Topic by adelikat

A real DS and DeSmuME runs in 59.82fps (more precisely, 33513982/6/355/263) but the emulator treated the framerate as 60fps when it shows a movie length in the dialog. Then, I commited a quick fix to the svn yesterday. After posting the fix, I noticed that TASvideos also calculates a movie length in 60fps. This wrongly shortens a movie length a few seconds X(

I am usually available on Discord server or Twitter.

Posted: 7/21/2009 1:49 PM

Quote

Captain_Forehead

Player (79)

Joined: 7/7/2008
Posts: 873
Location: Utah

I still got nothing.

Posted: 7/21/2009 3:54 PM
(Edited: 7/21/2009 4:46 PM)

Quote

arukAdo

Experienced player (961)

Joined: 12/3/2008
Posts: 939
Location: Castle Keep

After a few tests on DoS and PoR i can tell 9.4 is more laggy than 9.2, it create lag where there wasnt before, not in "huge" quantity but simply few frames here and there It would be great to have some words from authors (zeromus ?) because it ruin movies making Using 9.4 in its current status to do a castlevania movie is clearly a waste All new features are cool, the emu seem to run great, but this lag problem is just too much an hassle for players and for doing TAS, i realy hope an answer or some infos on this, else i guess whe are compel to use 9.2+ .... and no (castlevania...) movies will be produced under 9.4 I want to add here : all those small problems like frame advance and whatnot, i dont care, im fine with them and i will not whine on that matter, it doesnt impact much on movie making or on movie stability, while the lag ruin everything... Edit: well ill take this oppertunity to request something usefull, if whe could have some kind of "reload movie" feature, it would allow to edit the movie file while its playing (very usefull for testing), for now whe need to stop movie and then go file/play movie dialog box, witch is kinda long and not productive

Tasvideos.org personal page

Posted: 7/21/2009 4:44 PM

Quote

moozooh

Senior Moderator

Joined: 8/4/2005
Posts: 5777
Location: Away

Maybe it's worth waiting for stabilization of the emulation core then. You knew full well it was far from final when adelikat released the first rerecording version. No point in complaining that it ruins something now. For the record, this was the precise reason I wasn't too big on the idea of porting rerecording features to immature emulators.

Warp wrote:

Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.

Posted: 7/21/2009 4:53 PM

Quote

Captain_Forehead

Player (79)

Joined: 7/7/2008
Posts: 873
Location: Utah

Yeah, maybe I should just wait.

Posted: 7/21/2009 5:06 PM

Quote

arukAdo

Experienced player (961)

Joined: 12/3/2008
Posts: 939
Location: Castle Keep

I "complain" because i happen to be tasing castlevania and sadly for me it have extra lag in 9.4, so i cant use it (also castlevania is needing sram... thats why im interrested in 9.4) Apparently some others games dont have this problem, even are getting less lag Also moozooh, i dont see how sarcasms on early release help at all, i would call that trolling, sry, i dont want to look offensive, but your not even a taser....

Tasvideos.org personal page

Posted: 7/21/2009 5:11 PM

Quote

moozooh

Senior Moderator

Joined: 8/4/2005
Posts: 5777
Location: Away

arukAdo wrote:

I "complain" because i happen to be tasing castlevania and sadly for me it have extra lag in 9.4, so i cant use it

Then don't use it.

arukAdo wrote:

Also moozooh, i dont see how sarcasms on early release help at all, i would call that trolling, sry, i dont want to look offensive, but your not even a taser....

There was no sarcasm whatsoever in my previous post. The responsibility for using an unstable emulator is yours, and yours alone. (Don't worry, you look offensive either way.)

Warp wrote:

Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.

Posted: 7/22/2009 6:29 PM

Quote

zeromus

Editor, Emulator Coder

Joined: 8/7/2008
Posts: 1156

arukAdo wrote:

It would be great to have some words from authors (zeromus ?) because it ruin movies making

Lag is of no concern to me. I am trying to make the emulation flawless. That means changing the timing sometimes. Sometimes it makes more lag, sometimes it makes less. You'll thank me later.

Posted: 7/23/2009 9:38 AM

Quote

andymac

Experienced player (623)

Joined: 11/30/2008
Posts: 650
Location: a little city in the middle of nowhere

Priority one is to get it working, and the second priority is to make it run fast. This is true for any programming cycle. This is quite a broad generalisation, there are definitely exceptions to this, but DesMuMe is not one of them. Don't complain about speed, beleive me, they already know.

Measure once. Cut twice.

Posted: 7/25/2009 7:09 AM

Quote

zeromus

Editor, Emulator Coder

Joined: 8/7/2008
Posts: 1156

andymac wrote:

Priority one is to get it working, and the second priority is to make it run fast. This is true for any programming cycle. This is quite a broad generalisation, there are definitely exceptions to this, but DesMuMe is not one of them. Don't complain about speed, beleive me, they already know.

Well, I exert considerable energy making it run fast also. But lag isn't running slow. Lag is bad emulation. An artifically laggy game is a buggily emulated game. So in the case of lag, it IS just trying to get it working. Desmume lags because of features which arent emulated yet except with abysmally timed hacks. Also I should add that I have now personally witnessed mysterious desynchs in 0.9.4. I am looking into it but haven't figured it out yet.

Posted: 7/25/2009 12:30 PM

Quote

adelikat

He/Him

Emulator Coder, Site Developer, Site Owner, Expert player (3574)

Joined: 11/3/2004
Posts: 4754
Location: Tennessee

9.4 and 9.4 dev will be accepted emulators for site submission. HOWEVER, do not use them. Everyone should use 9.2+ for their movies. It has all the tools of 9.4 and then some (including fixed frame advance). In addition, it is the same sync as 9.2 resulting in better comparisons and compatibility for all site movies. 9.4 has timing differences but none are the results of "better" emulation. Just bad timing in a different way that 9.2's bad timing. So there is really no benefit to using 9.4.

It's hard to look this good. My TAS projects

Posted: 7/25/2009 1:27 PM

Quote

gocha

Any

Emulator Coder, Former player

Joined: 6/21/2006
Posts: 401
Location: Japan, Nagoya

adelikat wrote:

9.4 has timing differences but none are the results of "better" emulation.

At least, 0.9.4 has a better sound quality (higher resolution, stream playback fix) than 0.9.2+.

I am usually available on Discord server or Twitter.

Posted: 7/25/2009 5:26 PM

Quote

ShinyDoofy

Former player

Joined: 12/5/2007
Posts: 716

I have an issue with a recorded avi file. There are 110914 video frames (spread across 6 <=2GB files encoded with ffv1). The site's code calculates a movie's length with 60 fps, desmume's avi runs at 59.826 fps. The sound has a length of 1866.4 seconds, this results in 59.4269 fps for the video to sync. Is that the real framerate? Frameskip was set to to 0 for avi recording, together with Cosine interpolation, ADPCM caching was off (is there an option to turn off the latter on the GTK build?). I thought I had seen some svn commits for the main branch that - is anybody else seeing this with gocha's binary from 7/25/2009 12:02:21 or am I doing something wrong? /edit: It also desyncs when playing the video with a WAV dump. :( //edit2: Nevermind, I screwed up in fiddling with the video stream.

Posted: 8/11/2009 2:51 PM

Quote

DDRKhat

Player (37)

Joined: 9/9/2006
Posts: 388

Any chance of getting a "Begin recording from savestate" ?

A whisper in the wind~~

Posted: 8/11/2009 4:59 PM

Quote

adelikat

He/Him

Emulator Coder, Site Developer, Site Owner, Expert player (3574)

Joined: 11/3/2004
Posts: 4754
Location: Tennessee

DDRKhat wrote:

Any chance of getting a "Begin recording from savestate" ?

Eventually.

It's hard to look this good. My TAS projects

Posted: 8/11/2009 8:12 PM

Quote

DDRKhat

Player (37)

Joined: 9/9/2006
Posts: 388

Is it a hard feature to do or something. are you not able to just view the source Gens and take it's recording coding and implement it? Or is it coded in a certain way that it only works with the Gens application? I can't admit to diagnosing the source too much so.

A whisper in the wind~~

Posted: 8/16/2009 9:44 AM
(Edited: 6/28/2021 3:41 PM)

Post subject: DeSmuME 0.9.4+

Quote

gocha

Any

Emulator Coder, Former player

Joined: 6/21/2006
Posts: 401
Location: Japan, Nagoya

Again, for making a new movie, you should use 0.9.4 instead of 0.9.2, since it emulates things better. One of major advantages is sound emulation. You'll hear really bad sound with 0.9.2, especially when you play a game which uses a streamed sound. Anyway, I've made an improved version of 0.9.4 official! Features My 0.9.4+ has:

Less desyncs (savestate has been updated, but you can read old savestates as well)
Lua scripting with advanced graphic functions (they works only in gui.register callback)
Path settings
State rewinding (it's not tested at all, oh well)
Magnification filters like hq2x
Lag reduction options (details described below)
Other little bugfixes (ex. frame advance improvement)

Still, it should sync fine with 0.9.4 movies. Let's try it now! Note: these codes are ported from the svn trunk. Downloads

Download DeSmuME 0.9.4+ win32 binary (r2964)
Experimental: Download DeSmuME 0.9.4+ win32 binary (r2881) (fixes/causes some graphic glitches after loading state)
Source code is available as a branch in the official svn repository of DeSmuME. Checkout https://desmume.svn.sourceforge.net/svnroot/desmume/branches/rerecording_0_9_4_plus (as usual, I haven't maintained non-windows port at all. However, I think it can be compiled with no big changes. If someone is interested to maintain them, please do)

About Lag Reduction Options Currently, there are 2 options. They increases speed, but they also causes a trouble for some games (for example: impossible to boot), because of whatever desmume's bad timing emulation. - Fix ARM Cycle Count (recommend on) This fixes the wrong cycle counting in armcpu_exec and instruction cycles of {S} instructions. - Fast Fetch-Execute (not much recommended at the moment) If enabled, the emulator will process fetch-execute cycle parallelly. If you want to remove lags which cannot be removed by Fix Arm Cycle Count, it's worth to enable this option. However, keep in mind that the option causes an emulation glitch for some games (see games affected by fast fetch-execute for examples). These options will not be recorded to a movie. Upload logs (JST; GMT+9) 2009-10-15 22:55 - Lua: add LuaBitOp, register manipulation functions, and memory hook functions. (r2964) 2009-09-16 14:20 - Experimental fix for Tetris DS graphical issue. (r2881) 2009-09-14 01:20 - Fix Lua bit operation functions to handle the sign bit (bit 31) correctly. (r2872) 2009-09-09 13:05 - Lag reduction option is separated to 2 options. Added savefile hack menu option from 0.9.4 branch for upgrading savefile sizes when one has been careless and used the wrong savefile size. (r2855) 2009-09-08 13:45 - Improved LagReduction timing. To know what has been changed, see the diff of r2844. (r2845) 2009-09-06 11:00 - Add LagReduction option to INI secretly (r2831; default = 0 (0.9.4 non-dev timing). The timing of LagReduction=1 is different from both of 0.9.4 dev and non-dev). Improved input display a little (r2832) 2009-08-22 23:30 - Lua drawing now works but only in gui.register callback (r2779) 2009-08-18 15:30 - Fix random desync (make 'samples' in SPU.cpp to be zero at SPU_Reset) (r2769) 2009-08-16 17:00 - First upload (r2766)

I am usually available on Discord server or Twitter.

Posted: 9/6/2009 4:33 AM

Post subject: Test video of LagReduction option

Quote

gocha

Any

Emulator Coder, Former player

Joined: 6/21/2006
Posts: 401
Location: Japan, Nagoya

I put a test video of LagReduction option here. http://www.youtube.com/watch?v=ZDM1jxu_TR4

I am usually available on Discord server or Twitter.

Posted: 9/6/2009 7:20 PM

Quote

moozooh

Senior Moderator

Joined: 8/4/2005
Posts: 5777
Location: Away

That's very good news that deserves much more recognition than it has received so far. How does it work? Does it mean it won't lag even in places where lag is naturally occurring?

Warp wrote:

Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.

Posted: 9/6/2009 10:17 PM

Quote

gocha

Any

Emulator Coder, Former player

Joined: 6/21/2006
Posts: 401
Location: Japan, Nagoya

moozooh wrote:

Does it mean it won't lag even in places where lag is naturally occurring?

To reduce lags which also occurs in real hardware is not my attempt, but

Sir VG wrote:

The map screen almost seems to run TOO fast actually. It's weird - it just feels a hair fast to me.

Hmm, is it indeed? For reference: http://www.youtube.com/watch?v=bk0u94bf21M The update is quite simple.

Index: ./src/armcpu.cpp
===================================================================
--- ./src/armcpu.cpp	(revision 2830)
+++ ./src/armcpu.cpp	(revision 2831)
@@ -494,46 +494,49 @@
   return TRUE;
 }
 
+bool fixCycleCount = false;
 
 template<int>
 u32 armcpu_exec()
 {
-        u32 c = 1;
+	u32 c = fixCycleCount ? 0 : 1;
 
	//this assert is annoying. but sometimes it is handy.
	//assert(ARMPROC.instruct_adr!=0x00000000);
 
 #ifdef GDB_STUB
	if (ARMPROC.stalled)
		return STALLED_CYCLE_COUNT;
 
	/* check for interrupts */
	if ( ARMPROC.irq_flag) {
		armcpu_irqException( &ARMPROC);
	}
 
-        c = armcpu_prefetch(&ARMPROC);
+	c += armcpu_prefetch(&ARMPROC);
 
	if ( ARMPROC.stalled) {
		return c;
	}
 #endif
 
 	if(ARMPROC.CPSR.bits.T == 0)
 	{
		if((TEST_COND(CONDITION(ARMPROC.instruction), CODE(ARMPROC.instruction), ARMPROC.CPSR)))
 		{
 			if(PROCNUM==0) {
 #ifdef WANTASMLISTING
				char txt[128];
 				des_arm_instructions_set[INSTRUCTION_INDEX(ARMPROC.instruction)](ARMPROC.instruct_adr,ARMPROC.instruction,txt);
 				printf("%X: %X - %s\n", ARMPROC.instruct_adr,ARMPROC.instruction, txt);
 #endif
 				c += arm_instructions_set_0[INSTRUCTION_INDEX(ARMPROC.instruction)]();
			}
 			else
 				c += arm_instructions_set_1[INSTRUCTION_INDEX(ARMPROC.instruction)]();
 		}
+		else if (fixCycleCount)
+			c++;
 #ifdef GDB_STUB
         if ( ARMPROC.post_ex_fn != NULL) {
             /* call the external post execute function */

I am usually available on Discord server or Twitter.

Posted: 9/9/2009 10:53 PM

Quote

DDRKhat

Player (37)

Joined: 9/9/2006
Posts: 388

DDRKhat wrote:

Any chance of getting a "Begin recording from savestate" ?

Not meaning to pester. This would be helpful for a few little videos I would like to do to provide "proof" of how some things work (I.E some Yu-Gi-Oh! examples)

A whisper in the wind~~

Posted: 9/15/2009 11:16 PM

Quote

FractalFusion

Editor, Expert player (2079)

Joined: 6/15/2005
Posts: 3282

In Tetris DS, sometimes saving and reloading causes part of the screen to shift upward. This occurs when using DeSmuMe-rr v0.9.4+. This does not occur with the official DeSmuMe v0.9.4 or with DeSmuMe-rr v0.9.2+.

Posted: 9/16/2009 2:08 AM

Quote

gocha

Any

Emulator Coder, Former player

Joined: 6/21/2006
Posts: 401
Location: Japan, Nagoya

FractalFusion wrote:

In Tetris DS, sometimes saving and reloading causes part of the screen to shift upward. This occurs when using DeSmuMe-rr v0.9.4+. This does not occur with the official DeSmuMe v0.9.4 or with DeSmuMe-rr v0.9.2+.

0.9.4+ sometimes causes a graphic glitch when a state is loaded. I experienced that the emu ignores the fade level of the screen on state load. If it's also occurs on the latest svn binary, it should be known by official developer team as well. (my fade problem doesn't occur, it seems) Edit: ported a old fix from trunk. It fixes the issue but also introduces a new issue (at least for OoE).

I am usually available on Discord server or Twitter.

Posted: 9/16/2009 5:36 PM

Quote

querschnitt

Joined: 7/7/2009
Posts: 9

hey guys. i can't get desmume 094 to work.. when i start it, nothing happens! it just crashed! and the send error report dialog appears :(

Forum DeSmuME DeSmuME emulator development

Forum

DeSmuME

DeSmuME emulator development