This is an enhanced version of the SNESGSS music engine for SNES homebrew games. Because the original tracker's exporter is hardcoded for the orginal SNESGSS driver, I made a new one in Python 3 that's based on the original export code.
New features include
- Local samples - have per-song samples that get loaded in alongside the songs that use them
- Commands to enable and configure echo
- Ability to play sound effects on music channels
- Take multiple SNESGSS module files as input and combine them
- Acknowledge SPC driver commands in a safer way
The faster SPC700 loader from libSFX is included. The music compression could also use some improvements - it seems to be particularly slow in Python, and doesn't try to compress as aggressively as it could have in order to try taking less time.
The driver has been converted to use a ca65 macro pack, allowing it to integrate with a ca65 program more easily, and display debugging information in Mesen-S. It's interfaced with similarly to the original driver, but the command IDs have been changed to make room for more commands.
The way you send commands has also been changed - you're supposed to read APU3 ($2143), wait for APU0 ($2140) to be nonzero, write the new command to APU0 (as well as any parameters to APU1-3), and then wait until APU3 has a different value than it had before.
The exportgss.py script handles taking SNESGSS modules and turning it into data the driver can take. Syntax is exportgss.py data.s enum.s List.gss Of.gss Songs.gss
The exporter outputs a data file (ca65 format) as well as a file containing a Music enum, and a SoundEffect enum, giving you an easy way to get the ID of the song or sound effect you want to play.
Local samples are specified by prefixing the instrument name with L_ but you can change the prefix in exportgss.py, as well as decide whether you want the exporter to accommodate streaming or not.
BRR conversion is split into a separate program, just using the original BRR conversion code from SNESGSS. It could be converted into Python later if desired, but it was a lot easier to just leave as-is than everything else.
It should be easy to build with any C compiler - gcc gssbrr.c -o gssbrr for example.
The BRR conversion tool is used by calling it with the command-line parameters length loop_start loop_end loop_enable loop_unroll souce_rate source_volume eq_low eq_mid eq_high resample_type downsample_factor ramp_enable source in order, and it outputs three text lines, consisting of the BRR size (in bytes, decimal), BRR loop point (in bytes, decimal), and the converted sample as a long hexadecimal string.
You'll want to add something like this to the MEMORY section of your linker configuration:
SPCZEROPAGE: start = $0004, size = $00EC;
SPCRAM: start = $0200, size = $FDC0;
And something like this to your SEGMENTS section (replacing ROM15 with wherever you want to store the data):
SPCIMAGE: load = ROM15, run=SPCRAM, align = $100, define=yes;
SPCZEROPAGE:load = SPCZEROPAGE, type=zp, optional=yes;
SPCBSS: load = SPCRAM, type = bss, align = $100, optional=yes;
The exporter will also use a segment named SongData and will currently put all of the music data (including local samples) in that segment.
65816-side functions are in blarggapu.s. You can call spc_boot_apu to load the driver and global music data, then you want to send the INITIALIZE command. Commands are sent by calling GSS_SendCommand with the command ID in the accumulator. For commands that take parameters, GSS_SendCommandParamX will use X, with the low byte going to APU2 and high byte going to APU3.
- APU2 = Volume
- APU3 = Change speed
- APU2 = Volume
- APU3 = Channel mask
- APU1 = Volume
- APU2 = Effect number
- APU3 = Pan
Starts a sound effect. The channel to play on is given in the upper nybble of the command ID - see the following example:
lda #255
sta APU1 ; volume
lda #SoundEffect::Jump
sta APU2 ; effect
lda #128
sta APU3 ; Pan
lda #GSS_Commands::SFX_PLAY|$80
sta APU0
When a song starts, it reserves a number of channels for music playback, starting from the first one. Sound effects cannot play on reserved channels, but you can specify "channel" 8, 9 or 10, and it will play on a virtual channel corresponding to channels 7, 6 and 5 respectively. When a sound effect is playing on a virtual channel, the music stops using the corresponding real channel and the sound effect plays instead.
- No parameters Must be called after loading the driver in - initializes variables and DSP registers.
- APU2 = Stereo, if nonzero The driver defaults to mono - change it here if needed.
- No parameters Starts playing whatever song has been loaded into audio RAM.
- No parameters Stops all music channels and marks all channels as free for sound effects.
- APU2 = Pause if nonzero
- No parameters Stops all channels, including any sound effects.
- APU2 = Number of pages (256 byte units) to send over
Starts the faster loader, with the destination hardcoded to the place music should be loaded.
GSS_LoadSongcan use this.
Jumps to the built-in SPC bootloader at $ffc9.
- APU2 = Volume
- APU3 = Channel mask
- APU2 = High byte of the start of the echo buffer
- APU3 = Echo delay (and buffer size) in 2KB chunks. 0 is invalid, and represents just 4 bytes.
- APU2 = FIR set
- APU3 = Feedback volume (the volume at which the echo-output is mixed back to the echo-input).
From FullSNES:
- Volume (-128..+127) (negative = phase inverted) (sample=sample*vol/128)
- Medium values (like 40h) produce "normal" echos (with decreasing volume on each repetition). Value 00h would produce a one-shot echo, value 7Fh would repeat the echo almost infinitely (assuming that FIRx leaves the overall volume unchanged).
To change the FIR sets available you can go to fir_table in the SPC driver. The preset options are 0=simple echo, 1=multi echo, 2=low pass echo, 3=high pass echo.