minifmod still used a lot?
category: general [glöplog]
with a chipmusic a kkrunchied 64k intro was 4k smaller after changed mini to micro
A lot smaller
i think XM music just sounds cool and the old good sample music reminds people golden ages of demoscene. softsynths are nice but you know... the sound is often too plastic... so yeah we've been using minifmod ana mxmplay : )
ufmod !! got to try that.
For my own use, I did a port of minifmod to full CPP and big endian compatibleness. So I traced almost the whole code and I must say... some things are damn shitty in minifmod (as the docs warn): a xm with more than 32 chans * crash * in the memory. (there are 64 hardcoded channels, that are 32+32 dbl buffers for sample swicth... so there are 32 chans in fact
the ugliest things is when a loop sample volume is switch off with "C00" or A01 +A01.... the chan is always mixed forever !!! I mean it adds things multiplied by zero a lot...
completely useless operation that takes lots of CPU in some xms... a simple jump and a multipy can fix that.
Also, the effect automat is a "switch hell", and in some cases effects are of course not played as they should.
For my own use, I did a port of minifmod to full CPP and big endian compatibleness. So I traced almost the whole code and I must say... some things are damn shitty in minifmod (as the docs warn): a xm with more than 32 chans * crash * in the memory. (there are 64 hardcoded channels, that are 32+32 dbl buffers for sample swicth... so there are 32 chans in fact
the ugliest things is when a loop sample volume is switch off with "C00" or A01 +A01.... the chan is always mixed forever !!! I mean it adds things multiplied by zero a lot...
completely useless operation that takes lots of CPU in some xms... a simple jump and a multipy can fix that.
Also, the effect automat is a "switch hell", and in some cases effects are of course not played as they should.
... and for information, uncompressed executable size of the whole minifmod seems to be between 6 and 7 Kb...
yes, too much. switch to kebby's k2. or write a granular synth for ps and me ;P
Quote:
"the musicians using it" would be me and smash. funny no one's said that about V2 for instance.
On the other hand Smash and you can say nothing bad has so far been produced with the FLT synth. I can't. :)
too late, he already did...
right back atcha :)
krabob: as the volume can be turned back on, at least the position needs to be tracked. but yes, mixing it is a waste of time. then again, Satan and/or Santana loves the lazy ones.
®: Yeah, we could now start a lenghty discussion why you loathe that one particular tune of yours that _doesn't_ follow your usual "recipes of success" (as you put it on your blog thing) so much.
But do you really want that? :)
But do you really want that? :)
Quote:
For my own use, I did a port of minifmod to full CPP and big endian compatibleness.
I remember back in 2005 there was a post at wasm.ru bbs about porting minifmod to pure C in order to make it crossplatform and so. Wheren't you the guy who posted that topic? :)
Quote:
some things are damn shitty in minifmod (as the docs warn):
Yeah, there many bugs in minifmod: crashing on "bad" xm files, memory leaks, L & R channels being swapped and some more. Me and Asterix fixed all of the above in uFMOD. If you could find any bug in the following release (coming soon), please, let us know.
Quote:
so there are 32 chans in fact
That's what the standard says. I never seen an XM with more than 32 channels.
kb_: that's bull, tunes could be made good even if they don't adhere to the Recipe™. the d&t tune could be a lot better in its own field, without making it a sellout. on a side note, i'm surprised someone actually read the rant closely enough to refer to it ;)
Is the Minifmod pure C/C++ port available somewhere? :-)
Quote:
krabob: as the volume can be turned back on, at least the position needs to be tracked. but yes, mixing it is a waste of time. then again, Satan and/or Santana loves the lazy ones.
Yeah ! I was about to make the mistake, But I thought about this. In the C mixer I did, If I know the volume will be 0 for a whole part and channel, I jump over the sound buffer AND use a multiply yo "clip" the sample itself in case the volume come back !!
Quote:
That's what the standard says. I never seen an XM with more than 32 channels.
(?) Well, I'm not a fastracker2 expert enough, but I have some xm with 36 channels or more here, for a xm collection I used to validate my sources. I thought 256 was the maximum. To be exact, the minifmod mixer has a "32 channels server", and * remaps * the xm chans on the mixer chans
as samples are launch. (Wait... I got to verify that.)
... and to complete my bullshits, someone Asked me to release this full C (no asm)+bigendianness+nointtofloatcast version. So I may do that soon.
Krabob, and? :-)
Quote:
Well, I'm not a fastracker2 expert enough, but I have some xm with 36 channels or more here, for a xm collection
Probably saved with Modplug Tracker, which didn't care much about format compatibility.
anyone know how good/bad ufmod is?
It's good enough. It's based on fmod, which in turn is what xmplay uses, so...
xmplay uses bass, not fmod. check your fax.
right...
my bad
it's not really 100% compatible to any other tracker, i.e. it neither emulates all FT2 glitches, nor all <insert random tracker which plays XMs differnt from FT2> glitches. But that's mostly details and understandable, judging the size of the whole thing. But yeah, better double-check if you tunes play correctly.
We used ufmod on "Transform", it has some problems with instrument vibrato in XMs (they're not smooth like other players and sometimes don't work at all) There's also weird bugs with portamento and vibrato fx commands, possibly some more I forget. I had to make quite a few changes to the track which is why the 'Rebel without a pause' bit is so buried. Looking at the files the built-in packing came out as:
original mod - 81,978
basic pack - 81,054 (just takes off headers)
full pack - 79,820
I did some bitstreaming tests with raw track data, think you can get it down by another couple of k that way (after the demo is compressed) then you just need to reconstruct the file at runtime.
Yeah more modules in intros please, demand that 32k from your coder today. ;)
original mod - 81,978
basic pack - 81,054 (just takes off headers)
full pack - 79,820
I did some bitstreaming tests with raw track data, think you can get it down by another couple of k that way (after the demo is compressed) then you just need to reconstruct the file at runtime.
Yeah more modules in intros please, demand that 32k from your coder today. ;)