumb, one for stock.

I guess it's also time to start trudging back through all that Sandbox stuff, and somehow scrounge up a machine (real or virtual) to run Debian or Ubuntu. Bleah. (Any chance I could just integrate the thumb toolchain into the Qt SDK? )
 

The Following User Says Thank You to Copernicus For This Useful Post:
Posts: 1,203 | Thanked: 3,027 times | Joined on Dec 2010
#26
if it weren't for cssu-t to cssu-s testing path I'd suggest merging thumb back into cssu-t as i think most are using it anyway. it is still a possibility, just means eventually thumb packages would land in stable.

if your then looking at "future" maemo, an update to linaro/thumb toolchain would be worth a look. a proper fix to the xlib-xcb issue with hw keyboard from qt5 work so we only need one copy of xlib. trying to port hildon to newer gtk or update qildon, see how much patching of gtk can be avoided along the way because of the tablet style focus of gtk these days.
 
Posts: 3,328 | Thanked: 4,476 times | Joined on May 2011 @ Poland
#27
Originally Posted by Android_808 View Post
if your then looking at "future" maemo, an update to linaro/thumb toolchain would be worth a look. a proper fix to the xlib-xcb issue with hw keyboard from qt5 work so we only need one copy of xlib. trying to port hildon to newer gtk or update qildon, see how much patching of gtk can be avoided along the way because of the tablet style focus of gtk these days.
But gtk3 != gtk2
And gtk3 could be brought for N900 only if some of the basic libraries were upgraded (don't remember which exactly. glib was one of them, and it was a version distributed only as random debs)
And gtk2 update is what would fall into "Maemo Reloaded"
For the N900's future, I'd vote for tracker upgrade (if it's feasible at all) - our old tracker is sloooow.
__________________
If you want to support my work, you can donate by PayPal or Flattr

Projects no longer actively developed: here
 

The Following 2 Users Say Thank You to marmistrz For This Useful Post:
Posts: 254 | Thanked: 509 times | Joined on Nov 2011 @ Canada
#28
Originally Posted by pichlo View Post
If that's the case then I've been misunderstanding Thumb all this time. My understanding was that Thumb instructions are slightly less efficient than ARM. The real benefit of Thumb is that the compiled binaries are smaller and thus take less memory space and require less swapping.
I have read that this is the case, thumb instructions are smaller but maybe less efficient, but in reality, all the thumb applications I've tried including ones I've compiled myself either perform the same, or slighter better AND are smaller.

The better/same performance may be the fact we're using a newer compiler though and not an aspect of thumb itself.

Anyway, the bottom line is, that I can see no reason not to have all applications thumb compiled and I've been using thumb on my device since FMG released it basically.
 

The Following 3 Users Say Thank You to shawnjefferson For This Useful Post:
Estel's Avatar
Posts: 5,028 | Thanked: 8,613 times | Joined on Mar 2011
#29
Originally Posted by Copernicus View Post
Er, does that mean that I should be creating thumb versions of my various apps?
That's why I've written 'where applicable". Don't worry, I see you as very self-respecting (and by others) developer (not that you should care about my humble opinion, but you know what my point is about, I hope).

From my experience, thumb-compiled memory savings is most important for programs that either run all the time, or have potential to run in background (media player), and most of your programs are rather of a kind that get "full user focus" when running.

OTOH...

Originally Posted by shawnjefferson View Post
but in reality, all the thumb applications I've tried including ones I've compiled myself either perform the same, or slighter better AND are smaller.
...is true, as benefits of updated linaro/thumb compiler manifest itself in code running more efficient on CPU in most cases, too. So, summing it up, if you can, thumb versions for your programs wouldn't hurt (at worst case, it wouldn't provide any improvement, apart from slightly lower size/RAM usage).

If, by any chance, you could do like gdizzz and set up mini-repo that would host all your thumb programs, it would be heaven - this way, user can just define this custom repo in application manager and forget about really tiresome thread-download-link hunting, everytime new version of something is released.

Of course I realize that not everyone have resources (or time, which is most likely case, nowadays) for setting up personal repo, though.

/Estel
__________________
N900's aluminum backcover / body replacement
-
N900's HDMI-Out
-
Camera cover MOD
-
Measure battery's real capacity on-device
-
TrueCrypt 7.1 | ereswap | bnf
-
Hardware's mods research is costly. To support my work, please consider donating. Thank You!
 

The Following 4 Users Say Thank You to Estel For This Useful Post:
Posts: 254 | Thanked: 509 times | Joined on Nov 2011 @ Canada
#30
Originally Posted by Estel View Post
If, by any chance, you could do like gdizzz and set up mini-repo that would host all your thumb programs, it would be heaven - this way, user can just define this custom repo in application manager and forget about really tiresome thread-download-link hunting, everytime new version of something is released.
While it's better than having thumb apps in threads on TMO (I myself am guilty of that), we should not be striving for individual repos, but a central thumb repo, IMO.
 

The Following 5 Users Say Thank You to shawnjefferson For This Useful Post:
Reply


 
Forum Jump


All times are GMT. The time now is 09:56.