1
0
mirror of https://github.com/cookiengineer/audacity synced 2025-10-11 00:53:46 +02:00

Update Nyquist to v3.09.

This commit is contained in:
Leland Lucius
2015-04-07 22:10:17 -05:00
parent f88b27e6d8
commit 9fb0ce5b82
358 changed files with 26327 additions and 7043 deletions

View File

@@ -9,7 +9,7 @@
#include "cext.h"
#include "stkchorus.h"
void stkchorus_free();
void stkchorus_free(snd_susp_type a_susp);
typedef struct stkchorus_susp_struct {
@@ -23,12 +23,12 @@ typedef struct stkchorus_susp_struct {
struct stkEffect *mych;
} stkchorus_susp_node, *stkchorus_susp_type;
#include "stkint.h"
#include "stkint.h"
void stkchorus_n_fetch(register stkchorus_susp_type susp, snd_list_type snd_list)
void stkchorus_n_fetch(snd_susp_type a_susp, snd_list_type snd_list)
{
stkchorus_susp_type susp = (stkchorus_susp_type) a_susp;
int cnt = 0; /* how many samples computed */
int togo;
int n;
@@ -56,6 +56,7 @@ void stkchorus_n_fetch(register stkchorus_susp_type susp, snd_list_type snd_list
if (susp->terminate_cnt != UNKNOWN &&
susp->terminate_cnt <= susp->susp.current + cnt + togo) {
togo = susp->terminate_cnt - (susp->susp.current + cnt);
if (togo < 0) togo = 0; /* avoids rounding errros */
if (togo == 0) break;
}
@@ -67,6 +68,7 @@ void stkchorus_n_fetch(register stkchorus_susp_type susp, snd_list_type snd_list
* AND cnt > 0 (we're not at the beginning of the
* output block).
*/
if (to_stop < 0) to_stop = 0; /* avoids rounding errors */
if (to_stop < togo) {
if (to_stop == 0) {
if (cnt) {
@@ -89,9 +91,7 @@ void stkchorus_n_fetch(register stkchorus_susp_type susp, snd_list_type snd_list
s1_ptr_reg = susp->s1_ptr;
out_ptr_reg = out_ptr;
if (n) do { /* the inner sample computation loop */
*out_ptr_reg++ = (sample_type) (stkEffectTick(mych_reg, *s1_ptr_reg++))
;
*out_ptr_reg++ = (sample_type) (stkEffectTick(mych_reg, *s1_ptr_reg++));
} while (--n); /* inner loop */
susp->mych = mych_reg;
@@ -118,103 +118,9 @@ void stkchorus_n_fetch(register stkchorus_susp_type susp, snd_list_type snd_list
} /* stkchorus_n_fetch */
void stkchorus_s_fetch(register stkchorus_susp_type susp, snd_list_type snd_list)
{
int cnt = 0; /* how many samples computed */
int togo;
int n;
sample_block_type out;
register sample_block_values_type out_ptr;
register sample_block_values_type out_ptr_reg;
register struct stkEffect * mych_reg;
register sample_type s1_scale_reg = susp->s1->scale;
register sample_block_values_type s1_ptr_reg;
falloc_sample_block(out, "stkchorus_s_fetch");
out_ptr = out->samples;
snd_list->block = out;
while (cnt < max_sample_block_len) { /* outer loop */
/* first compute how many samples to generate in inner loop: */
/* don't overflow the output sample block: */
togo = max_sample_block_len - cnt;
/* don't run past the s1 input sample block: */
susp_check_term_log_samples(s1, s1_ptr, s1_cnt);
togo = min(togo, susp->s1_cnt);
/* don't run past terminate time */
if (susp->terminate_cnt != UNKNOWN &&
susp->terminate_cnt <= susp->susp.current + cnt + togo) {
togo = susp->terminate_cnt - (susp->susp.current + cnt);
if (togo == 0) break;
}
/* don't run past logical stop time */
if (!susp->logically_stopped && susp->susp.log_stop_cnt != UNKNOWN) {
int to_stop = susp->susp.log_stop_cnt - (susp->susp.current + cnt);
/* break if to_stop == 0 (we're at the logical stop)
* AND cnt > 0 (we're not at the beginning of the
* output block).
*/
if (to_stop < togo) {
if (to_stop == 0) {
if (cnt) {
togo = 0;
break;
} else /* keep togo as is: since cnt == 0, we
* can set the logical stop flag on this
* output block
*/
susp->logically_stopped = true;
} else /* limit togo so we can start a new
* block at the LST
*/
togo = to_stop;
}
}
n = togo;
mych_reg = susp->mych;
s1_ptr_reg = susp->s1_ptr;
out_ptr_reg = out_ptr;
if (n) do { /* the inner sample computation loop */
*out_ptr_reg++ = (sample_type) (stkEffectTick(mych_reg, (s1_scale_reg * *s1_ptr_reg++)))
;
} while (--n); /* inner loop */
susp->mych = mych_reg;
/* using s1_ptr_reg is a bad idea on RS/6000: */
susp->s1_ptr += togo;
out_ptr += togo;
susp_took(s1_cnt, togo);
cnt += togo;
} /* outer loop */
/* test for termination */
if (togo == 0 && cnt == 0) {
snd_list_terminate(snd_list);
} else {
snd_list->block_len = cnt;
susp->susp.current += cnt;
}
/* test for logical stop */
if (susp->logically_stopped) {
snd_list->logically_stopped = true;
} else if (susp->susp.log_stop_cnt == susp->susp.current) {
susp->logically_stopped = true;
}
} /* stkchorus_s_fetch */
void stkchorus_toss_fetch(susp, snd_list)
register stkchorus_susp_type susp;
snd_list_type snd_list;
{
long final_count = susp->susp.toss_cnt;
void stkchorus_toss_fetch(snd_susp_type a_susp, snd_list_type snd_list)
{
stkchorus_susp_type susp = (stkchorus_susp_type) a_susp;
time_type final_time = susp->susp.t0;
long n;
@@ -229,53 +135,53 @@ void stkchorus_toss_fetch(susp, snd_list)
susp->s1_ptr += n;
susp_took(s1_cnt, n);
susp->susp.fetch = susp->susp.keep_fetch;
(*(susp->susp.fetch))(susp, snd_list);
(*(susp->susp.fetch))(a_susp, snd_list);
}
void stkchorus_mark(stkchorus_susp_type susp)
void stkchorus_mark(snd_susp_type a_susp)
{
stkchorus_susp_type susp = (stkchorus_susp_type) a_susp;
sound_xlmark(susp->s1);
}
void stkchorus_free(stkchorus_susp_type susp)
void stkchorus_free(snd_susp_type a_susp)
{
deleteStkEffect(susp->mych);
stkchorus_susp_type susp = (stkchorus_susp_type) a_susp;
deleteStkEffect(susp->mych);
sound_unref(susp->s1);
ffree_generic(susp, sizeof(stkchorus_susp_node), "stkchorus_free");
}
void stkchorus_print_tree(stkchorus_susp_type susp, int n)
void stkchorus_print_tree(snd_susp_type a_susp, int n)
{
stkchorus_susp_type susp = (stkchorus_susp_type) a_susp;
indent(n);
stdputstr("s1:");
sound_print_tree_1(susp->s1, n);
}
sound_type snd_make_stkchorus(sound_type s1, double baseDelay, double depth, double freq, double mix, rate_type sr)
sound_type snd_make_stkchorus(sound_type s1, double baseDelay, double depth, double freq, double mix)
{
register stkchorus_susp_type susp;
/* sr specified as input parameter */
rate_type sr = s1->sr;
time_type t0 = s1->t0;
int interp_desc = 0;
sample_type scale_factor = 1.0F;
time_type t0_min = t0;
/* combine scale factors of linear inputs (S1) */
scale_factor *= s1->scale;
s1->scale = 1.0F;
/* try to push scale_factor back to a low sr input */
if (s1->sr < sr) { s1->scale = scale_factor; scale_factor = 1.0F; }
falloc_generic(susp, stkchorus_susp_node, "snd_make_stkchorus");
susp->mych = initStkChorus(baseDelay, depth, freq, round(sr));
stkEffectSetMix(susp->mych, mix);
/* select a susp fn based on sample rates */
interp_desc = (interp_desc << 2) + interp_style(s1, sr);
switch (interp_desc) {
case INTERP_n: susp->susp.fetch = stkchorus_n_fetch; break;
case INTERP_s: susp->susp.fetch = stkchorus_s_fetch; break;
default: snd_badsr(); break;
}
susp->susp.fetch = stkchorus_n_fetch;
susp->terminate_cnt = UNKNOWN;
/* handle unequal start times, if any */
if (t0 < s1->t0) sound_prepend_zeros(s1, t0);
@@ -284,8 +190,8 @@ stkEffectSetMix(susp->mych, mix);
/* how many samples to toss before t0: */
susp->susp.toss_cnt = (long) ((t0 - t0_min) * sr + 0.5);
if (susp->susp.toss_cnt > 0) {
susp->susp.keep_fetch = susp->susp.fetch;
susp->susp.fetch = stkchorus_toss_fetch;
susp->susp.keep_fetch = susp->susp.fetch;
susp->susp.fetch = stkchorus_toss_fetch;
}
/* initialize susp state */
@@ -304,8 +210,8 @@ stkEffectSetMix(susp->mych, mix);
}
sound_type snd_stkchorus(sound_type s1, double baseDelay, double depth, double freq, double mix, rate_type sr)
sound_type snd_stkchorus(sound_type s1, double baseDelay, double depth, double freq, double mix)
{
sound_type s1_copy = sound_copy(s1);
return snd_make_stkchorus(s1_copy, baseDelay, depth, freq, mix, sr);
return snd_make_stkchorus(s1_copy, baseDelay, depth, freq, mix);
}