From 4121657da9d2a8dbb1bfff1dbab98f2330efb1db Mon Sep 17 00:00:00 2001
From: KubaPro010 <kuba.pro010@gmail.com>
Date: Thu, 27 Mar 2025 21:19:33 +0100
Subject: [PATCH] this

---
 lib/filters.c    | 43 +++++++++++++++++--------------------------
 lib/oscillator.c | 24 +++++-------------------
 lib/oscillator.h |  1 -
 3 files changed, 22 insertions(+), 46 deletions(-)

diff --git a/lib/filters.c b/lib/filters.c
index 52333ce..aba54c3 100644
--- a/lib/filters.c
+++ b/lib/filters.c
@@ -33,39 +33,30 @@ void init_lpf(LPFFilter *filter, float cutoff, int sample_rate) {
 
 float process_lpf(LPFFilter *filter, float x) {
 #if USE_NEON
-	float32x4_t x_vec = vdupq_n_f32(x);
-	float32x4_t y_vec = x_vec;
+	float32_t input = x;
 
-	float32x4_t *A_vec = (float32x4_t*)filter->A;
-	float32x4_t *d1_vec = (float32x4_t*)filter->d1;
-	float32x4_t *d2_vec = (float32x4_t*)filter->d2;
-	float32x4_t *w1_vec = (float32x4_t*)filter->w1;
-	float32x4_t *w2_vec = (float32x4_t*)filter->w2;
+	float32x4_t y_vec = vdupq_n_f32(input);
 
-	for (int i = 0; i < LPF_ORDER; i += 4) {
-		float32x4_t w1_term = vmulq_f32(*d1_vec, *w1_vec);
-		float32x4_t w2_term = vmulq_f32(*d2_vec, *w2_vec);
-		float32x4_t w0_new = vaddq_f32(vaddq_f32(w1_term, w2_term), y_vec);
+	for (int i = 0; i < LPF_ORDER; i++) {
+		float32x4_t d1_vec = vdupq_n_f32(filter->d1[i]);
+		float32x4_t d2_vec = vdupq_n_f32(filter->d2[i]);
+		float32x4_t w1_vec = vdupq_n_f32(filter->w1[i]);
+		float32x4_t w2_vec = vdupq_n_f32(filter->w2[i]);
+		float32x4_t A_vec = vdupq_n_f32(filter->A[i]);
 
-		float32x4_t two_w1 = vmulq_n_f32(*w1_vec, 2.0f);
-		y_vec = vmulq_f32(*A_vec, vaddq_f32(vaddq_f32(w0_new, two_w1), *w2_vec));
+		float32x4_t w1_term = vmulq_f32(d1_vec, w1_vec);
+		float32x4_t w2_term = vmulq_f32(d2_vec, w2_vec);
+		float32x4_t w0_new_vec = vaddq_f32(vaddq_f32(w1_term, w2_term), y_vec);
 
-		*w2_vec = *w1_vec;
-		*w1_vec = w0_new;
+		float32x4_t two_w1 = vmulq_n_f32(w1_vec, 2.0f);
+		float32x4_t output_term = vaddq_f32(w0_new_vec, vaddq_f32(two_w1, w2_vec));
+		y_vec = vmulq_f32(A_vec, output_term);
 
-		A_vec++;
-		d1_vec++;
-		d2_vec++;
-		w1_vec++;
-		w2_vec++;
+		filter->w2[i] = filter->w1[i];
+		filter->w1[i] = vgetq_lane_f32(w0_new_vec, 0);
 	}
 
-	float32x2_t y_low = vget_low_f32(y_vec);
-	float32x2_t y_high = vget_high_f32(y_vec);
-	float32x2_t y_sum = vadd_f32(y_low, y_high);
-	y_sum = vpadd_f32(y_sum, y_sum);
-
-	return vget_lane_f32(y_sum, 0);
+	return vgetq_lane_f32(y_vec, 0);
 #else
 	float y = x;
 	for (int i = 0; i < LPF_ORDER; i++) {
diff --git a/lib/oscillator.c b/lib/oscillator.c
index 37268b8..b206bea 100644
--- a/lib/oscillator.c
+++ b/lib/oscillator.c
@@ -35,22 +35,8 @@ float get_oscillator_cos_multiplier_ni(Oscillator *osc, float multiplier) {
 }
 
 void advance_oscillator(Oscillator *osc) {
-	#if USE_NEON  // Use NEON if available
-		float32x4_t v_phase = vdupq_n_f32(osc->phase);
-		float32x4_t v_increment = vdupq_n_f32(osc->phase_increment);
-		float32x4_t v_twopi = vdupq_n_f32(M_2PI);
-	
-		v_phase = vaddq_f32(v_phase, v_increment);
-		uint32x4_t v_mask = vcgeq_f32(v_phase, v_twopi);  // Check if phase >= 2π
-		float32x4_t v_wrapped = vsubq_f32(v_phase, v_twopi);
-		v_phase = vbslq_f32(v_mask, v_wrapped, v_phase);
-	
-		osc->phase = vgetq_lane_f32(v_phase, 0);
-	
-	#else  // Scalar fallback if NEON is not available
-		osc->phase += osc->phase_increment;
-		if (osc->phase >= M_2PI) {
-			osc->phase -= M_2PI;
-		}
-	#endif
-	}
\ No newline at end of file
+	osc->phase += osc->phase_increment;
+	if (osc->phase >= M_2PI) {
+		osc->phase -= M_2PI;
+	}
+}
\ No newline at end of file
diff --git a/lib/oscillator.h b/lib/oscillator.h
index e5dfcd8..778c98d 100644
--- a/lib/oscillator.h
+++ b/lib/oscillator.h
@@ -2,7 +2,6 @@
 
 #include "constants.h"
 #include <math.h>
-#include "optimization.h"
 
 typedef struct {
 	float phase;