Oh well, thankfully that won't make a significant difference (less than a kilobyte I'm sure). Besides, my testing is still internally consistent, and if other people will use the encode I provided it will be consistent with their results as well.
I've updated the post above with new data.
[EDIT]
Alright, done with motion estimation testing. Short summary: never use SATD Exhaustive (me=tesa); use uneven multi-hexagon at high range if you want a lot more speed at very little cost.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
Alright, I see. You're copypasting when it syncs and playing by yourself when it doesn't!
Can you estimate how many of these 742 frames of improvement come from Comicalflop's run, and how many from your own input?
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
Test system spec:
Intel Core i5-750 (4 cores @2.67 GHz, everything on defaults);
4 GB dual-channel DDR-1333 (not that it really matters);
Windows 7 64-bit;
x264 rev. 1602 64-bit from http://x264.nl/ with threads=auto.
Sample length: 3600 frames.
The Gritty Details, vol. 1: me=esa vs. me=tesa vs. me=umh.(Bitrate in kbps, size in bytes, speed in fps.)
Conclusion #1.
The quality/size difference between esa and tesa is absolutely negligible. While the speed hit is fixed regardless of the search range (~15%), improvements in size are within at most 0.3%, and they happen only at the point where the whole thing is slow as hell anyways. Meaning this time is better spent elsewhere.
Conclusion #2.
Uneven multi-hexagon search at high range outperforms exhaustive in quality/speed ratio until a certain threshold, where motion estimation quality stops rising (see conclusion #3). Exhaustive search can still improve quality in this case, but at an incredible performance hit: up to ~4% reduction in bitrate at up to 10x lower encoding speed. The value is limited by the video's resolution (320 for Genesis).
Conclusion #3.
The sweet spot for umh is ~192, increasing it past this point doesn't hurt speed much, but doesn't return any benefits either. For (t)esa, efficiency drops rapidly after ~160, while setting it to over 200 equals to torture.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
So I took the same segment and tried to make a RGB32 capture, but ran into a problem with colors. So I made a YV12 capture instead.
FFV1: http://www.mediafire.com/?noziymmdjjz (AVI, 147 MB)
H.264 lossless: http://www.mediafire.com/?0yymim2ynhz (MKV, 17 MB)
Going to run a few tests on it as well.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
ME range isn't about the amount of motion, it's about pixels traveling a maximum distance to be still recognized as vectors. For instance, if at least something in your game is scrolling at 32 pixels/frame, then merange=32 is warranted to pick it up. If nothing does, there will be no benefit whatsoever.
But I'm fairly sure that the motion search value is essentially multiplied by the amount of reference frames; i. e., with merange=32 and ref=16 it will search not only the 32 pixels around the given pixel for one with the same attributes compared to the previous frame, but comparing to 16 previous frames. This needs confirmation, though.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
It may (and often does) need less time due to the difference in production, because a successful attempt at an unassisted speedrun takes only as much time to do as is needed to get from start to finish, since it's done at normal speed. To illustrate, one of the Trauma Center runs on SDA took Lucid Faia two attempts to do (because he's that good). So please don't escalate this subject.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
Depends on how much time it will take to develop such wrapper for Crysis-level games. In 6-7 years (an optimist, aren't I) an average machine would probably be enough!
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
Well, I'm not wellbe6's mother, and I hope neither of us is, but it feels strange to me that we educate and foster newbies on our newbie board and treat them with hostility if they happen to post on the main part of the forum.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
Now, let's not be rude. The guy is obviously not experienced with forums (how experienced can a 8 year old be?), mocking him or driving him away without letting understand what exactly he is doing wrong would only make his behavior worse, even if it won't be us who will have to deal with it in the future.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
Not having seen gocha's TAS in a couple years allowed me to once again shake my head in disbelief at the marvelous hole-in-ones. How much brain power does one need to think such complex trajectories through?
Anyway, thank you, this run was great. Please TAS more!
Also, many thanks for the cutsceneless encode, gocha, it was very nice.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
In my opinion using sci-fi as a medium to convey some kind of sociological, political, moral or other such type of message does not necessarily make it good sci-fi, nor is it a requisite. A sci-fi story can be good even if it does not contain any deep message, or even deep and involved story-telling.
Prerequisite or not, it's how it's been done originally for nearly the first hundred years of sci-fi's existence, and I think it's a good thing. If a work of fiction has something else to say aside from describing unrealistic events happening in unrealistic circumstances, it only increases its value, doesn't it? I believe it's not even the fact of political, social, or philosophical subtext that matters, but how well does a book allow to read deeply into itself. But then again, there's this brilliant episode of South Park. :D
Warp wrote:
I once read the opinion that one requisite of good sci-fi is that it tells a story (or contains relevant elements, plot points or messages) which couldn't be easily told in a non-sci-fi story (usually also non-fantasy, as the line between sci-fi and fantasy can often be quite blurred when talking about story-telling). If a sci-fi story could be told equally easily and fluently in a completely non-sci-fi non-fantasy setting (for example by moving the setting to European colonialists invading some native cultures somewhere, thus introducing a significant difference in technology between two cultures) then it makes the whole point of it being sci-fi kind of moot, from the story-telling point of view. The sci-fi becomes decoration and not an integral part of the story.
This opinion has merit, but it really is dodgy, because the sci-fi being a decoration and not an integral part of the story boils down to a matter of opinion.
Is, say, Neon Genesis Evangelion (excuse the trivial example) a series about giant robots fighting giant aliens with biblical symbols to save the world from an impending disaster? That can't easily be told in non-fiction, that's for sure. Or is it a series about fragility of humans and their relationships and problems of self-determination? This requires as much fiction as going outside and talking to the next person. So which of the two is a decoration, and which one is the real thing? Maybe I'm mistaken, but there can never be a solid consensus.
Warp wrote:
The Matrix trilogy could probably not be told very fluently in a non-sci-fi setting.
The main issue with the Matrix is that it can't be told fluently at all. :D
But you know, it would be an interesting challenge to try and describe the events of the 2nd and 3rd movies in a concise manner without touching any of the symbolism every scene is so rich in and still have it make sense.
Warp wrote:
Watching John with the machine, it was suddenly so clear. The terminator wouldn't stop, it would never leave him. It would never hurt him or shout at him or get drunk and hit him or say it was too busy to spend time with him. And it would die to protect him. Of all the would-be fathers that came over the years, this thing, this machine, was the only thing that measured up. In an insane world, it was the sanest choice.
Yeah, I actually consider Sarah's narration to be one of the things that make T2 a good movie, heh.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
Weird, I thought he was present at at least one of them. Or maybe I'm confusing that with some members visiting him.
Well, whatever, let's not interfere with DevilSpree asserting himself on a site where admins can do whatever they want.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
When I say "in person", I mean you can go there and see him play in front of you like he does on the videos. It has been done. Shame that you're still in denial.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
(Bitrate in kbps, size in bytes, speed in fps.)
Conclusion #1.
The quality/size difference between esa and tesa is absolutely negligible. While the speed hit is fixed regardless of the search range (~15%), improvements in size are within at most 0.3%, and they happen only at the point where the whole thing is slow as hell anyways. Meaning this time is better spent elsewhere.
Conclusion #2.
Uneven multi-hexagon search at high range outperforms exhaustive in quality/speed ratio until a certain threshold, where motion estimation quality stops rising (see conclusion #3). Exhaustive search can still improve quality in this case, but at an incredible performance hit: up to ~4% reduction in bitrate at up to 10x lower encoding speed. The value is limited by the video's resolution (320 for Genesis).
Conclusion #3.
The sweet spot for umh is ~192, increasing it past this point doesn't hurt speed much, but doesn't return any benefits either. For (t)esa, efficiency drops rapidly after ~160, while setting it to over 200 equals to torture.
