Skip to content

Latest commit

 

History

History
575 lines (473 loc) · 27.5 KB

mml-syntax.md

File metadata and controls

575 lines (473 loc) · 27.5 KB

MML Syntax

The audio-engine and MML syntax is still in development and subject to change without notice.

Header

Lines starting with # define metadata.

Song metadata:

  • #Title
  • #Game
  • #Composer
  • #Author
  • #Copyright
  • #Date
  • #License

The song tempo can be set with either #Tempo or #Timer:

  • #Tempo number - Sets the number of times a 48-tick cycle will occur in a minute.
    • With a #Zenlen value of 96 (default), specifies the number of half-notes in a minute
    • With a #Zenlen value of 192, specifies the number of quarter-notes in a minute
  • #Timer number - Sets the tick clock. Each tick is Timer * 0.125 milliseconds (or Timer / 8000Hz seconds) long.

The following adjust default values:

  • #Zenlen number - Set the default whole-note length (default 96)

Echo values:

  • #EchoLength - Echo buffer size in milliseconds.
    • This value must be a multiple of 16
    • The echo buffer requires EchoLenth / 16 * 2048 bytes of Audio-RAM.
    • The echo buffer is placed at the end of Audio-RAM.
    • An EchoLength of 0 does not disable the echo buffer. When EchoLength is 0, 256 bytes of Audio-RAM will be allocated to the echo buffer and the echo buffer is 4 bytes in size.
    • The default EchoLength value is 0.
  • #MaxEchoLength - Maximum echo buffer size in milliseconds.
    • If MaxEchoLength is not supplied, EchoLength will be used.
  • #FirFilter - FIR filter (C0 - C7 S-DSP registers)
    • The FIR filter must contain 8 space separated values.
    • Values prefixed with a $ are hexadecimal ($00-$FF)
    • Values without a prefix are decimal (-128 to 127)
    • WARNING: These values are copied to the C0 - C7 S-DSP registers without any overflow checks.
      An overflowing FIR filter can pop. Excessive filter feedback will continue to get louder and louder, exploding your song.
      Headphone users should turn down their volume when playing with echo filters/feedback.
    • The compiler will reject an FIR filter whose absolute sum is > 128 (unless this limit is disabled).
  • #DisableFirFilterLimit - Disables the FIR filter absolute sum limit.
  • #EchoFeedback - Echo feedback setting (-128 to 127) (EFB S-DSP register)
    • WARNING: This value is copied to the EFB S-DSP register without any overflow checks.
      Excessive echo feedback will continue to get louder and louder, exploding your song.
      Headphone users should turn down their volume when playing with echo filters/feedback.
  • #EchoVolume - 1 or 2 echo volume arguments (0 to 127)
    • If only 1 value is given, it will used for both channels
  • #EchoInvert - Invert echo volume
    • mono, left, right - echo flags
    • both - invert both left and right channels in all modes
    • none - disable echo invert
    • See \ei command for more details

The following will only affect exported .spc files:

  • #SpcSongLength - Override the .spc file song length (in seconds)
    • The maximum value is 999 seconds
  • #SpcFadeout - The fadeout length in milliseconds
    • The maximum value is 99999 milliseconds
    • The default fadeout is 0 (no fadeout)

Instruments

Lines starting with @ contain instrument definitions.

Instrument ids can be decimal numbers or alphanumeric.

  • @<id> <sample_name> - Create an instrument called id, that references the instrument sample_name in the samples.json file.
  • @<id> <sample_name> adsr <attack> <decay> <sustain_level> <sustain_rate> - Creates an instrument that overrides the ADSR values.
  • @<id> <sample_name> gain <gain_value> - Creates an instrument that overrides will disable ADSR and sets the GAIN S-DSP register to gain_value.

Channels

Lines starting with A - H select which track the MML will be played on.

The channels must be separated from the MML by a space or tab.

Multi-Channels

A single MML line can be played on multiple channels. Do not separate the channels with spaces. If a channel is repeated, the line will only be played once.

For example:

A   c
CDE a b

Will play c on channel A and a b on channels C, D and E.


NOTE: Playing a MML line on multiple channels duplicates the MML data for each channel.

Subroutines

Lines starting with ! are subroutines.

!<id> <MML>

Subroutines allow you to deduplicate MML code.

For example:

!1 a b c

A !1 d e !1

Would play a b c d e a b c on channel A.


The subroutine id can be a decimal or alphanumeric.

Subroutines can expand across multiple lines:

!a a

!z d e
!z f g

Subroutine !a will output a.
Subroutine !z will output d e f g


This is not a substitution macro, each subroutine/channel has the following independent settings:

  • Whole note length (zenlen)
  • Octave
  • Transpose (semitone offset)
  • Default length
  • Quantization value
  • MP Vibrato
  • MD automatic detune by cents

MML Syntax

An MML command cannot span multiple lines. Loops can span multiple lines.


id format:

  • If the id is a number, no space is required after the id.
  • If the id is an alphanumeric, a space is required after the id.

length format: [[%]duration][.]

  • The note length is whole-note/duration (if there is no %).
  • The % prefix will specify duration in ticks (clock-value).
  • Adding one or more dots . will extent the note length by half.
  • Examples:
    • a4 - Plays a for a quarter-note.
    • a8 - Plays a for an eighth-note.
    • a4. - Plays a for 1.5 × quarter-note length (a4 & a8).
    • a4.. - Plays a for 1.75 × quarter-note length (a4 & a8 & a16).
    • a%20 - Plays a for 20 ticks.
    • a%20. - Plays a for 30 ticks (20+10).

Required parameters are enclosed in <>.
Optional parameters are enclosed in [].
Numbers can be decimal, or hexadecimal when prefixed with $ (ie, $ff)

  • c d e f g a b - Play note

    • + adds a sharp.
    • - adds a flat.
    • Multiple + and - can be used to increase/decrease the pitch by a semitone.
    • A optional length can be added after the note. If there is no length, the default length (l) will be used.

  • P<pitch>[,length] - Play pitch

    • Sets the VxPITCH register to pitch and plays a note. For example, P$1000 will play the BRR sample at the native sample rate.

  • PR<sample_rate>[,length] - Play BRR sample at a given sample rate (in Hertz)

    • Sets the VxPITCH register to play the current BRR sample at the given sample rate.
    • For example: PR32000 will play the BRR sample at the native 32KHz sample rate (P$1000) and PR16000 will it at 16Khz sample rate (P$0800).

  • PF<freq>[,length] - Play instrument a given frequency (in Hertz)

    • Uses the instrument frequency to generate a play-pitch instruction
    • Cannot be used on samples

  • s[number][,length] - Play samplelf.

    • If this command is used with a sample, number is the sample-rate list index (default is 0).
    • If this command is used with an instrument, number is the note-id.

  • n<number> - Play the integer MIDI note at the default length

  • ^[length] - Tie, extends the duration of the previous note or rest

  • & - Slur. Used to tie two notes together without a key-off event

  • &<length> - Alternative Tie syntax

  • r[length] - Rest

    • r rest commands send a key-off event.
    • r rest commands can be tied, ie r2 ^8.
    • Three or more consecutive r rest commands will be merged by the MML parser. For example, r4 r8 r8 will be merged into a two r4 r4 commands.
    • Very long r rests will be optimised into a loop. For example, when Whole Note Length is 96, r1 r1 r1 r1 r1 r1 r1 r1 r1 r1 will be converted to r1 [r%216]4.

  • w[length] - Wait

    • A w wait will not send a key-off event. Useful for changing pan or volume in the middle of a slurred note.
    • w wait commands can be tied, ie w2 ^8.
    • Multiple w wait commands will be merged by the MML parser. For example, w4 w8 w8 will be merged into a single w2 command.
    • Very long w waits will be optimised into a loop. For example, when Whole Note Length is 96, w1 w1 w1 w1 w1 w1 w1 w1 w1 w1 will be converted to [w%240]4.

  • N<0-31>[,length] - play noise

    • Noise is disabled on the next note or P play-pitch command
    • CAUTION: There is only 1 noise generator.
      This command will change the noise frequency on all channels playing noise.
    • CAUTION: The instrument is used to determine the length of the noise.
      • If the instrument does not loop, the noise is played for the length of the instrument's sample.
      • If the instrument loops, the noise is played until key-off.

  • N- - disable noise


  • @<id> - Set instrument
    • The instrument should not be changed in the middle of a slur.
  • ?@<id> - Set subroutine instrument hint
    • Sets an instrument hint for {} portamento, MP and MD commands without emitting an set_instrument instruction.
    • The ?@ command can only be used in subroutines.
    • The ?@ hint will be used when previewing subroutines in the GUI.
    • CAUTION: A subroutine with a ?@ hint can only be called if the instrument source frequency matches the ?@ hint.
  • A<attack>,<decay>,<sustain_level>,<sustain_rate> - Set the channel's ADSR register
  • G - Set the channel's GAIN register (disables ADSR)
    • G<gain> - write the raw 8 bit value to the GAIN register
    • GF<value> - fixed envelope (0-127)
    • GD<rate> - linear decrease envelope (0-31)
    • GE<rate> - exponential decrease envelope (0-31)
    • GI<rate> - linear increase envelope (0-31)
    • GB<rate> - bent increase envelope (0-31)
  • GT - Temporarily sets the channel's GAIN register (the envelope will be restored on key-off)
    • GT0 - disable temporary GAIN
    • GT<gain> - write the raw 8 bit value to the GAIN register
    • GFT<value> - fixed temporary envelope (0-127)
    • GDT<rate> - linear temporary decrease envelope (0-31)
    • GET<rate> - exponential decrease temporary envelope (0-31)
    • GIT<rate> - linear increase temporary envelope (0-31)
    • GBT<rate> - bent increase temporary envelope (0-31)
    • GT - reuse previous temporary GAIN

  • o<0..7> - Set octave
  • > - Increase octave
  • < - Decrease octave
  • _<-128..+128> - Sets transpose. Changes the pitch of each note played by param semitones. The default is 0.
  • __<-128..+128> - Relative transpose. Adds param to the transpose setting.

  • l<length> - Set the default length
  • C<4..256> - Set the whole note length (in ticks).
    • Default is #Zenlen or 96 if #Zenlen not set.

  • v<0..16> - Set coarse volume
  • V<0..255> - Set fine volume
  • v-<1..16> - Decrement volume (coarsely)
  • v+<1..16> - Increment volume (coarsely)
  • V-<1..255> - Decrement volume (finely)
  • V+<1..255> - Increment volume (finely)
  • p<0..128> - Set pan
    • A pan of 0 will output to the left speaker only. A pan of 128 will output to the right speaker only.
    • The default pan is 64 (centered).
    • Panning is disabled if the stereo flag is false (mono output).
  • px<-64..+64> - Set pan
    • Negative px values pan to the left. px-64 is 100% to the left.
    • Positive px values pan to the right. px+64 is 100% to the right.
    • px0 is centered
    • Panning is disabled if the stereo flag is false (mono output).
  • p-<1..128> - Decrement pan (pan to the left)
  • p+<1..128> - Increment pan (pan to the right)

  • vs+<1..16>,<ticks> - Volume slide up (coarse volume)

    • Volume slide is disabled when the volume is set or adjusted.
    • Volume slide is disabled on overflows and underflows.

  • vs-<1..16>,<ticks> - Volume slide down (coarse volume)

  • Vs+<1..255>,<ticks> - Volume slide up (fine volume)

  • Vs-<1..255>,<ticks> - Volume slide down (fine volume)

  • v~<1..8>,<quarter_wavelength_in_ticks> - Tremolo (coarse volume)

    • Tremolo will amplitude modulate the channel volume in a sawtooth pattern every tick.
    • The first parameter is the center-to-peak depth in volume units.
    • quarter_wavelength_in_ticks controls the rate (speed) of the tremolo.
    • Tremolo is disabled when the volume is set or adjusted.
    • Tremolo is disabled on overflows and underflows.

  • V~<1..127>,<quarter_wavelength_in_ticks> - Tremolo (fine volume)

  • ps+<1..128>,<ticks> - Pan slide up

    • Pan slide is disabled when the pan is set or adjusted.
    • Pan slide is disabled on overflows and underflows.

  • ps-<1..128>,<ticks> - Pan slide down

  • p~<1..64>,<quarter_wavelength_in_ticks> - Panbrello

    • Panbrello will amplitude modulate the pan in a sawtooth pattern every tick.
    • The first parameter is the center-to-peak depth in pan units.
    • quarter_wavelength_in_ticks controls the rate (speed) of the panbrello.
    • Panbrello is disabled when the pan is set or adjusted.
    • Panbrello is disabled on overflows and underflows.

  • i[L][R][M] - Set channel phase invert
    • L - invert left channel if in surround mode
    • R - invert right channel if in surround mode
    • M - invert both channels if in mono or stereo mode
    • Multiple invert flags can be set at once. For example, iLR will invert the left and right channels in surround mode but no channels will be inverted in mono or stereo mode.
    • If no L/R/M is given, all 3 flags will be set.
    • CAUTION: There must be a space after the i command.
    • CAUTION: L and R are ignored in stereo mode.
  • iB - Phase invert with the left, right and mono invert-flags set
  • i0 - Disable phase invert

  • E - Enable echo
  • E1 - Enable echo
  • E0 - Disable echo

  • [ <mml> ]<2..255> - Loop
    • Repeats a section of MML a given number of times
    • The broken chord command {{ }} consumes a loop.
  • [ <mml1> : <mml2> ]<2..255> - Loop
    • Repeats a section of MML a given number of times, skipping the MML after the : on the last loop.
    • For example: [c : e r]4 will play c e r c e r c e r c
  • !<id> - Call subroutine
    • A subroutine may change the instrument.
    • Calling a subroutine disables MP vibrato.
  • L - Set loop point
    • If a loop point is set, the channel will restart at the loop point when the channel has reached the end.
  • ; - Comment. The text between a ; and a new line is ignored.
  • | - Divider. Has no effect and can be used for aesthetic reasons (ie, splitting a line into bars).

  • {[pitch1] <pitch2>} [total_length] [, delay_length] [, portamento_speed] - Portamento

    • Smoothly transition between two pitches.
    • If delay_length is set, there will be a delay between pitch1 key-on and the pitch-shift. For example: {df}4,8 will delay the portamento by a eighth note (d8 & {df}8).
      • If delay_length is set, it must be less than total_length.
    • The S-DSP PITCH register will be incremented/decremented by a fixed value on every tick. This value is calculated by the MML compiler and can be overridden by setting portamento_spped.
      • The MML compiler needs to know which instrument is playing the pitch. If you are using portamento in a subroutine, you will need to either set an instrument (@) before playing a portamento or manually set portamento_speed.
      • portamento_speed can be set without setting delay_length, ie: {df}4,,50
    • Can be tied (^) and slurred (&) like a regular note
    • The octave can be changed inside the braces. For example: {a > c}2 and {o3 c o4 c}2
    • The pitches can be P, PR or PF commands
    • pitch1 is optional if the previous note is slurred and known (ie, not at the start of a loop)

  • {{<pitch list>}} [total_length] [, note_length] [, tie] - Broken Chord

    • Quickly cycle through the pitches for a given duration (length).
    • For example, {{ceg}} will quickly alternate between c, e and g every tick with a single key-on event.
    • total_length the total length of the broken chord.
    • note_length is the length of each note played.
      • By default this value is %1 if tie is 1 (default) or %2 if tie is 0.
    • The tie flag indicates if the pitches are tied or not.
      • If 1 (default), the pitches will be tied, a single key-on event and key-off event will be emitted.
      • If 0, There will be a key-on and key-off event for each pitch.
    • The octave can be changed inside the braces (like portamento).
    • The pitches can be P, PR or PF play-pitch commands
    • This command consumes a loop.
    • Examples (Whole Note Length is 96, default l is 4):
      • {{ceg}} expands to [c%1 & : e%1 & g%1 &]8 e%2
      • {{ce-g}}2,32 expands to [c32 & e-32 & g32 &]5 c32
      • {{c}}4,,0 expands to [c%2]12
      • {{de}}4,,0 expands to [d%2 e%2]6
      • {{fg}}4,%3,0 expands to [f%3 g%3]4
      • {{ab}}4,16,0 expands to [a16 b16]2

  • ~0 - Disable manual vibrato

  • ~<pitch_offset_per_tick>,<quarter_wavelength_in_ticks> - Manual vibrato

    • Generates a set_vibrato audio instruction bytecode.
    • Vibrato will add or subtract a pitch_offset_per_tick value to the S-DSP PITCH register in a sawtooth pattern on every tick.
    • quarter_wavelength_in_ticks value controls the rate of vibrato.
    • Manual vibrato will disable MP Vibrato.
    • Manual vibrato will persist across subroutine calls.

  • MP0 - Disable MP vibrato

    • MP Vibrato does not take effect immediately. Vibrato will be disabled on the next note.

  • MP<depth_in_cents>,<quarter_wavelength_in_ticks> - MP Vibrato

    • depth_in_cents controls the depth of the vibrato, in cents either side of the note (half-extent).
    • quarter_wavelength_in_ticks value controls the rate of vibrato.
    • When MP Vibrato is activated the MML compiler will calculate a pitch_offset_per_tick value for each subsequent note played. This greatly simplifies and speedups the SPC-700 code, but requires the MML compiler to know which instrument is playing the note.
    • MP Vibrato does not take effect immediately. The vibrato starts on the next note played.
    • CAUTION: MP Vibrato does not persist across subroutine calls. When calling a ! subroutine, MP vibrato is temporally disabled and MP vibrato resumes after the ! mml subroutine returns.

  • Q<1-8> - Quantize
    • Cuts the note by param/8. The default value is 8.
    • Has no effect on slurred notes.
    • For example: Q4 c4 will cut the notes length in half (4/8) and play c8 r8
    • Q8 (the default) will not cut notes.
  • Q%<0-255> - Fine Quantize
    • Same as Q, except it cuts the notes by param/256.
  • Q<1-7>,<temp-gain> / Q%<0-255>,<temp-gain> - Quantize with Temp-GAIN
    • Instead of a hard 1 tick keyoff, this command will a set temp-GAIN, then rest.
    • For example: Q4,D10 c4 will expand to c8 & GDT10 r8
    • Q<n>,D<rate> - Quantize with linear decrease envelope (0-31)
    • Q<n>,E<rate> - Quantize with exponential decrease envelope (0-31)
    • Q<n>,I<rate> - Quantize with linear increase envelope (0-31)
    • Q<n>,B<rate> - Quantize with bent increase envelope (0-31)
    • Q<n>,F<value> - Quantize with fixed envelope (1-127)
    • Q<n>,<gain> - Quantize with a raw GAIN value (1-255)
    • CAUTION: This command changes temporary GAIN on the next note.
  • q<1-254>[,<1-255>][,<gain>] - early release (sound cut)
    • First parameter: number of ticks to cut the note earlier then normal.
    • Second parameter: minimum ticks before a note is cut (default 1 tick).
    • Third parameter: temporally change the envelope to GAIN(gain) on early-release.
    • If there is no GAIN parameter, the note will be key-offed on early-release.
    • Examples:
      • q4 c%24 is equivalent to c%20 w%4.
      • q20,8 c%48 is equivalent to c%28 w%20.
      • q20,8 c%24 is equivalent to c%9 w%11 (8 ticks playing c, then key-off, then wait for 11 more ticks).
    • The gain parameter is useful for custom release envelopes (ie, q10,D16 or q8,E28).
      • Of the various ways to write custom release envelopes, the q command uses the least audio-RAM.
    • The minimum ticks parameter can be skipped:
      • q<n>,D<rate> - linear decrease envelope (0-31)
      • q<n>,E<rate> - exponential decrease envelope (0-31)
      • q<n>,I<rate> - linear increase envelope (0-31)
      • q<n>,B<rate> - bent increase envelope (0-31)
      • q<n>,F<rate> - fixed envelope (1-127)
    • CAUTION: the GAIN rate may need to be changed if the song's tempo changes.
    • CAUTION: q persists across subroutine calls and : skip-last-loop commands.
  • q0 - Disable early-release.

  • D<-16383..+16383> - detune
    • Adds an offset to the VxPITCH register in play-note and portamento instructions.
    • CAUTION: P play pitch command is unaffected by detune.
    • CAUTION: detune is not disabled when the instrument is changed.
    • CAUTION: detune persists across subroutine calls
  • D0 - disable detune
  • MD<-600..+600> - automatic detune by cents
    • Detunes play-note and portamento commands by a given number of cents.
    • MD automatically calculates the VxPITCH detune and adds a D detune command before each note. For example, MD+20 c d e would expand to D+25 c D+28 d D+31 e.
    • CAUTION: MD increases the song data by 2 or 3 bytes per note. If you have a lot of notes with a fixed MD cents, you could clone and detune the instrument instead.
  • MD0 - disable automatic detune
    • CAUTION: Also disables D detune

  • PM - enable pitch modulation
    • The previous channel's OUTX (sample * envelope) value is used as the pitch modulation source.
    • Only channels B to F can be pitch modulated.
    • Channels G & H cannot be pitch-modulated as they would conflict with sound-effect channel ducking.
    • CAUTION: tad-gui's Play song from cursor feature cannot accurately emulate pitch-modulation.
      • The rust bytecode interpreter does not fully emulate key-on, portamento, vibrato or envelopes.
      • A pitch-modulation source will only key-on if pitch-modulation is on when playback starts.
  • PM0 - disable pitch modulation

  • \asm { <asm_code> } - bytecode assembly
    • Instructions can be separated by | dividers or newlines.
    • Loops inside \asm blocks must be self-contained. The start_loop and end_loop instructions must be within the same \asm block.
    • See bytecode assembly documentation.

The Following commands affect all channels:

  • t<bpm> - Set song tempo (see #Tempo for the bpm format)
  • T<64-255> - Set song tick-clock (see #Timer for the parameter format)

Global echo commands:

  • CAUTION: hexadecimal numbers in echo commands are byte values. For example \efb $c0 is the same as \efb -64.
  • WARNING: These commands modify the echo S-DSP registers without any checks.
    • Excessive echo feedback will continue to get louder and louder, exploding your song.
    • An overflowing FIR filter can pop. Excessive filter feedback will continue to get louder and louder, exploding your song.
    • Headphone users should turn down their volume when playing with echo filters/feedback.

  • \evol <0..127> - Set the left and right global echo volume to the same value
  • \evol <left 0..127>,<right 0..127> - Set the left and right global echo volume
  • \evol <-127..+127> - Adds the parameter to the left and right global echo volume
  • \evol <left -127..+127>,<right -127..+127> - Adds the stereo parameters to the left and right global echo volumes
  • \efb <-128..127> - Sets the echo feedback
  • \efb+ <1..127>[,max] - Increment echo feedback
    • If max is set, the echo feedback be <= max.
  • \efb- <1..128>[,min] - Decrement echo feedback
    • If min is set, the echo feedback be >= min.
  • \fir { c0 c1 c2 c3 c4 c5 c6 v7 } - Set all 8 FIR filter tap coefficients
  • \ftap <tap 0-7>,<value -128..127> - Set a single FIR filter tap to the given coefficient value
  • \ftap+ <tap 0-7>,<1..127>[,max] - Increment a single FIR filter tap coefficient
    • If max is set, the FIR tap be <= max.
  • \ftap- <tap 0-7>,<1..128>[,min] - Decrement a single FIR filter tap coefficient
    • If min is set, the FIR tap be >= min.
  • \ei <flags> - Invert echo volume
    • 0 - disable echo invert
    • B - invert both left and right channels in all modes
    • L - invert left channel if in surround mode
    • R - invert right channel if in surround mode
    • M - invert both channels if in mono or stereo mode
    • Multiple L/R/M flags can be set at once. For example, iLR will invert the left and right channels in surround mode but no channels will be inverted in mono or stereo mode.
  • \edl <delay> - Change echo delay buffer length in milliseconds
    • This value must be a multiple of 16 and cannot be greater than #MaxEchoLength
    • CAUTION: changing the echo delay while echo is active will cause a glitch.
      • Shrinking the echo buffer delay will drop a section of the echo buffer.
      • Increasing the echo buffer delay will reveal and process uninitialized echo buffer memory.
    • CAUTION: this command does not take effect immediately.
      • EDL register writes only take effect when the S-DSP reaches the end of the echo buffer.
      • Echo volume and feedback should be 0 when changing echo delay.
      • Echo volume and feedback should be restored old-EDL + new-EDL milliseconds after the \edl command. This wait will also clear the uninitialized portion of echo buffer.
    • CAUTION: The play-from-cursor feature of the GUI cannot emulate the echo buffer position

Engine Limitations

This audio engine was designed to be simple, to that effect the following limitations apply:

  • Song channels G and H are also used for sound effects.
    • When a sound effect is playing, song channels G and/or H will be ducked (temporally muted).
    • The sound effects window (SFX button or Menu > Audio > Sound Effects) can be used to test how channel ducking affects the song.
  • To prevent popping, there needs to be a 8 millisecond delay between key-off and key-on events. This is accomplished by:
    • Setting a minimum tick-clock of 8ms.
    • Subtracting a single tick from every non-slurred note and adding a 1-tick rest after a key-off event.
    • The minimum length of a non-slurred note is 2 ticks.
  • All songs and sound effects use the same samples (no sample swapping).
  • Loops and subroutines share a stack.
    • The stack is 21 bytes per channel.
    • 3 bytes of stack are required for a loop.
    • 2 bytes of stack are required for a subroutine call.
    • The compiler will refuse to compile MML that would overflow the stack.
  • The echo buffer is always active, even if #EchoLength is 0.
  • A 250ms delay is required to initialize and clear the echo buffer before it can be written to. All songs will be delayed a minimum of 250ms before the song starts, even if the song does not use echo.

To simplify and speedup the SPC-700 code:

  • All pitches and pitch-offsets are precalculated.
    • The pitches are stored in a 256 entry pitch table.
    • There is no overflow or underflow checks when accessing the pitch table.
    • The MML compiler will check for pitch-out-of-range errors, but only if it knows what instrument the note is play in.
  • Pitch effects (portamento and vibrato) are precalculated by the MML compiler.
    • To calculate the pitch-offset-per-tick value, the MML compiler needs to know which instrument is playing which portamento/vibrato note.