0
mirror of https://github.com/torvalds/GuitarPedal.git synced 2026-08-18 13:13:35 +00:00
Files
torvalds-GuitarPedal/Audio/analyze.h
Linus Torvalds 0c1b9c3db3 Split Software/ into the four things it actually was
'Software' was the directory everything that was not KiCad ended up in,
which stopped describing anything a while ago - Validation and the web
app are software too.  Worse, it put the shared parts inside the
firmware, where they read as the firmware's own.

They are not.  Effects/ has three consumers built from it: the firmware,
Validation's bench, and the web app's controls, all generated from the
same POT: comments by gen_effects.py.  Audio/ has two - the bench
compiles the same biquads, the same envelope followers and the same
single_sample(), which is the whole reason a measurement on a
workstation says anything about the pedal.  Neither belongs under
Firmware/, so neither is under it any more:

  Effects/    one file per effect
  Audio/      the DSP they are built from, and the audio loop
  Firmware/   the rest of what runs on the pedal, and the submodules
  WebMIDI/    the web app
  scripts/    what the build runs
  Validation/ unchanged
  Hardware/, Documentation/, Images/

CMakeLists.txt and the wrapper Makefile move to the top with them,
because the build now consumes four of those directories and generates
into a fifth.  board.local and build/ come along; MIDI_CC_MAP.md is
generated into Documentation/ rather than into the old Software/ root.

scripts/ goes with the build rather than staying under the firmware,
because six of the ten had nothing to do with the firmware: gen_effects.py
reads Effects/ and writes to three different places, pow2/log2/quarter_sine
generate Audio/'s tables, check-readme.py compares Effects/ against the
README, and server.py serves the web app.  Four of them are invoked from
Validation, which was reaching into Firmware/ for tooling - the same
burying this commit is undoing.  The four that really are about the
firmware are ELF checks the top-level build drives anyway, and a second
scripts directory would only be a second place to look.

C includes say "Audio/foo.h" and the generated map says
"Effects/bar.h", with the repository root on the include path for both
the firmware and the bench.  Spelling the directory out rather than
relying on a bare name is what keeps Audio/cycles.h shimmable: a quoted
include searches the including file's own directory first.

The submodules are renamed as well as moved.  git mv updates their paths
but leaves the section names, and 'Software/pico-sdk' surviving in
.gitmodules would be the word this commit removes, still load-bearing.
That meant the nested modules under pico-sdk too - six .git files
pointing into .git/modules/Software - which is why 'git submodule update
--init --recursive' is worth running once after pulling this.

Verified rather than assumed: a clean configure and build, make check
(failing only on the missing-eeprom case it already failed on),
check-effects, all four analysis pages reproducing every series and
drawing every chart, and a flash to the board that still measures a
routed reverb where it did before.

One latent bug fell out of it.  bench/coeff declared only quarter_sine.h
of the three generated math tables, and Audio/util.h includes pow2.h and
log2.h as well - so building that target with an empty gen/ could never
have worked.  'make bench' builds bench/bench first, which generates all
three, so it stayed hidden until this rebuilt everything from nothing.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2026-08-11 13:48:26 -07:00

64 lines
1.6 KiB
C

#define FFT_SHIFT 13
#define FFT_SIZE (1 << FFT_SHIFT)
#define ANALYZE_RING_SHIFT 14
#define ANALYZE_RING_SIZE (1 << ANALYZE_RING_SHIFT)
#define ANALYZE_RING_MASK (ANALYZE_RING_SIZE - 1)
//
// NOTE! This is accessed from both cores, but the
// logic is that the audio core only writes to 'ring_buf'
// and increments 'write_index'. The UI core independently
// reads from the ring buffer and maintains the read index.
//
struct analyze_state {
float ring_buf[ANALYZE_RING_SIZE];
unsigned int write_index;
unsigned int read_index;
} analyzer;
// Hann function using the quarter_sine table. We don't
// do the standard "(1-cos(x))/2", we do "sin^2(x/2)"
// instead, and only use half the sine cycle.
//
// Half a sine cycle is the same as walking the quarter
// cycle forward and then backward.
static inline float hanning(unsigned int idx)
{
const int fractional_bits = FFT_SHIFT - QUARTER_SINE_STEP_SHIFT -1;
float frac = u32_to_fraction(idx << (32 - fractional_bits));
idx >>= fractional_bits;
unsigned int next = idx+1;
if (idx >= QUARTER_SINE_STEPS) {
idx = QUARTER_SINE_STEPS*2 - idx;
next = idx-1;
}
float sin = linear(frac, quarter_sin[idx], quarter_sin[next]);
return sin*sin;
}
// 4x downsampled data into continuous lock-free ring buffer
static inline void analyze_process_sample(sample_t sample)
{
// Downsample by 4x
static float sample_sum;
static int count;
float left = sample.left;
left += sample_sum;
if (3 & ++count) {
sample_sum = left;
return;
}
sample_sum = 0;
unsigned int idx = analyzer.write_index;
analyzer.ring_buf[idx & ANALYZE_RING_MASK] = left;
smp_store_release(&analyzer.write_index, idx + 1);
}