This is the official Trismegistus with Ex 0.9 download thread

#21
I just setup some tests, and it does indeed look like Som is drawing multiple frames (without ever committing the first one) when leaving the intro screens and whenever leaving a menu. In the case of the "Take this item?" menu the message box I setup actually crashes the game, which is something I've only scene before when Som didn't close its BeginScene() with an EndScene() call after changing the display device or resolution.

I kinda wonder if Som might be drawing multiple frames that never see the light of day whenever things get slow. That would definitely explain the slowness anyway. It polls the inputs several times between individual frames, which is not something you'd really expect given the general shoddiness of Som in general. It's probably not happening but it's probably worth investigating just to be sure.
Reply

#22
Wow... this is really bizarre. Today Dark Destiny is playing pretty damn fast for me and not slowing down at all while moving!?

The polls per frame number is also extremely high, the keyboard is getting checked 31 times per frame at 60fps??? It's pretty much playing like it always does... it gets slow in the bad parts as usual, but that's to be expected.

I'm pretty positive I did not change a damn thing programming wise. The only things that have changed is I disabled DirectSound hardware acceleration and I just ran a program that fixes my mouse when its advanced features are not working. I doubt the mouse could explain it because I do that all the time, though I'm kinda new to this computer. The only other explanation is just a random Som bug running amok that doesn't always kick in.

So hopefully there was just something peculiar to my computer causing the pathological slowdown. 31 ppf sounds really weird. But it may be Som just polls the inputs as much as it can for some reason or something. It's methodology in that dept does not seem very sound. Like I'm pretty sure checking the mouse that many times won't return any new data because mouse data is relative and its resolution is not that high. I know anyway the system I've setup to cope with Som's oddball input polling can't likely help at that rate. I guess I'll have to up the maximum Confused
Reply

#23
Shit it was the hardware "acceleration" ... WTF!?

It's good thing that sound card would pop on sound effects... otherwise I would not have found out short of setting up a custom profiling framework (which might yet happen one day)

This is probably why Microsoft dumped hardware acceleration for DirectSound (see another recent post of mine) for Vista/7. I guess I will have to revisit that post.

My guess is the card was trying to deliver on the 3D sound somehow, and the player moving would cause it to recalibrate its sound effects etc. It pisses me off you can't disable acceleration programmatically or even force things into software mode (near as I can tell it just doesn't take)

I swear most people in technology industries have no fucking clue what they're even doing. It's just crap everywhere you look.

EDITED: The only way to disable acceleration is to modify the registry. I'm assuming a restart is not required. For XP players it might even be worth setting up a message box to explain that they have hardware acceleration on and probably shouldn't / might cause problems for them ... and let them disable it in one click. The only problem is if you actually have a nice sound card or don't have like a dual/quad-core (pretty common now days) setup or hyperthreading™
Reply

#24
PS: Trismegistus is not super fast still, or near fullspeed, but it's serviceable / generally smooth sailing, even though sometimes it hangs around 10fps. Still feels realtime / playable.

Btw... if you wanna know what the framerate is for this build, you can add <fps/> to the window title setting in the Ex.ini file. Or you can uncomment or whatever the stuff in the [Output] section.

PPS: @Todd, you really should replace those custom columns in DD with something more practical. They really are game killers. I suspect they have way too many polygons or something.

EDITED: Strangely enough the crazy polls per frames numbers (31 at 60fps 100s at lower framerates) only happen in the Colosseum map. You can see ppf in the Output stuff (F5) ... it's a number I've not been aware of for very long. Those numbers seem pretty absurd though. Other maps without skies seemed pretty regular ppf wise.
Reply

#25
(2011-01-17, 06:37 AM)Holy Diver link Wrote:[quote author=Verdite link=topic=428.msg5121#msg5121 date=1295197723]
I got no slow down on my pc with ex and SoM, dunno whats going on there.

Whatever it is is not Ex related. I just posted it in this thread for the hell of it. Everything seems to drop to its knees whenever Som rebuilds the scene... ie. when you move. It does reveal though that Som is doing a lot more work when you're moving, so Ex could definitely take advantage of that and turn on antialiasing techniques when not moving... which are not needed while moving because everything is moving too much to notice jaggies anyway.
[/quote]

Just retracting this bit in lieu of discovering it was my soundcard causing the extreme slowdown on the move... not Som/my low profile CPU.
Reply

#26
Actually, my new theory on polls per frame is probably there is just some kind of global event that checks the keyboard/joypad independently of everything else. I don't know how else to describe it. The base number changes with each map. It multiplies when the framerates drop down, suggesting if the theory holds that the events are probably synchronized with the timers regardless of whether or not the frame can be drawn in time.

Anyway... this will make doing analog right more complicated... the mouse doesn't do analog right on the Colosseum map because the 31+ polls break the smoothing algorithm (signal smoothing / not camera smoothing)

@Todd,,, do you have like about 30 of any one kinda thing on the Colosseum map Confused
Reply

#27
^It turns out that every "always on" event that does something as far as I can tell raises the number of polls per frame on that map. I can't think of any plausible justification for this, or use for it under any circumstances... ‎  so I'm assuming it's a bug / will have to find time to isolate the real input poll versus the phantom ones.

If you can think of any reason an always on event should care about the state of the keyboard/joypad you win a cookie (of the internet variety) Coffee

Hell, at least a lot of new bugs are being found out Smokin


PS: We can only wonder if there are any other side effects thanks to this bug. Like what else does each new "always on" event cause to happen Geno
Reply

#28
I think i could guess why... Its probably because the always on is probably like a data attachment to the character... So basically everywhere they go, the system will be checking i the event triggers are available yet. So yeah, i think always on -> data attached to character (and is read by the game constantly?)

You know about the "take this item" discussion we've been having... Well you know you've thought about it too much when you have a dream about it, which i just had... ‎  Dazed noo.
Reply

#29
(2011-01-18, 06:14 PM)Verdite link Wrote:I think i could guess why... Its probably because the always on is probably like a data attachment to the character... So basically everywhere they go, the system will be checking i the event triggers are available yet. So yeah, i think always on -> data attached to character (and is read by the game constantly?)

You know about the "take this item" discussion we've been having... Well you know you've thought about it too much when you have a dream about it, which i just had... ‎  Dazed noo.

Well "always on" events by definition are not triggered. The trigger event was the first one I tried. It's the only one that would make any sense, but that would be a damn hackish way to do it. But anyway, the explanation is surely more like the input data gathering routine is just in sync with the timer routine (since it's independent of the refresh rate) ... who knows why it works this way. It's pretty obvious its not intentional. Som has so many bugs that all have about one thing in common... they don't jump right out at you until you've spent a minimum of time with it. Which obviously From' never did...

I'm glad anyway Sony has a bare minimum of quality standards... otherwise the console games would probably all be this bad too Rolleyes
Reply

#30
The collosseum has a high number of npc/enemies that appear dependant on counter triggers...every other map has very few if any. ‎  I dont know if that may have anything to do with it.

Other things the game may monitor that could affect your numbers randomly between maps that are not event based -
-dropped items staying on the map, including gold that is not picked up.
-whether or not single appearance enemies have appeared, im sure there must be some variables tracked for that
-the status of a chest that has/has not yet been opened.
-the status of a door that has/has not been unlocked.
-enemies that only appear a limited number of times is probably tracked in the memory you are looking at

Another thing, are the number of polls at all affected by the draw distance in a map?
- Todd DuFore (DMPDesign)
Site Founder
Reply





Users browsing this thread:
6 Guest(s)