demod: lowpass reorder speedup

demod/iq_svcl/iq_fm.c

@@ -274,13 +274,13 @@ static int lowpass_init(float f, int taps, float **pws) {
 }
 
 static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
-    ui32_t n;
-    ui32_t s = sample % taps;
-    double complex w = 0;
+    float complex w = 0;     // -Ofast
+    int n;
+    int S = taps-1 - (sample % taps);
     for (n = 0; n < taps; n++) {
-        w += buffer[n]*ws[taps+s-n]; // ws[taps+s-n] = ws[(taps+sample-n)%taps]
+        w += buffer[n]*ws[S+n]; // ws[taps+s-n] = ws[(taps+sample-n)%taps]
     }
-    return (float complex)w;
+    return w;
 // symmetry: ws[n] == ws[taps-1-n]
 }
 

demod/mod/demod_mod.c

@@ -465,7 +465,7 @@ static int f32read_cblock(dsp_t *dsp) {
     int n;
     int len;
     float x, y;
-    ui8_t s[4*2*dsp->decM]; //uin8,int16,flot32
+    ui8_t s[4*2*dsp->decM]; //uin8,int16,float32
     ui8_t *u = (ui8_t*)s;
     short *b = (short*)s;
     float *f = (float*)s;
@@ -625,7 +625,7 @@ static float complex lowpass0(float complex buffer[], ui32_t sample, ui32_t taps
     }
     return (float complex)w;
 }
-static float complex lowpass1(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+static float complex lowpass1d(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
     ui32_t n;
     ui32_t s = sample % taps;
     double complex w = 0;
@@ -635,7 +635,19 @@ static float complex lowpass1(float complex buffer[], ui32_t sample, ui32_t taps
     return (float complex)w;
 // symmetry: ws[n] == ws[taps-1-n]
 }
+//static __attribute__((optimize("-ffast-math"))) float complex lowpass1bf()
 static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    float complex w = 0;     // -Ofast
+    int n;
+    int S = taps-1 - (sample % taps);
+    for (n = 0; n < taps; n++) {
+        w += buffer[n]*ws[S+n]; // ws[taps+s-n] = ws[(taps+sample-n)%taps]
+    }
+    return w;
+// symmetry: ws[n] == ws[taps-1-n]
+}
+//static __attribute__((optimize("-ffast-math"))) float complex lowpass2()
+static float complex lowpass2(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
     float complex w = 0;     // -Ofast
     int n;
     int s = sample % taps; // lpIQ