Forums

I'm part-way through a reworking of in_vgm. Skip past the first paragraph if you don't care why.

in_vgm was the first thing I ever wrote in C. It shows. It's a big spaghetti mess of every coding style, naming convention, beginner error and confusion you can think of, the most pressing of which being, almost everything is in a single monster file. So I've rebuilt it from the ground up based on (hopefully) cleaner abstractions and interfaces, although it's still pure C for maximum portability, with OO-ish abstractions to make it possible to do things like decode 2 files at the same time. In theory, this means it ought to be possible to support FB2K, WMP, XMMS, etc with minimal work, just plumbing the APIs into each other. I make no guarantees of this; Winamp is still my primary target.

As it is at the moment, it has just become able to play Master System, Game Gear and Mega Drive VGMs. Almost every other feature is not done yet. So here's what's done:

- Plugged in my SN76489 core to support SMS, GG, Coleco, BBC etc
- Reworked YM2612 core from Genesis Plus GX to be re-entrant (untested), and plugged it in, to support Mega Drive/Genesis
- Plugged in EMU2413 core
- Implemented YM2612 sample optimisation support

Here's a list of features in the current in_vgm that aren't done, but definitely will be (eventually):

- Seeking
- Looping support (including Loop Forever)
- Per-channel muting
- YM2151 support
- Metadata support - including titles, lengths, Media Library support and a bunch of "guessing" stuff I implemented in the current beta to get things like track numbers and album artists
- Transcoding/CD burning/Replay Gain calculation support
- Fadeout on end
- Replay Gain
- File Information window
- Configuration window
- VGM7z support (even if nobody uses it... we will use it here, eventually)
- Playback rate adjustment
- Immediate updates from the config window (where possible)

Here's a list of features in the current in_vgm that I'm not sure if I'll bother with:

- "Now Playing" HTML display of track info. Current versions of Winamp don't support this, but they do support Unicode metadata and modern operating systems are quite capable of displaying it.
- Per-channel panning and panning randomisation. Did anyone ever use this? I find it makes the music less "flat" but the randomness screws up a lot of the time, and setting it manually is just a lot of hard work.
- Selectable YM2612 core
- Ability to play files directly from URLs (by downloading them to a temp file)

And here's a list of new super-duper features that this new reworking ought to make possible:

- High sampling rate output
- Original-rate sound chip emulation followed by high-quality resampling
- The ability to actually maintain the code without causing nightmares
- Those ports I mentioned before

So... what do you think? What suggestions do you have for how in_vgm 2 can be even better than the old in_vgm?

Betas now appearing here:

http://www.smspower.org/maxim/forumstuff/in_vgm_latest/

...as mentioned above, a lot of features are missing.

Great! Just, don't over engineer ( reminder ). Do you actually need feedback from this early beta now?
A bar I know in Kyoto tends to play various chiptunes including VGM.

Had a small issue with MSVCR71D.DLL
It goes in Winamp folder (not in Plugins).
The old version wasn't reliant on this. :(

Bock wrote

Great! Just, don't over engineer ( reminder ).

That article is exactly what stopped me dong it, but all attempts to refactor were ultimately making it worse. I'm copying/pasting/tidying as I go rather than ignoring the old code.

Bock wrote

Do you actually need feedback from this early beta now?

To be honest, I need encouragement and fun ideas to work on as a break from the more boring parts.

DMEnduro wrote

Had a small issue with MSVCR71D.DLL
It goes in Winamp folder (not in Plugins).
The old version wasn't reliant on this. :(

The new one won't be. The current beta is a debug build, hence it needs the *D.DLL runtime. The release build was taking about 10 minutes and after 4 or so last-minute bugs (hence >40 minutes to fix them) I couldn't be bothered to wait.

Seeking is nice!

Is the Genesis Plus GX core based on Gens? If not, it would be nice to have selectable cores.

Landale wrote

Is the Genesis Plus GX core based on Gens? If not, it would be nice to have selectable cores.

Gens is one of the most inaccurate cores by now.

No, it's not based on Gens.

http://code.google.com/p/genplus-gx/

The new sound engine is apparently borrowed from Mame. Good thing :)

Tom wrote

Gens is one of the most inaccurate cores by now.

The biggest diffrence between Kega, Gens, Mame and Regen cores is the volume balance between PSG, FM and DAC etc. That's the most importent aspect when we talk about accuracy.. I wish Steve Snake could release his Kega source code..

A DirectShow filter (which I guess you alluded to by mentioning WMP) would be incredibly useful as that would let any sane media player handle VGMs under Windows.

I uploaded a new version that includes seeking and looping. This time it's a release build with no runtime DLL needed.

http://www.smspower.org/maxim/forumstuff/in_vgm_latest/

Maxim wrote

I uploaded a new version that includes seeking and looping. This time it's a release build with no runtime DLL needed.

Doesn't work with XMPlay at all.
Loading any VGM file results in a hang, going to the plugin panel closes the program silently.
Maybe you made an unicode only build? Those will work only with Winamp 5, Winamp 2 (and XMPlay/Foobar with their Winamp plugin loaders) is no go.

How odd. It's not Unicode. It's XMPlay crashing the plugin, so I can debug it.

Does not work with Winamp 5 neither here, I tried under XP (application is immediately closed) and Vista (crash messagebox with debugging stuff)

in_vgm.dll is clearly the problem, removing it from /plugins makes winamp loading back again

never had this problem with older versions

Well, it's a beta :( but I am running it in Winamp 5... do you have any weird setup? Old version of Winamp? Other plugins installed?

Maxim wrote

How odd. It's not Unicode. It's XMPlay crashing the plugin, so I can debug it.

Do you mean "can't debug" perhaps?
If so, I'll wait for a release version before reporting to the XMPlay's creator. Hopefully it's just some left over debug code acting up or some other stuff like that.

I mean, I can run it in debug mode under XMPlay and see what happens.

Maxim wrote

I mean, I can run it in debug mode under XMPlay and see what happens.

Goodie, sorry for misunderstanding. :)
Tried it with the newest daily XMPlay build and it behaves exactly the same.

The plugin must be doing something wrong to be failing like this. It ought to be simple to figure out what and make it better again.

Note that I'm releasing frequent builds to encourage myself with the feedback. They will be tend to be quite buggy.

Edit: Yes, it was a bug in the code that determines the place to load/save the settings. If you haven't enabled saving them in your user profile, it goes looking for the Winamp directory and the bug was there; I had enabled it.

ok, I got it reworking (at least, not crashin... my vgz files seems to be unsupported) by loading a previous version first

to get back on topic, nice features in my mind would be, regarding genesis playback:

- real-time mixer table or highly configurable filters so I could try reproduce the exact sound of my old console (this could also be useful to find filter configuration for different console versions)

- YM2612 HQ resampling (emulation at the original frequency)

hey dude. whenever you can, send over any source changes (if at all) involving either the vgm format itself or (if possible) the modified ym2612 code so i can attempt to update vgmtool with it. if the ym2612 is just the new fm.c from genplus linked earlier, i can get that myself :)

also, it is still my hope that i can change vgmtool to the point where the only difference btwn win and mac versions is the gui code...

There's no format changes. I'm using fm.c from Genesis Plus GX with modifications to move all the state variables into the YM2612 struct, and then to have the struct be passed in/out of functions rather than have a global; this should allow multiple cores to run simultaneously (untested) to allow things like multi-threaded decoding (eg. for transcoding while listening).

I posted today's effort:

http://www.smspower.org/maxim/forumstuff/in_vgm_latest/

- Fixed the crashing bug if you don't have a recent Winamp with profile saving enabled
- Started work on the config dialog, lots of framework stuff that ought to make maintenance 1000x easier.

Maxim wrote

I posted today's effort:

- Fixed the crashing bug if you don't have a recent Winamp with profile saving enabled
- Started work on the config dialog, lots of framework stuff that ought to make maintenance 1000x easier.

Works as a charm now, thanks.

08 Episode 1, Page 2-2.vgz from P2612 Comix Zone set (good test case since it has that GEMS instrument playing right at the beginning of the song) is still played wrong. M1 player plays the file a bit better (still not 100% like Kega, but definitely much less suckage). Since M1 also uses same fm.c code IIRC, maybe you don't render stuff in native OPN frequency yet (55466 or 55467 depending on how you round things up).

That's right, it's 44100Hz for now until I get around to adding in SRC.

Maxim wrote

That's right, it's 44100Hz for now until I get around to adding in SRC.

those music (those which use the "bug")relies on very-high frequencies and will require a good low-pass filter as well to sound completely "like the real hardware"

Kega has indeed the more realistic filtering I've ever heard in any emulator...

Hmm, so I may need to do some filtering too? :(

How come with the latest beta it doesn't show the track lengths?

Track lengths don't show up until you play the track because the metadata functions aren't in yet. If you set it to loop forever it doesn't show lengths either - they're infinite.

Not much new today - I added UI for the rate adjustment, which was already in there anyway (but backwards). I spent far too much time trying (and failing) to make Ctrl+Tab work in the config window, but did fix a few things to make it keyboard-navigable again.

The default playback rate is set at 66Hz, although a strange choice I'm not questioning why you've done that (testing I guess). I saw that, checked through the tabs, then hit cancel. Playing a track, it was still playing at the normal rate. went back in and changed loop times to 2 and then hit ok and the rate change kicked in only then.

Track lengths are showing up, and rate and loop affect them correctly it seems.

I'm not sure why it'd turn up as 66Hz. It does import the settings from the old plugin, are you sure you didn't change it in there? It defaults to 60.

Anyway, I got some time to work on it today; it now has GD3 support (mostly, helps to have a newer Winamp because the older metadata APIs aren't hooked up yet) and fixed a bug that only appeared in the optimised build, which is why I've been releasing unoptimised builds. I added a few of the metadata options too.

http://www.smspower.org/maxim/forumstuff/in_vgm_latest/

I suppose it may have been set in the old version.
It's just happened again, when updating to the current version. Dropped the dll in the plugins folder, started Winamp and hit play. Normal 60Hz playback even though it was left at 66 from before. It may be loading the setting, but then ignoring it until it is set from the dialog?
EDIT: This is happening EVERY time Winamp is started. The playback setting is forgotten, even though it is set (and stays) in the dialog.

OK, I fixed this bug.

I'll try to release an update tonight. I've added support for any output sampling rate - so you can play back at 96kHz or 192kHz or 8kHz or whatever you like, so long as your sound card supports it. I'm halfway through adding "high quality" mode - where each chip is run at its own internal output rate, and then resampled to the output rate using a high-quality conversion library - but it's somewhat broken right now. It also requires me to include a 1MB DLL full of lookup tables, but hopefully the quality will be worth it.

Maxim,

if you are going to use libsamplerate (Secret Rabbit Code) as I did, you could eventually get ride of the SINC_MEDIUM and SINC_BEST methods to save some memory space (their lookup tables are indeed huge)...

SINC_FAST gave me very good result, at least, far better that simple linear interpolation and I could not hear any differences with the two above ones.

Also, if you have time, it might be interesting to provide an additional samplerate conversion method using Blargg's blip buffer library, just to see what's best ;-)

good luck

Indeed, I just did that; it's only 86KB now. It's a shame to have to ship a DLL, but the latest SRC doesn't build in MSVC any more due to some floating point conversion tweaks that make use of C99 functions, so I have to compile a DLL in MinGW.

I notice you use src_simple; the docs suggest that's a bad idea quality-wise (no continuity of conversion across invocations). I also see you use a trivial float<->int conversion, that's almost certainly wasting a lot of CPU (depending on how efficient that is on the Broadway) - see http://www.mega-nerd.com/FPcast/ and also http://assemblyrequired.crashworks.org/2009/01/12/why-you-should-never-cast-floa... which is more console-specific.

Audiophiles tend to give you a hard time about everything, I'm sure they'd say you need the extra dynamic range noise floor bandwidth signal to noise ratio that the better filters produce, but they do seem to need huge amounts of CPU.

Does blip_buffer do rate conversion? Or do you mean feeding it the wave as a "blip" sequence? I'm generally not a fan of his BLS approach...

yes my implementation is indeed very limited as I didn't really put much time in it, just a few to find the best quality/speed ratio setting to use and setup the interface to see how that was working

thanks for pointing this out anyway, I'll probably try to improve that part when I have time

about blip bufffer, I was just told recently that it could be used in emulators to emulate chips at their original frequence so yes, I assume it does some kind of samplerate conversion

I'm far from being an audio expert though :/

Well, Blip_buffer has Sms_Snd_Emu next to it.

http://www.smspower.org/maxim/forumstuff/in_vgm_latest/

Sample rate conversion is proving to be very tricky to get right. So here's a build without it, with configurable output sampling rate (so you can do 96kHz etc) and the rate adjustment bug fixed.

Hey Max; longtime fan of your IN_VGM plugin, first time poster to SMS Power.

It's about your WIP builds, namely the 20090127 build. I installed it, and when my random playlist got to a vgm, Winamp sucked one of my Pentium D cores dry! And that's just trying to initialize playback, namely the oscilloscope only showed a sinewave while this was happening!

This happened no matter what value I set the Playback Rate; yes, even the original 44.1KHz! I hadda revert to the 20090125 build cause of this. Sorry.

OS: Windows XP SP2 (32-bit)
Soundcard: Onboard Realtek HD Audio
Player: Winamp v5.541
Output Method: Nullsoft DirectSound Output v2.47 (d)

A particular VGM or any one?

Any one, sadly. All it does is hang, not bothering to even emit the first note. x.x

oop, completely forgot; all my VGM's are MD/Genesis VGM's. Yeah, I know; shoulda said in my first post! 8D

JFD62780 wrote

It's about your WIP builds, namely the 20090127 build. I installed it, and when my random playlist got to a vgm, Winamp sucked one of my Pentium D cores dry! And that's just trying to initialize playback, namely the oscilloscope only showed a sinewave while this was happening!

This happened no matter what value I set the Playback Rate; yes, even the original 44.1KHz! I hadda revert to the 20090125 build cause of this. Sorry.

I can confirm this happening with XMPlay as well. Again, 20090125 version plays fine.

Then it's probably a config thing, like the last time it failed in xmplay. I ought to test in there more.

Hmm ok, more probs with the Jan 27th build, im using v5.541, and with the plugin installed, the software crashes on open, it seems to be associated with the file ntdll.dll, is this a file that could be needed to make the plugin function correctly?

When I revert to v0.36 beta (31st Jan '08) everything works fine :)

You'd have a hard time booting your Windows box if you hadn't the ntdll library installed... :)

For info, I don't have any of those problems on my WinXP SP3 + P4 HT.

The main vgm7z idea is really good, but I think the system should be changed.

If we open a .vgm7z file in Winamp, it should show the root entry, eg. Sonic.vgm7z and under that entry it should expand all files, which are part of that archive.

The files should be automatically decompressed to a folder (for example the User Temp dir) and after playing these files in_vgm should automatically delete extracted files & folder.

Take a look at in_zip or SNESAmp.

SNESAmp has a nice way to play compressed files too, RSN files (Rar compressed .spc files) can be directly played and handled in Winamp.

The last time I checked the vgm7z feature, it simply had the ability to decompress the files to a folder when I open a vgm7z file in Winamp.

After playing, the extracted files were still on my drive and because of this, the main feature made not much sense. Why should I use .vgm7z files, when the plugin always creates a folder with decompressed files and I manually have to delete the files?

Otherwise good job with your plugin I'm using it for an age now. :)

finally got back to my windows box; DLed the 20090125 beta cuz of the complaints abt the 0127 version, and the preferences dialog only shows content in the "playback" and "tags" tabs, the other tabs contain empty fieldsets.

winamp pro 5.55 build 2353

using xp sp3 + hacked uxtheme and a leopard msstyles, if that makes any difference. screenshots attached.

(edit: BTW, just tried the 0127 on winamp 5.55 and get the same no-play error as mentioned before me)

invgm_tab_notworking_1.PNG (17.33 KB)

ym2413 tab not working

invgm_tab_notworking_2.PNG (17.35 KB)

ym2612 tab not working

invgm_tab_working_1.PNG (19.04 KB)

playback tab working

invgm_tab_working_2.PNG (21.39 KB)

tags tab working

double post for the following reason:

in_vgm 20090125 does NOT play vgms converted from cyms. the cyms play fine using ymamp 2.5 (the last version updated by DrO for the 5.5 series of winamp) and the conversions USED to work w/in_vgm 0.36. what gives? attached is a zip of two cyms dumped using kawaks - "stage clear" from knights of the round, and "raid" from strider - and the vgms converted using my not yet public build of vgmtool 2r6 (aka vgmtool 3 alpha; maxim has this version, but i don't think i sent the code updates yet).

i am, of course, only using 20090125 cuz 0127 doesn't play at all. also, neither 0125 nor 0127 allow viewing file info for ANY vgm; i forget if this was mentioned before and if it was intentional, but i do miss it for some reason. album art doesn't show either (the folder.png/folder.jpg rename trick in the vgm's folder apparently doesn't work anymore).

cyms+vgms.zip (20.02 KB)

two cyms and their conversion results

Can confirm that 0127 wont work at all.

Any VGM I load will get as far as setting the song title but then sits at the seeking position infinitly and causes Winamp to no longer respond to messages.

0125 will play VGMs, but given I'm not a heavy user of VGM playing I'm not sure what if anything unusual i could be looking out for.

EDIT:
Running Vista Ultimate 64bit if thats relevant.

@ Neologix

Maxim wrote

- Started work on the config dialog, lots of framework stuff that ought to make maintenance 1000x easier.

Author	Message
Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	in_vgm 2 - suggestions please Posted: Sun Jan 18, 2009 4:49 pm Last edited by Maxim on Sun Jan 18, 2009 6:15 pm; edited 1 time in total
	I'm part-way through a reworking of in_vgm. Skip past the first paragraph if you don't care why. in_vgm was the first thing I ever wrote in C. It shows. It's a big spaghetti mess of every coding style, naming convention, beginner error and confusion you can think of, the most pressing of which being, almost everything is in a single monster file. So I've rebuilt it from the ground up based on (hopefully) cleaner abstractions and interfaces, although it's still pure C for maximum portability, with OO-ish abstractions to make it possible to do things like decode 2 files at the same time. In theory, this means it ought to be possible to support FB2K, WMP, XMMS, etc with minimal work, just plumbing the APIs into each other. I make no guarantees of this; Winamp is still my primary target. As it is at the moment, it has just become able to play Master System, Game Gear and Mega Drive VGMs. Almost every other feature is not done yet. So here's what's done: - Plugged in my SN76489 core to support SMS, GG, Coleco, BBC etc - Reworked YM2612 core from Genesis Plus GX to be re-entrant (untested), and plugged it in, to support Mega Drive/Genesis - Plugged in EMU2413 core - Implemented YM2612 sample optimisation support Here's a list of features in the current in_vgm that aren't done, but definitely will be (eventually): - Seeking - Looping support (including Loop Forever) - Per-channel muting - YM2151 support - Metadata support - including titles, lengths, Media Library support and a bunch of "guessing" stuff I implemented in the current beta to get things like track numbers and album artists - Transcoding/CD burning/Replay Gain calculation support - Fadeout on end - Replay Gain - File Information window - Configuration window - VGM7z support (even if nobody uses it... we will use it here, eventually) - Playback rate adjustment - Immediate updates from the config window (where possible) Here's a list of features in the current in_vgm that I'm not sure if I'll bother with: - "Now Playing" HTML display of track info. Current versions of Winamp don't support this, but they do support Unicode metadata and modern operating systems are quite capable of displaying it. - Per-channel panning and panning randomisation. Did anyone ever use this? I find it makes the music less "flat" but the randomness screws up a lot of the time, and setting it manually is just a lot of hard work. - Selectable YM2612 core - Ability to play files directly from URLs (by downloading them to a temp file) And here's a list of new super-duper features that this new reworking ought to make possible: - High sampling rate output - Original-rate sound chip emulation followed by high-quality resampling - The ability to actually maintain the code without causing nightmares - Those ports I mentioned before So... what do you think? What suggestions do you have for how in_vgm 2 can be even better than the old in_vgm?

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Sun Jan 18, 2009 6:14 pm
	Betas now appearing here: http://www.smspower.org/maxim/forumstuff/in_vgm_latest/ ...as mentioned above, a lot of features are missing.

Bock Site Admin Joined: 08 Jul 2001 Posts: 8652 Location: Paris, France	Posted: Mon Jan 19, 2009 3:41 am
	Great! Just, don't over engineer ( reminder ). Do you actually need feedback from this early beta now? A bar I know in Kyoto tends to play various chiptunes including VGM.

DMEnduro Joined: 27 Apr 2005 Posts: 420 Location: Australia	Posted: Mon Jan 19, 2009 4:08 am
	Had a small issue with MSVCR71D.DLL It goes in Winamp folder (not in Plugins). The old version wasn't reliant on this. :(

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Mon Jan 19, 2009 9:31 am
	Bock wrote Great! Just, don't over engineer ( reminder ). That article is exactly what stopped me dong it, but all attempts to refactor were ultimately making it worse. I'm copying/pasting/tidying as I go rather than ignoring the old code. Bock wrote Do you actually need feedback from this early beta now? To be honest, I need encouragement and fun ideas to work on as a break from the more boring parts. DMEnduro wrote Had a small issue with MSVCR71D.DLL It goes in Winamp folder (not in Plugins). The old version wasn't reliant on this. :( The new one won't be. The current beta is a debug build, hence it needs the *D.DLL runtime. The release build was taking about 10 minutes and after 4 or so last-minute bugs (hence >40 minutes to fix them) I couldn't be bothered to wait.

Niloct Joined: 25 Dec 2005 Posts: 607 Location: São Paulo - Brazil	Posted: Mon Jan 19, 2009 5:02 pm
	Seeking is nice!

Landale Joined: 22 Aug 2005 Posts: 45 Location: Sweden	Posted: Mon Jan 19, 2009 5:31 pm
Landale Joined: 22 Aug 2005 Posts: 45 Location: Sweden	Is the Genesis Plus GX core based on Gens? If not, it would be nice to have selectable cores.

Tom Joined: 16 May 2002 Posts: 1356 Location: italy	Posted: Mon Jan 19, 2009 5:44 pm
Tom Joined: 16 May 2002 Posts: 1356 Location: italy	Landale wrote Is the Genesis Plus GX core based on Gens? If not, it would be nice to have selectable cores. Gens is one of the most inaccurate cores by now.

Kaminari Joined: 25 Jun 2005 Posts: 50 Location: Paris, France	Posted: Mon Jan 19, 2009 6:44 pm
	No, it's not based on Gens. http://code.google.com/p/genplus-gx/ The new sound engine is apparently borrowed from Mame. Good thing :)

Landale Joined: 22 Aug 2005 Posts: 45 Location: Sweden	Posted: Mon Jan 19, 2009 7:19 pm
Landale Joined: 22 Aug 2005 Posts: 45 Location: Sweden	Tom wrote Gens is one of the most inaccurate cores by now. The biggest diffrence between Kega, Gens, Mame and Regen cores is the volume balance between PSG, FM and DAC etc. That's the most importent aspect when we talk about accuracy.. I wish Steve Snake could release his Kega source code..

benryves Joined: 12 Apr 2005 Posts: 391 Location: London, United Kingdom	Posted: Mon Jan 19, 2009 7:25 pm
	A DirectShow filter (which I guess you alluded to by mentioning WMP) would be incredibly useful as that would let any sane media player handle VGMs under Windows.

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Mon Jan 19, 2009 8:31 pm
	I uploaded a new version that includes seeking and looping. This time it's a release build with no runtime DLL needed. http://www.smspower.org/maxim/forumstuff/in_vgm_latest/

Knurek Guest	Posted: Mon Jan 19, 2009 10:02 pm
Knurek Guest	Maxim wrote I uploaded a new version that includes seeking and looping. This time it's a release build with no runtime DLL needed. Doesn't work with XMPlay at all. Loading any VGM file results in a hang, going to the plugin panel closes the program silently. Maybe you made an unicode only build? Those will work only with Winamp 5, Winamp 2 (and XMPlay/Foobar with their Winamp plugin loaders) is no go.

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Tue Jan 20, 2009 9:15 am
	How odd. It's not Unicode. It's XMPlay crashing the plugin, so I can debug it.

Christophe Joined: 23 Feb 2007 Posts: 18	Posted: Tue Jan 20, 2009 2:24 pm
Christophe Joined: 23 Feb 2007 Posts: 18	Does not work with Winamp 5 neither here, I tried under XP (application is immediately closed) and Vista (crash messagebox with debugging stuff) in_vgm.dll is clearly the problem, removing it from /plugins makes winamp loading back again never had this problem with older versions

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Tue Jan 20, 2009 3:06 pm
	Well, it's a beta :( but I am running it in Winamp 5... do you have any weird setup? Old version of Winamp? Other plugins installed?

Knurek Guest	Posted: Tue Jan 20, 2009 3:36 pm
Knurek Guest	Maxim wrote How odd. It's not Unicode. It's XMPlay crashing the plugin, so I can debug it. Do you mean "can't debug" perhaps? If so, I'll wait for a release version before reporting to the XMPlay's creator. Hopefully it's just some left over debug code acting up or some other stuff like that.

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Tue Jan 20, 2009 4:07 pm
	I mean, I can run it in debug mode under XMPlay and see what happens.

Knurek Guest	Posted: Tue Jan 20, 2009 4:54 pm
Knurek Guest	Maxim wrote I mean, I can run it in debug mode under XMPlay and see what happens. Goodie, sorry for misunderstanding. :) Tried it with the newest daily XMPlay build and it behaves exactly the same.

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Tue Jan 20, 2009 5:05 pm
	The plugin must be doing something wrong to be failing like this. It ought to be simple to figure out what and make it better again. Note that I'm releasing frequent builds to encourage myself with the feedback. They will be tend to be quite buggy. Edit: Yes, it was a bug in the code that determines the place to load/save the settings. If you haven't enabled saving them in your user profile, it goes looking for the Winamp directory and the bug was there; I had enabled it.

Christophe Joined: 23 Feb 2007 Posts: 18	Posted: Tue Jan 20, 2009 7:30 pm
Christophe Joined: 23 Feb 2007 Posts: 18	ok, I got it reworking (at least, not crashin... my vgz files seems to be unsupported) by loading a previous version first to get back on topic, nice features in my mind would be, regarding genesis playback: - real-time mixer table or highly configurable filters so I could try reproduce the exact sound of my old console (this could also be useful to find filter configuration for different console versions) - YM2612 HQ resampling (emulation at the original frequency)

neologix Joined: 13 Nov 2007 Posts: 93 Location: New York, NY, USA	Posted: Tue Jan 20, 2009 7:32 pm
	hey dude. whenever you can, send over any source changes (if at all) involving either the vgm format itself or (if possible) the modified ym2612 code so i can attempt to update vgmtool with it. if the ym2612 is just the new fm.c from genplus linked earlier, i can get that myself :) also, it is still my hope that i can change vgmtool to the point where the only difference btwn win and mac versions is the gui code...

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Tue Jan 20, 2009 9:11 pm
	There's no format changes. I'm using fm.c from Genesis Plus GX with modifications to move all the state variables into the YM2612 struct, and then to have the struct be passed in/out of functions rather than have a global; this should allow multiple cores to run simultaneously (untested) to allow things like multi-threaded decoding (eg. for transcoding while listening). I posted today's effort: http://www.smspower.org/maxim/forumstuff/in_vgm_latest/ - Fixed the crashing bug if you don't have a recent Winamp with profile saving enabled - Started work on the config dialog, lots of framework stuff that ought to make maintenance 1000x easier.

Knurek Guest	Posted: Tue Jan 20, 2009 10:34 pm
Knurek Guest	Maxim wrote I posted today's effort: - Fixed the crashing bug if you don't have a recent Winamp with profile saving enabled - Started work on the config dialog, lots of framework stuff that ought to make maintenance 1000x easier. Works as a charm now, thanks. 08 Episode 1, Page 2-2.vgz from P2612 Comix Zone set (good test case since it has that GEMS instrument playing right at the beginning of the song) is still played wrong. M1 player plays the file a bit better (still not 100% like Kega, but definitely much less suckage). Since M1 also uses same fm.c code IIRC, maybe you don't render stuff in native OPN frequency yet (55466 or 55467 depending on how you round things up).

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Tue Jan 20, 2009 10:55 pm
	That's right, it's 44100Hz for now until I get around to adding in SRC.

Eke Guest	Posted: Wed Jan 21, 2009 8:46 am
Eke Guest	Maxim wrote That's right, it's 44100Hz for now until I get around to adding in SRC. those music (those which use the "bug")relies on very-high frequencies and will require a good low-pass filter as well to sound completely "like the real hardware" Kega has indeed the more realistic filtering I've ever heard in any emulator...

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Wed Jan 21, 2009 9:58 am
	Hmm, so I may need to do some filtering too? :(

Jacko Joined: 01 Feb 2004 Posts: 1464 Location: Sunny ol Tamworth, New England NSW AU	Posted: Wed Jan 21, 2009 2:11 pm
	How come with the latest beta it doesn't show the track lengths?

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Wed Jan 21, 2009 9:13 pm
	Track lengths don't show up until you play the track because the metadata functions aren't in yet. If you set it to loop forever it doesn't show lengths either - they're infinite. Not much new today - I added UI for the rate adjustment, which was already in there anyway (but backwards). I spent far too much time trying (and failing) to make Ctrl+Tab work in the config window, but did fix a few things to make it keyboard-navigable again.

DMEnduro Joined: 27 Apr 2005 Posts: 420 Location: Australia	Posted: Thu Jan 22, 2009 1:01 am
	The default playback rate is set at 66Hz, although a strange choice I'm not questioning why you've done that (testing I guess). I saw that, checked through the tabs, then hit cancel. Playing a track, it was still playing at the normal rate. went back in and changed loop times to 2 and then hit ok and the rate change kicked in only then. Track lengths are showing up, and rate and loop affect them correctly it seems.

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Sun Jan 25, 2009 3:39 pm
	I'm not sure why it'd turn up as 66Hz. It does import the settings from the old plugin, are you sure you didn't change it in there? It defaults to 60. Anyway, I got some time to work on it today; it now has GD3 support (mostly, helps to have a newer Winamp because the older metadata APIs aren't hooked up yet) and fixed a bug that only appeared in the optimised build, which is why I've been releasing unoptimised builds. I added a few of the metadata options too. http://www.smspower.org/maxim/forumstuff/in_vgm_latest/

DMEnduro Joined: 27 Apr 2005 Posts: 420 Location: Australia	Posted: Tue Jan 27, 2009 12:32 am
	I suppose it may have been set in the old version. It's just happened again, when updating to the current version. Dropped the dll in the plugins folder, started Winamp and hit play. Normal 60Hz playback even though it was left at 66 from before. It may be loading the setting, but then ignoring it until it is set from the dialog? EDIT: This is happening EVERY time Winamp is started. The playback setting is forgotten, even though it is set (and stays) in the dialog.

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Tue Jan 27, 2009 9:06 am
	OK, I fixed this bug. I'll try to release an update tonight. I've added support for any output sampling rate - so you can play back at 96kHz or 192kHz or 8kHz or whatever you like, so long as your sound card supports it. I'm halfway through adding "high quality" mode - where each chip is run at its own internal output rate, and then resampled to the output rate using a high-quality conversion library - but it's somewhat broken right now. It also requires me to include a 1MB DLL full of lookup tables, but hopefully the quality will be worth it.

Eke Guest	Posted: Tue Jan 27, 2009 3:06 pm
Eke Guest	Maxim, if you are going to use libsamplerate (Secret Rabbit Code) as I did, you could eventually get ride of the SINC_MEDIUM and SINC_BEST methods to save some memory space (their lookup tables are indeed huge)... SINC_FAST gave me very good result, at least, far better that simple linear interpolation and I could not hear any differences with the two above ones. Also, if you have time, it might be interesting to provide an additional samplerate conversion method using Blargg's blip buffer library, just to see what's best ;-) good luck

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Tue Jan 27, 2009 3:43 pm
	Indeed, I just did that; it's only 86KB now. It's a shame to have to ship a DLL, but the latest SRC doesn't build in MSVC any more due to some floating point conversion tweaks that make use of C99 functions, so I have to compile a DLL in MinGW. I notice you use src_simple; the docs suggest that's a bad idea quality-wise (no continuity of conversion across invocations). I also see you use a trivial float<->int conversion, that's almost certainly wasting a lot of CPU (depending on how efficient that is on the Broadway) - see http://www.mega-nerd.com/FPcast/ and also http://assemblyrequired.crashworks.org/2009/01/12/why-you-should-never-cast-floa... which is more console-specific. Audiophiles tend to give you a hard time about everything, I'm sure they'd say you need the extra dynamic range noise floor bandwidth signal to noise ratio that the better filters produce, but they do seem to need huge amounts of CPU. Does blip_buffer do rate conversion? Or do you mean feeding it the wave as a "blip" sequence? I'm generally not a fan of his BLS approach...

Eke Guest	Posted: Tue Jan 27, 2009 5:32 pm
Eke Guest	yes my implementation is indeed very limited as I didn't really put much time in it, just a few to find the best quality/speed ratio setting to use and setup the interface to see how that was working thanks for pointing this out anyway, I'll probably try to improve that part when I have time about blip bufffer, I was just told recently that it could be used in emulators to emulate chips at their original frequence so yes, I assume it does some kind of samplerate conversion I'm far from being an audio expert though :/

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Tue Jan 27, 2009 5:53 pm
	Well, Blip_buffer has Sms_Snd_Emu next to it.

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Tue Jan 27, 2009 8:30 pm
	http://www.smspower.org/maxim/forumstuff/in_vgm_latest/ Sample rate conversion is proving to be very tricky to get right. So here's a build without it, with configurable output sampling rate (so you can do 96kHz etc) and the rate adjustment bug fixed.

JFD62780 Joined: 27 Jan 2009 Posts: 13 Location: Tampa, FL, USA	Posted: Tue Jan 27, 2009 10:21 pm
	Hey Max; longtime fan of your IN_VGM plugin, first time poster to SMS Power. It's about your WIP builds, namely the 20090127 build. I installed it, and when my random playlist got to a vgm, Winamp sucked one of my Pentium D cores dry! And that's just trying to initialize playback, namely the oscilloscope only showed a sinewave while this was happening! This happened no matter what value I set the Playback Rate; yes, even the original 44.1KHz! I hadda revert to the 20090125 build cause of this. Sorry. OS: Windows XP SP2 (32-bit) Soundcard: Onboard Realtek HD Audio Player: Winamp v5.541 Output Method: Nullsoft DirectSound Output v2.47 (d)

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Tue Jan 27, 2009 11:52 pm
	A particular VGM or any one?

JFD62780 Joined: 27 Jan 2009 Posts: 13 Location: Tampa, FL, USA	Posted: Wed Jan 28, 2009 12:07 am
	Any one, sadly. All it does is hang, not bothering to even emit the first note. x.x oop, completely forgot; all my VGM's are MD/Genesis VGM's. Yeah, I know; shoulda said in my first post! 8D

Knurek Guest	Posted: Wed Jan 28, 2009 11:47 pm
Knurek Guest	JFD62780 wrote It's about your WIP builds, namely the 20090127 build. I installed it, and when my random playlist got to a vgm, Winamp sucked one of my Pentium D cores dry! And that's just trying to initialize playback, namely the oscilloscope only showed a sinewave while this was happening! This happened no matter what value I set the Playback Rate; yes, even the original 44.1KHz! I hadda revert to the 20090125 build cause of this. Sorry. I can confirm this happening with XMPlay as well. Again, 20090125 version plays fine.

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14745 Location: London	Posted: Thu Jan 29, 2009 9:52 am
	Then it's probably a config thing, like the last time it failed in xmplay. I ought to test in there more.

Jacko Joined: 01 Feb 2004 Posts: 1464 Location: Sunny ol Tamworth, New England NSW AU	Posted: Fri Jan 30, 2009 3:09 pm
	Hmm ok, more probs with the Jan 27th build, im using v5.541, and with the plugin installed, the software crashes on open, it seems to be associated with the file ntdll.dll, is this a file that could be needed to make the plugin function correctly? When I revert to v0.36 beta (31st Jan '08) everything works fine :)

Kaminari Joined: 25 Jun 2005 Posts: 50 Location: Paris, France	Posted: Fri Jan 30, 2009 8:16 pm
	You'd have a hard time booting your Windows box if you hadn't the ntdll library installed... :) For info, I don't have any of those problems on my WinXP SP3 + P4 HT.

Koopa Guest	Posted: Sat Jan 31, 2009 1:04 am
Koopa Guest	The main vgm7z idea is really good, but I think the system should be changed. If we open a .vgm7z file in Winamp, it should show the root entry, eg. Sonic.vgm7z and under that entry it should expand all files, which are part of that archive. The files should be automatically decompressed to a folder (for example the User Temp dir) and after playing these files in_vgm should automatically delete extracted files & folder. Take a look at in_zip or SNESAmp. SNESAmp has a nice way to play compressed files too, RSN files (Rar compressed .spc files) can be directly played and handled in Winamp. The last time I checked the vgm7z feature, it simply had the ability to decompress the files to a folder when I open a vgm7z file in Winamp. After playing, the extracted files were still on my drive and because of this, the main feature made not much sense. Why should I use .vgm7z files, when the plugin always creates a folder with decompressed files and I manually have to delete the files? Otherwise good job with your plugin I'm using it for an age now. :)

neologix Joined: 13 Nov 2007 Posts: 93 Location: New York, NY, USA	Posted: Sun Feb 01, 2009 4:34 am
	finally got back to my windows box; DLed the 20090125 beta cuz of the complaints abt the 0127 version, and the preferences dialog only shows content in the "playback" and "tags" tabs, the other tabs contain empty fieldsets. winamp pro 5.55 build 2353 using xp sp3 + hacked uxtheme and a leopard msstyles, if that makes any difference. screenshots attached. (edit: BTW, just tried the 0127 on winamp 5.55 and get the same no-play error as mentioned before me) invgm_tab_notworking_1.PNG (17.33 KB) ym2413 tab not working invgm_tab_notworking_2.PNG (17.35 KB) ym2612 tab not working invgm_tab_working_1.PNG (19.04 KB) playback tab working invgm_tab_working_2.PNG (21.39 KB) tags tab working

neologix Joined: 13 Nov 2007 Posts: 93 Location: New York, NY, USA	Posted: Sun Feb 01, 2009 5:39 am
	double post for the following reason: in_vgm 20090125 does NOT play vgms converted from cyms. the cyms play fine using ymamp 2.5 (the last version updated by DrO for the 5.5 series of winamp) and the conversions USED to work w/in_vgm 0.36. what gives? attached is a zip of two cyms dumped using kawaks - "stage clear" from knights of the round, and "raid" from strider - and the vgms converted using my not yet public build of vgmtool 2r6 (aka vgmtool 3 alpha; maxim has this version, but i don't think i sent the code updates yet). i am, of course, only using 20090125 cuz 0127 doesn't play at all. also, neither 0125 nor 0127 allow viewing file info for ANY vgm; i forget if this was mentioned before and if it was intentional, but i do miss it for some reason. album art doesn't show either (the folder.png/folder.jpg rename trick in the vgm's folder apparently doesn't work anymore). cyms+vgms.zip (20.02 KB) two cyms and their conversion results

thatawesomeguy Joined: 25 Jul 2007 Posts: 732 Location: Melbourne, Australia	Posted: Sun Feb 01, 2009 7:14 am
	Can confirm that 0127 wont work at all. Any VGM I load will get as far as setting the song title but then sits at the seeking position infinitly and causes Winamp to no longer respond to messages. 0125 will play VGMs, but given I'm not a heavy user of VGM playing I'm not sure what if anything unusual i could be looking out for. EDIT: Running Vista Ultimate 64bit if thats relevant.

Kaminari Joined: 25 Jun 2005 Posts: 50 Location: Paris, France	Posted: Sun Feb 01, 2009 5:56 pm
	@ Neologix Maxim wrote - Started work on the config dialog, lots of framework stuff that ought to make maintenance 1000x easier.

Forums

View topic - in_vgm 2 - suggestions please