Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
dwangoAC wrote:
I've created a new WIP which completes E1L1 and represents a significant improvement over my previous WIP. The lurch in the first prison cell combined with improved enemy AI manipulation resulted in a faster completion time with no lost health. I'm not precisely certain how much faster this is as I let the movie run a bit longer the first time after pulling the elevator switch but it's definitely faster.
Looks like this new version is 19 frames (271ms) faster than the old one after fade to black at end (I extended the previous movie to pull the elevator switch).
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
Zeyn0 wrote:
Hum, i find something interesting for my problem of "Start from a save state" ; with the Dega emulator, the first input in the controller data format of the movie files (.mmv) should start at 00F0 x 04, but in all my movie files, the first input always start at 62F0 x 04, and i try to just copy and paste at 00F0 x 04 but the movie can't be played after that.
Please help !
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
MESHUGGAH wrote:
edit2: 2170 frames. wall clip optimized better
Holy heck, I thought 2183 was near optimal and you cut 13 frames below that.
One thing, it is possible to use soft reset instead of reset button without losing time (since reset button is unverifiable)?
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
I noticed that in the room with the well known secret, you turn your back on the door while it is opening, and backstrafe on the next room.
Diagonal backstrafing is slower than diagonal forward strafing... Is there not enough time to turn 90 degrees while the door out the next room is opening or is there something on the way?
Quick encode (I had the Lua script on): http://ilari.tasvideos.org/w3d-e1l1.mkv
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
CoolKirby wrote:
xxNKxx wrote:
My real nickname is N?K, but I can't create name N?K at here because rules, that's why I must create name xxNKxx
Thanks for your reply. I'll fix both descriptions so they credit you accurately.
[user:N?K] is wrong. Unfortunately, due to a site bug, it is impossible to do the thing properly[1].
[1] One can't override the link text in user link.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
For debugging the desync, finding a place where it can desync and bracketing the point with savestates (the first savestate desyncs, the second doesn't) as tight as possible would be helpful...
Edit: Looks like I found the bug:
If mouse callback occurs at exact instruction boundary, the events associated with it are emitted with the same timecode as mouse callback.
Now, on playing that back, it is undefined (what actually happens depends on when last event was) if the mouse callback or event callback should run first.
If event callback runs first, the result could very well be a desync.
Fix is to set lower bound for event timestamp to be one nanosecond in the future, as this guarantees that the event callback will run after mouse callback, even if both are during the same instruction.
------
As for the run itself, it looks quite sloppy.
- The angles that should be 45 degrees aren't.
- I would try changing angles to be suitable for the next room during waiting for door.
* 90 degree turn takes about 0.25 seconds. There should be enough time during door opening sequence.
- Getting the machinegun looks to waste fair amount of time.
* Based on SDA speedrun, machinegun won't speed things that much and one is on path (via deathdrop from first officer encountered) on level 2.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
RNG state to index mapping algorithm:
linear_transform = function(x, a, b)
return (math.floor(a / 65536) * (x % 65536) * 65536 + (a % 65536) * x + b) % 4294967296;
end
rng_state_to_index = function(state)
local A = 0x2925141D;
local B = 0xD6DAEBE3;
local C = 1;
local D = 2;
local i;
local j = 0;
for i=0,31 do
if (state % D) ~= 0 then
state = linear_transform(state, A, B);
j = j + C;
end
B = linear_transform(A, B, B);
A = linear_transform(A, A, 0);
C = C * 2;
D = D * 2;
end
return j;
end
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
AngerFist wrote:
Well to be on the safe side, I am going to restart the whole run using 2.2.1. I had merely tased a thousand frames before so no harm done.
Er, just check if the exiting movie file syncs on all ROMs on 2.2.1 and if it does, just switch to that version then (the movie file format is compatible)?
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
BrotherMojo wrote:
Ilari wrote:
Not faster in long run: 60 sp/f (running) vs. 64 sp/f (jetpacking).
...wow, really? Doesn't running normally move you at 55 sp/f? The conveyor adding only 5 sp/f to that seems pretty strange, especially compared to how slowly you move when running opposite to a conveyor.
I just checked again. The figures (60sp/f and 25sp/f) are correct. Surprising for sure.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
AngerFist wrote:
Now I really just need someone to assure it wont desync using 2.1.5 movie file in 2.2.1..
Why not check if the movie syncs with 2.2.1 and if it does, start using that version on both computers?
AFAIK, there are sync differences between 2.1.5 and 2.2.1... May not be relevant ones, but there are.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
BrotherMojo wrote:
How about running on the conveyors that are travelling the same direction you are? It feels a bit faster than Jetpacking when doing it over long distances, but that could be an unfounded belief, and on top of that I'm not sure about in short bursts.
Not faster in long run: 60 sp/f (running) vs. 64 sp/f (jetpacking). But in short bursts, situation could be different since there is no acceleration for running.
I did some peeking on speed and acceleration data. Found all sorts of odd stuff like:
* Horizontal acceleration in air depends on your horizontal speed.
* The acceleration of gravity is not the same for jumps and jetpack cutoff.
* If holding alt while running off ledge (no fuel), the game treats you as jumping instead of instantly resetting your velocity like falling off cliff.
* If jetpacking from standstill, there is no horizontal accelration the first frame, but if jetpacking from run, there is acceleration on the first frame (added to your existing speed).
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
MetalStorm wrote:
Not entirely sure how to do this, but is there a way to display text without any background?
You mean with the game showing underneath?
As for doing that with lua-gd, does specifying the background color using image:colorAllocateAlpha() to be fully transparent work?
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
Lex wrote:
It was likely a coincidence unrelated to anything FCEUX did. FCEUX is free, open-source software. If there was something devastating in the FCEUX code, it would likely be found and removed by any of the various FCEUX devs and power-users.
Also, aren't operating systems are supposed to protect the system from damage from misbehaving applications?
Well, there is a possibility that a movie file from someone malicious could exploit FCEUX, but I view this as quite unlikely, and such things likely wouldn't cause the computer to "die".
Also what does "computer died" mean here?
- Crashed?
- Powered down?
- Crashed and won't boot again?
- Crashed and booting leads to blue screen of death?
- Powered down and won't boot again?
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
Svimmer wrote:
(bad image embed)
That's not valid image to hotlink.
Futhermore, that site seemingly offers the download as an .exe file (obviously unsafe).
Edit: After edit the hotlink still seems to be bad, but at least it doesn't offer an .exe.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
pacaya_fr wrote:
SoulCal wrote:
Btw, can I ask you to run my dtm file? You seem to have the correct ISOs to run the game, but you'd need to use a different Dolphin build.
My PC is very slow, and it was a **** to use Dolphin. I'm not sure I will be able to run it : I had a lot of crashes ! You want me to run your dtm file to watch ranking info ?
I think he wants you to just to see if the whole DTM runs properly.
But if you were able to extract ranking data in addition, that could be a nice addition...
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
natt wrote:
Are the missing sound effects OPL generated? If so, is the emulated game picking up the sound card correctly?
The initial startup messages say it is picking up SB (which contains OPL chip)...
Also, just tested, the level 0 has at least gem pickup sounds (which are OPL)...
Due to how OPL emulation is hacked in[1], starting dumping from any point after OPL chip has been used does not work correctly...
[1] Specifically, the emulator records all writes to OPL chip and OPL is emulated later. If starting dumping from latter point, the emulated OPL registers will contain wrong values, leading to glitched / missing sounds.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
slamo wrote:
Setting the CPUDiv to anything above 50 doesn't seem to work for me...I can get into the game and everything, but when I press "S" to start at the main menu it doesn't respond.
http://tasvideos.org/userfiles/info/6878890210379818
Settings & co:
- CPUDiv is 125 (8MIPS)
- Memory size is 384 (1.5MB)
- FPU is disabled.
- SB is detected.
Enters level 0 (spams up to make warpin animation faster, don't ask why).
Important addresses (valid for this startup):
- 00025CA3 (word): X position (32 subpixels per pixel?)
- 00025CA1 (word): Y position (32 subpixels per pixel?)
- 00027040 (word): Fuel (10000 is full, 0 is empty, seems to deplete at 10 units per frame)
- 00025872 (dword): Random number generator state.
The RNG formula is: x' = 0x15a4e35 * x + 1 (mod 2^32)
Edit: Corrected some addresses and added RNG.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
Svimmer wrote:
I have the setting turned on to receive notifications for replies, but I didn't see any in recent times
I heard reports about E-Mails not working and tested it myself, worked.
In that progress, corrected EHLO to match RDNS (this was like one day ago).
So I have little idea what is causing those delivery failures... :-/
