Menu
▾
▴
#962 Super Robot Taizen : OG SAGA Underoptimised.
Like the previously closed comment, the game is running slower then it should be. Performance is off the scale in all other games bar this one (which works fullspeed in nogba)
It Definitely appears to be inconsistencies in the core emulation, as i can enable all manners of settings (including the advanced SPU) and it will still (only) reach 54fps. changing the magnification filter also does not effect the FPS.
Games which are far more intensive gfx wise are running 60fps constantly with ease, so the only explanation is this game is running some intensely underoptimised functions in the backend somewhere.
i7 920 @ 3.4Ghz, GTX275.
its something that definitely needs looking into because something is definitely not right about the performance (especially when i can emulate windows emulating a PS2)
Discussion
Anonymous
p.s. advanced spu is slower, thats why it is optional
I run 49fps in the main engine on my c2d T7700 @ 2.4ghz. This seems about right for a game that runs at 99% ARM9 load, which is no emulator bug, but just the way the game is coded. Try making a profiler guided build. I don't want to hear comparisons to other emulators, things are the way they are for reasons I am not going to debate.
Ok, now that i know your numbers i can say for sure theres a issue with the emulator. We are both getting exactly the same FPS with no frameskipping, and yet i have a cpu that even in single threaded apps has a 40+% performance increase over yours at stock (and im running overclocked).
Theres no way this is a cpu limitation, Something is seriously choking the arm9 in this game that it doesn't matter how fast your cpu is or how much more efficient it is, it just can't keep up.
Theres absolute no way in heck that this game would have 0 scaling when going from a t7700 to a i7.
i told you, make a profiler guided build. I get 37FPS in a release fastbuild with no profiler guided optimizations. 37*1.4 = how many fps you get
Recent svn build seems to "shave" off quite a couple of fps from games and I do get about 45fps(avg) for this game.
T9400 2.53Ghz
Sometime soon I'll do a comparison to 0.9.5 release and figure out whether we lost speed and why. But I agree, if its any, its not very much. (desmume still needs all the help it can get)
I did a check on a few games comparing 0.9.5 with r3377(without frameskip) and it drops approx. 3-7fps (5fps avg). Pokemon Platinum drops about 10fps in big cities(Hearthome), fps will maintain consistent when cycling but drops suddenly if you stop cycling and do nothing(0.9.5 drops about 1fps whereas r3377 drops about 4fps).
Maybe it is my r3377 is not a profile guided build.
Also, I notice that frameskip is not working as well in recent build especially after skip5. I used to run with frameskip9 for my games and work very well except for some that need frameskip0.
Pokemon Platinum used to get flat 60fps(not in big city) but it now dropping to 54-60 and in big city(Hearthome), fps is about 50-55 and cycling is sliding whereas it was 57-60 previously for frameskip9 (previous svn with current svn). Not sure if there has been changes after ninsutja make the frameskip changes.
there hasnt. it must be an illusion. i dont see how anything you said has to do with the frameskipping. frameskipping "not working well" is not defined as failing to get the framerate you want. sometimes frameskipping every single frame can't even get up to 60fps.
dont bother comparing profiler guided to non profiler guided builds, there will be big differences in speed.
I am sure you are right since you are a developer.
I compare r3377 to r3220 for Pokemon Platinum with same settings and fly to Hearhome city (frameskip 9).
r3377 - cycle to and fro and fps (50-53) and you see the bicycle sliding instead of the leg doing the cycling.
r3220 - cycle to and fro and fps(59-60) and you see the bicycle riding with the leg doing the cycling.
I am sure my eyes is not playing trick on me.
If I revert out r3221, r3224 & r3225 from r3377, then I get the same speed improvement as r3220 i.e. I get the same fps etc in Pokemon Platinum as in r3220 for the same tested scenario.
You can use Indiana Jones and the Staff of Kings to do the comparison and the speed improvement is obvious.
It is possible that those revisions may have affected something.
Are you sure you havent chosen to use "directdraw software-forcer" added in r3221?
Can you send me a savefile where I can watch you fly to hearhome city?
And finally, are you aware that when you are frameskipping, depending on the exact framerate, you may miss frames such as legs going up and down? (but I will verify it myself if you send dsv)
and since you issued the challenge, I tested staff of kings in the very first 3d area of the game and get 69-70fps in both 0.9.5 and a brand new profiler-guided svn build. Perhaps the newer directdraw code is playing poorly with your system
1. I am not challenging you because you are a developer and I am a gamer.
2. I use "DirectDraw-HW" (always).
3. Do you mean dsm(savefile?) or use external apps to capture my test scenario?
4. Yes, I am aware that leg riding movement can be skipped during frameskipping just as Quantum game(Bond walking become sliding @60fps until it was fixed somehow).
5. I thought frame limiter limits fps to 60 and if you get above 69-70fps for r3377 and (r3377 minus new directdraw codes), there is no basis for comparison. If you run thru' with a non-profile guided build < 60fps, there could be a difference (as I did). Again it is not a challenge but how I look at your reply.
6. If you seriously feel that I am challenging your technical competency and upset about it, just let me know and I will stop posting bug report/comment.
7. You probably don't like my style of bug reporting just as I don't enjoy reading your comment.
8. If you still want me to attach my dsv after reading my reply, let me know.
quit being so sensitive. If everyone agreed then there would be nothing to talk about. Since we are talking, there is disagreement, and there are challenges and proofs.
You are a gamer and I am a developer. I don't have time to play games or barely to use the emulator at all. You have some insights that I do not.
I mean send a dsv so I can watch you riding a bike across the sky or whatever that game does.
When judging quantitative performance regressions, you must turn off the frame limiter, and the frame skipping. That is what I did in staff of kings.
When judging quality of frameskipping, you must turn on framelimiter and get the game running slow. I did that on staff of kings also, but I am not confident that I can tell any difference.
If the directdraw changes that you reverted have any real consequences, then it must simply be that it is slowing down the emulator on your system, and that results in qualitatively worse frameskipping... but only because your system is having that much more trouble due to a lower framerate--which you can only prove by turning off frameskip/framelimit and doing a precise performance measurement.
Download rar file from http://www.sendspace.com/file/c6rile
Pokemon Platinum:
limiterOff/frameskip0 limiterOn/frameskip9
r3377 40-45 50-55
r3377- 47-49 59-60
When limiter is off & frameskip 0, fps will "fly" off the roof during intro, when flying to another city etc.
The thing that I don't understand :
when frameskip9, both fps increase by about 10fps but riding bicycle is lierally sliding for r3377 while it is quite smooth with leg riding movement for r3377(minus new directdraw codes).
zeromus,
I have attached my dsv as you have requested.
May I know your conclusion?
i finally got around to looking at this but the file had expired or something
If you are talking to me (see reply below) otherwise ignore.
I deleted the file after I uploaded it.
I don't keep r3377 anymore (no longer interested).
with r3421, I am able to get 53-60 with limiter on/frameskip 9 (it is no longer an issue to me and no more sliding movement during cycling - can see leg movement).
You can close this discussion since the file is deleted by sendspace.