UPD: ARMEL guys are not counted.
I hope you know about the current state of OpenMW code: it is un-portable to non-x86 and un-playable on non-x86.
From my side, I've already made an issue report here: http://bugs.openmw.org/issues/564
Currently, I'm working on this issue inside porting2 branch https://github.com/dougmencken/openmw/commits/porting2 (code works, but for reading, and in a way nobody me nor Zini can accept
No-boy ca-re due to lack of non-x86 testers and developers
-
- Posts: 4
- Joined: 22 Feb 2013, 20:05
No-boy ca-re due to lack of non-x86 testers and developers
Last edited by dougmencken on 22 Feb 2013, 21:08, edited 1 time in total.
Re: No-boy ca-re due to lack of non-x86 testers and develope
Not wanna start to nitpick, but while the statement in the title of your bug report is correct, the non-x86 stuff is not. The code should work in its current state on little endian CPUs and bi-endian CPUs that are operated in little endian mode. I could be wrong about that, but I think that includes all major ARM platforms. The only (somewhat) current platform I am aware of that is incompatible is power pc.
-
- Posts: 4
- Joined: 22 Feb 2013, 20:05
Re: No-boy ca-re due to lack of non-x86 testers and develope
ARMEL is EL (endian-little), yes. But there're a lot of other architectures which are non little-endian. ARM (which is BE by default). MIPS. SPARC. HPPA. S390X. POWER. http://wiki.debian.org/ArchitectureSpecificsMemo
But if your goal is to support only EL, just mention it somewhere. Then I'll just make true fork after we will solve licensing issues (currently, I can just //copyright dougmenken every source) (adding copyrights is okay, it is not removing them).
UPD: PowerPC can operate in EL-mode too instead of BE. But it requires: PPC-EL linux kernel, PPC-EL std C library (glibc), and the whole dependency userspace built to be PPC-EL. And even after that, we will not have 32-bit buffers in OGRE (they are expected due to x86-only SSE! see also: https://ogre3d.atlassian.net/browse/OGRE-142).
But if your goal is to support only EL, just mention it somewhere. Then I'll just make true fork after we will solve licensing issues (currently, I can just //copyright dougmenken every source) (adding copyrights is okay, it is not removing them).
UPD: PowerPC can operate in EL-mode too instead of BE. But it requires: PPC-EL linux kernel, PPC-EL std C library (glibc), and the whole dependency userspace built to be PPC-EL. And even after that, we will not have 32-bit buffers in OGRE (they are expected due to x86-only SSE! see also: https://ogre3d.atlassian.net/browse/OGRE-142).
Last edited by dougmencken on 22 Feb 2013, 21:57, edited 2 times in total.
Re: No-boy ca-re due to lack of non-x86 testers and develope
No one ever said that.But if your goal is to support only EL, just mention it somewhere.
-
- Posts: 4
- Joined: 22 Feb 2013, 20:05
Re: No-boy ca-re due to lack of non-x86 testers and develope
Then my work has some value for OpenMW community? Good!Zini wrote:No one ever said that.But if your goal is to support only EL, just mention it somewhere.
UPD: Nice disappearing of messages and posts. It is funny.
- lgromanowski
- Site Admin
- Posts: 1193
- Joined: 05 Aug 2011, 22:21
- Location: Wroclaw, Poland
- Contact:
Re: No-boy ca-re due to lack of non-x86 testers and develope
It's not disappearing, it's just moderation - the first few posts from new users are moderated.dougmencken wrote:UPD: Nice disappearing of messages and posts. It is funny.
Re: No-boy ca-re due to lack of non-x86 testers and develope
I think I have seen several topics asking " will this run on Android?". This makes me think that you are right, we need it to be portable.
One question though: Will it run at all?
As far as i understand, OpenMW is very CPU hungry, which limits the possibilities of running it in anything other than powerful x86 desktop monsters and some notebooks.
But we need to answer one question: if the development goes on like it is doing now, will we be able to port OpenMW to different architectures? Or do we need to change something in the development process right now in order to let us port it in the future?
I can`t answer the question since i am only trying to learn C++ (partly because of OpenMW, mind it) and i have no idea which way to take.
P.S. Sorry for a bit of necroposting.
One question though: Will it run at all?
As far as i understand, OpenMW is very CPU hungry, which limits the possibilities of running it in anything other than powerful x86 desktop monsters and some notebooks.
But we need to answer one question: if the development goes on like it is doing now, will we be able to port OpenMW to different architectures? Or do we need to change something in the development process right now in order to let us port it in the future?
I can`t answer the question since i am only trying to learn C++ (partly because of OpenMW, mind it) and i have no idea which way to take.
P.S. Sorry for a bit of necroposting.
- sirherrbatka
- Posts: 2159
- Joined: 07 Aug 2011, 17:21
Re: No-boy ca-re due to lack of non-x86 testers and develope
because morrowind has a sick number of objects to render in every cell cpu must work heavy to "explain" for GPU what to render. If we could reduce number of batches performance would increase greatly. Gus worked on so called texture atlas and got huge performance gain:
long time ago, but the problem is that this method won't work in morrowind case because of large hi-ref textures used in replacers. At the moment there is only a plan for optimize openmw but this demonstrate what can be done.Beshara:
1 texture 400 FPS
"normal" OpenMW 100 FPS
Vivec:
1 texture: 150 FPS 30 batches
"normal" OpenMW: 40 FPS 500 batches
Re: No-boy ca-re due to lack of non-x86 testers and develope
There is no need for any changes. Android is little endian AFAIK, so there shouldn't be any problem with the current code base. This whole topic is a lot noise about very little anyway. The endianness only affects system that are a low priority target for OpenMW (e.g. older per-Intel Macs or systems running Solaris). While it would be nice to support these too eventually, there is no urgent need to act.But we need to answer one question: if the development goes on like it is doing now, will we be able to port OpenMW to different architectures? Or do we need to change something in the development process right now in order to let us port it in the future?
If your average Android hardware can handle OpenMW is an entirely different question though. As sirherrbatka explained already that Morrowind content is poorly optimised. We can partially compensate that by adding some optimisations to OpenMW. The only way to find out if that is sufficient, will be to try it out.
Re: No-boy ca-re due to lack of non-x86 testers and develope
My computer is an x86. OpenMW runs ok on my computer! It was a little difficult at first, but it sort of fixed itself when I goofed around with it.