a48e7b796c
Oops, EOF opcodes weren't being padded
2022-06-11 21:13:36 +01:00
b1bd8f7b19
Don't need most EOF
2022-06-11 17:39:57 +01:00
13fb60fed0
Add a lot of EOF variants - quality is good but they're mostly not used
2022-06-11 17:27:52 +01:00
1552715b49
Optimized EOF sequence
2022-06-04 14:53:33 +01:00
fbd6b956d1
Checkpoint working player
2022-06-04 14:06:52 +01:00
8ea4f150ad
Checkpoint - trying to mitigate EOF clicking
2022-05-24 22:19:40 +01:00
b97a625d7d
Working version of encoder. This reproduces the clicking that happens from end-of-frame, so we will need a new strategy
2022-05-16 21:11:17 +01:00
33ce46a19c
Second order difference equation for damped harmonic oscillator, from "Signal processing in C"
2022-05-16 10:43:40 +01:00
929ea026d9
Format comment
2020-12-29 14:33:22 +00:00
daee564255
Rename slowpath opcode to end_of_frame to better describe what it does.
...
Add command-line flags for audio normalization params.
2020-12-28 22:42:34 +00:00
ccba51eead
Add support for targeting 6502, for which the JMP (indirect) opcode
...
takes 5 cycles instead of 6.
2020-12-28 13:23:57 +00:00
38280e7d93
Add more audio file types
2020-12-28 13:23:18 +00:00
d4bc2d45e5
Parametrize options by command-line flags
2020-12-28 12:54:44 +00:00
4090c502a8
Add back preprocess_audio for converting audio sample rate. It's
...
faster to do this once when converting an audio file many times.
2020-12-28 12:54:39 +00:00
4666171aaa
Count and report on the number of speaker toggles/sec during encoding.
2020-12-28 12:30:29 +00:00
a1a1f33c21
Add partial support for 4-cycle NOP; NOP opcode pairs
...
Count and report on the number of speaker toggles/sec during encoding
2020-12-28 12:29:05 +00:00
089591737f
Regen
2020-12-24 14:44:19 +00:00
cef38dae59
Comment nit
2020-12-24 14:44:09 +00:00
4701dc14f1
Regen
2020-12-24 14:43:51 +00:00
93b63db558
Use the correct effective clock rate for a PAL Apple II, taking into account the long cycle.
2020-12-24 14:43:00 +00:00
0bb7871355
Allow padding out the full $300 page with opcodes, sorted by the
...
number of non-NOP opcodes to give the most opportunities for speaker
actuation.
Use float32 in opcode code.
2020-12-24 14:40:26 +00:00
c776b3cbe2
Refactor to allow passing in other opcode sequences
2020-12-07 21:06:00 +00:00
baca499d3b
Checkpoint
2020-12-07 20:58:34 +00:00
f1b6983dda
Last minute tweaks
2020-12-07 20:58:18 +00:00
706ba54f5b
Further significant optimizations:
...
- Switch some matrices to column-major order, since they usually have
more rows than columns. This turns out to be slightly faster.
- Precompute more of the speaker trajectory formula, since most of it
only depends on the matrix of applied voltages (which only depends
on frame_offset).
Bug fixes:
- make sure to leave enough 0-padding at the end of input data in case
we schedule a SLOWPATH towards the end, since we'll skip into this
buffer.
- Pad the last TCP frame with EXIT opcodes instead of TICK_00.
- Improve the eta counter behaviour by not setting a minimum tick
interval, and being more careful about when we update. This fixes
issues with short files
2020-12-07 20:48:15 +00:00
54369a620f
Wrap line
2020-12-07 20:48:05 +00:00
194da9a509
Check in generated player using NOP, NOP3 and STA opcodes, to max
...
cycle length 19. This seems to give the best quality so far.
2020-12-07 20:40:00 +00:00
1cf1ed25bb
Don't start opcode sequences with NOPs, the encoder mostly avoids
...
these so they just waste space without improving quality.
2020-12-07 20:38:58 +00:00
b8de1d4f7b
Fix a serious bug in cycle_lookahead, it was not flipping the opcode
...
voltage schedule based on the last value of the previous opcode.
In candidate_opcodes, return the hash of pruned_opcodes as well (since
this is relatively expensive to recompute), so this can be used to
precompute and cache partial results of applying the opcodes.
2020-12-07 20:34:57 +00:00
468d47966d
Optimize encoder by memoizing better, and expanding the recurrence
...
relation for the difference equation to use a closed-form solution
that can be vectorized better. This is now about 3x faster.
2020-12-03 14:02:36 +00:00
ccc8594c45
Clean up more.
...
Switch back to longest opcodes first, it is both smaller and slightly
better quality (~3%).
2020-10-15 20:28:26 +01:00
76d6bc8426
Integrate generated player code
2020-10-15 14:27:37 +01:00
22babbaa73
Parametrize normalization percentage
2020-10-15 13:08:33 +01:00
3ef80ec6cf
Overhaul README to describe current status. Still some work needed.
2020-10-15 11:31:25 +01:00
f17db175a5
Simplify
2020-08-25 21:57:42 +01:00
659d9906bd
Generates player audio code and encoder representation, by enumerating
...
all combinations of 6502 opcodes and deduplicating.
2020-08-25 21:54:46 +01:00
8122f65dc6
Optimize memory and speed.
2020-08-25 20:46:00 +01:00
938fbae488
Optimize evolve() by expanding recurrence relation
2020-08-24 22:28:28 +01:00
7a4a36af6b
Checkpoint WIP
...
- use exotic opcodes in player
- optimize encoder
2020-08-24 21:27:24 +01:00
5fb995c435
Credit Scott in README
2020-08-16 23:19:48 +01:00
4767ee51fd
Fixup player timings and opcode variants for 65c02 timings since JMP
...
(indirect) takes 5 cycles not 6! It should be possible to also
accommodate 6502 timings in a followup.
h/t to Scott Duensing who noticed that my sample audio sounded "a tad
slow", which turned out to be due to this 1-cycle difference (which
added up to almost an extra minute playback to an 8-minute song).
Add comments and tidy up the code a bit.
Flesh out README some more.
2020-08-16 23:15:30 +01:00
9e0e1fcbcb
One more fix
2020-08-15 23:24:25 +01:00
790c2bc1ec
Revise and update
2020-08-15 23:22:27 +01:00
86cd1e1114
1MHz encoder and player
...
- we simulate the speaker at full 1-cycle resolution, and
evaluate all possible opcodes that can be scheduled for the
given position in the TCP stream
- this lets us tick the speaker at any cycle interval >=9 (except 10
cycles)
2020-08-13 22:08:50 +01:00
eeeb457b23
Add an example file. This was encoded with lookahead 11.
2020-08-12 22:10:38 +01:00
08f3a58a7f
Add missing makefile and tidy up a bit.
2020-08-12 22:08:27 +01:00
9c08e31cb0
Remove dead code and assemble at $2000 instead.
2020-08-12 21:48:54 +01:00
ebe9cd688e
Happy 70th birthday Woz!
2020-08-12 00:14:20 +01:00
b35e4f86fb
Add shebang and tidy some more
2020-08-11 23:26:55 +01:00
1bbbd8c154
Merge preprocess_audio back in. We didn't end up getting a speed
...
boost from pypy, perhaps because the critical path is in numpy code.
Add some comments and clean up a bit.
2020-08-11 23:23:33 +01:00
kris