demod: lowpass cleanup

ui32_t n -> int n  helps -Ofast auto-vectorization

demod/iq_svcl/iq_base.c

@@ -305,7 +305,7 @@ static int f32_cblk(dsp_t *dsp) {
     int BL = dsp->decM * blk_sz;
     int len = BL;
     float x, y;
-    ui8_t s[4*2*BL]; //uin8,int16,flot32
+    ui8_t s[4*2*BL]; //uin8,int16,float32
     ui8_t *u = (ui8_t*)s;
     short *b = (short*)s;
     float *f = (float*)s;
@@ -488,20 +488,21 @@ 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) {
-    ui32_t n;
-    ui32_t s = sample % taps;
-    double complex w = 0;
+//static __attribute__((optimize("-ffast-math"))) float complex lowpass()
+static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    float complex w = 0;
+    int n; // -Ofast
+    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]
 }
-static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
-    float complex w = 0;     // -Ofast
+static float complex lowpass2(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    float complex w = 0;
     int n;
-    int s = sample % taps; // lpIQ
+    int s = sample % taps;
     int S1 = s+1;
     int S1N = S1-taps;
     int n0 = taps-1-s;

demod/iq_svcl/iq_dec.c

@@ -336,7 +336,7 @@ 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;
@@ -461,20 +461,21 @@ 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) {
-    ui32_t n;
-    ui32_t s = sample % taps;
-    double complex w = 0;
+//static __attribute__((optimize("-ffast-math"))) float complex lowpass()
+static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    float complex w = 0;
+    int n; // -Ofast
+    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]
 }
-static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
-    float complex w = 0;     // -Ofast
+static float complex lowpass2(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    float complex w = 0;
     int n;
-    int s = sample % taps; // lpIQ
+    int s = sample % taps;
     int S1 = s+1;
     int S1N = S1-taps;
     int n0 = taps-1-s;

demod/iq_svcl/iq_fm.c

@@ -273,9 +273,18 @@ static int lowpass_init(float f, int taps, float **pws) {
     return taps;
 }
 
+static float complex lowpass0(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    ui32_t n;
+    double complex w = 0;
+    for (n = 0; n < taps; n++) {
+        w += buffer[(sample+n+1)%taps]*ws[taps-1-n];
+    }
+    return (float complex)w;
+}
+//static __attribute__((optimize("-ffast-math"))) float complex lowpass()
 static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
-    float complex w = 0;     // -Ofast
-    int n;
+    float complex w = 0;
+    int n; // -Ofast
     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]

demod/mod/demod_mod.c

@@ -625,20 +625,20 @@ static float complex lowpass0(float complex buffer[], ui32_t sample, ui32_t taps
     }
     return (float complex)w;
 }
-static float complex lowpass1d(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
-    ui32_t n;
-    ui32_t s = sample % taps;
+static float complex lowpass1a(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
     double complex w = 0;
+    ui32_t n;
+    ui32_t S = taps + (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;
 // symmetry: ws[n] == ws[taps-1-n]
 }
-//static __attribute__((optimize("-ffast-math"))) float complex lowpass1bf()
+//static __attribute__((optimize("-ffast-math"))) float complex lowpass()
 static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
-    float complex w = 0;     // -Ofast
-    int n;
+    float complex w = 0;
+    int n; // -Ofast
     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]
@@ -646,11 +646,10 @@ static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t 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
+    float complex w = 0;
+    int n; // -Ofast
+    int s = sample % taps;
     int S1 = s+1;
     int S1N = S1-taps;
     int n0 = taps-1-s;

demod/multi/demod_base.c

@@ -699,20 +699,21 @@ 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) {
-    ui32_t n;
-    ui32_t s = sample % taps;
-    double complex w = 0;
+//static __attribute__((optimize("-ffast-math"))) float complex lowpass()
+static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    float complex w = 0;
+    int n; // -Ofast
+    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]
 }
-static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
-    float complex w = 0;     // -Ofast
-    int n;
-    int s = sample % taps; // lpIQ
+static float complex lowpass2(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    float complex w = 0;
+    int n; // -Ofast
+    int s = sample % taps;
     int S1 = s+1;
     int S1N = S1-taps;
     int n0 = taps-1-s;

imet/imet4iq.c

@@ -394,18 +394,28 @@ static int lowpass_init(float f, int taps, float **pws) {
     return taps;
 }
 
-
-static float complex lowpass1(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
-    ui32_t n;
-    ui32_t s = sample % taps;
+static float complex lowpass1a(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
     double complex w = 0;
+    ui32_t n;
+    ui32_t S = taps + (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;
 // symmetry: ws[n] == ws[taps-1-n]
 }
+//static __attribute__((optimize("-ffast-math"))) float complex lowpass()
 static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    float complex w = 0;
+    int n; // -Ofast
+    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 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

mk2a/mk2a1680mod.c

@@ -645,7 +645,7 @@ 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;
@@ -732,21 +732,31 @@ static int lowpass_init(float f, int taps, float **pws) {
     return taps;
 }
 
-
-static float complex lowpass1(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
-    ui32_t n;
-    ui32_t s = sample % taps;
+static float complex lowpass1a(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
     double complex w = 0;
+    ui32_t n;
+    ui32_t S = taps + (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;
 // symmetry: ws[n] == ws[taps-1-n]
 }
+//static __attribute__((optimize("-ffast-math"))) float complex lowpass()
 static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
-    float complex w = 0;     // -Ofast
+    float complex w = 0;
+    int n; // -Ofast
+    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 float complex lowpass2(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    float complex w = 0;
     int n;
-    int s = sample % taps; // lpIQ
+    int s = sample % taps;
     int S1 = s+1;
     int S1N = S1-taps;
     int n0 = taps-1-s;

scan/dft_detect.c

@@ -650,20 +650,21 @@ 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) {
-    ui32_t n;
-    ui32_t s = sample % taps;
-    double complex w = 0;
+//static __attribute__((optimize("-ffast-math"))) float complex lowpass()
+static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    float complex w = 0;
+    int n; // -Ofast
+    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]
 }
-static float complex lowpass(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
-    float complex w = 0;     // -Ofast
+static float complex lowpass2(float complex buffer[], ui32_t sample, ui32_t taps, float *ws) {
+    float complex w = 0;
     int n;
-    int s = sample % taps; // lpIQ
+    int s = sample % taps;
     int S1 = s+1;
     int S1N = S1-taps;
     int n0 = taps-1-s;