shithub: choc

Info  •  Files  •  Log  •  Branches

Fix OPL MIDI tempo calculations.

It turns out that the way that tempo has been calculated in OPL playback
has been broken for a long time. The mysterious "fudge factor" that I
had to apply to tempo calculations is actually completely unnecessary:
the byte-swapping in the MIDI_GetFileTimeDivision() function was being
done wrong, so the time division being used by the OPL MIDI code was
completely wrong. Presumably the multiply by 260 was close enough to an
8-bit bitshift that it worked okayish, but large enough time division
values would overflow a single byte and screw up.

This fixes long-running OPL playback problems in a number of WADs, most
notably Alien Vendetta. This *really* fixes #352.

--- a/src/i_oplmusic.c

+++ b/src/i_oplmusic.c

@@ -46,9 +46,6 @@

 #define PERCUSSION_LOG_LEN 16

-// TODO: Figure out why this is needed.

-#define TEMPO_FUDGE_FACTOR 260

-

 typedef struct

 {

     byte tremolo;

@@ -1198,8 +1195,7 @@

     // Get the number of microseconds until the next event.

     nticks = MIDI_GetDeltaTime(track->iter);

-    us = ((uint64_t) nticks * us_per_beat * TEMPO_FUDGE_FACTOR)

-       / ticks_per_beat;

+    us = ((uint64_t) nticks * us_per_beat) / ticks_per_beat;

     // Set a timer to be invoked when the next event is

     // ready to play.

--- a/src/midifile.c

+++ b/src/midifile.c

@@ -699,14 +699,14 @@

 unsigned int MIDI_GetFileTimeDivision(midi_file_t *file)

 {

-    short result = SHORT(file->header.time_division);

+    short result = SDL_SwapBE16(file->header.time_division);

     // Negative time division indicates SMPTE time and must be handled

     // differently.

     if (result < 0)

     {

-        // TODO: Figure this out.

-        return 96;

+        return (signed int)(-(result/256))

+             * (signed int)(result & 0xFF);

     }

     else

     {