FFT plugin: eliminate 64-bit math. This should result in faster and probably more accurate calculations.

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@25790 a1c6a512-1295-4272-9138-f99709370657

1 files changed, 21 insertions, 14 deletions

diff --git a/apps/plugins/fft/fft.c b/apps/plugins/fft/fft.c
index ae07179557..acc19fabe5 100644
--- a/apps/plugins/fft/fft.c
+++ b/apps/plugins/fft/fft.c
@@ -307,8 +307,8 @@ struct {
 /************************* End of globals *************************/
 
 /************************* Math functions *************************/
-#define QLOG_MAX 286286
-#define QLIN_MAX 1534588906
+#define QLOG_MAX 0x00040000 
+#define QLIN_MAX 0x5B000000
 #define QLN_10 float_q16(2.302585093)
 #define LIN_MAX (QLIN_MAX >> 16)
 
@@ -350,24 +350,31 @@ void apply_window_func(char mode)
 /* Calculates the magnitudes from complex numbers and returns the maximum */
 int32_t calc_magnitudes(bool logarithmic)
 {
-    int64_t tmp;
+    /* A major assumption made when calculating the Q*MAX constants 
+     * is that the maximum magnitude is 29 bits long. */
+
+    uint32_t tmp;
     size_t i;
 
-    int32_t max = -2147483647;
-    
-    /* Calculate the magnitude, discarding the phase.
-     * The sum of the squares can easily overflow the 15-bit (s15.16)
-     * requirement for fsqrt, so we scale the data down */
+    int32_t max = 0;
+
+    /* Calculate the magnitude, discarding the phase. */
     for (i = 0; i < ARRAYSIZE_PLOT; ++i)
     {
         tmp = output[i].r * output[i].r + output[i].i * output[i].i;
-        tmp <<= 16;
-
-        tmp = fsqrt64(tmp, 16);
-
-        if (logarithmic)
-            tmp = get_log_value(tmp & 0x7FFFFFFF);
 
+        if (tmp > 0x7FFFFFFF) tmp >>= 1; /* if our assumptions are correct,
+                                            this should never happen. It's just
+                                            a safeguard. */
+        if (tmp > 0)
+        {
+            tmp = fp_sqrt(tmp, 0); /* linear scaling, nothing
+                                      bad should happen */
+            tmp <<= 16;
+            if (logarithmic)
+                tmp = get_log_value(tmp);/* the log function
+                                            expects s15.16 values */
+        }
         plot[i] = tmp;
 
         if (plot[i] > max)