From 0bf7e586bc481552a1915a2ff62483cdc5cbc9cd Mon Sep 17 00:00:00 2001
From: Mark Borgerding <mark@borgerding.net>
Date: Tue, 2 Jan 2007 16:20:17 +0000
Subject: [PATCH] Memory footprint fix thanks to Galen Seitz

---
 CHANGELOG         |  3 +++
 Makefile          |  2 +-
 tools/kiss_fftr.c | 10 +++++-----
 3 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/CHANGELOG b/CHANGELOG
index 48debab..517de23 100644
--- a/CHANGELOG
+++ b/CHANGELOG
@@ -1,3 +1,6 @@
+1.2.7 
+  Shrank the real-fft memory footprint. Thanks to Galen Seitz.
+
 1.2.6 (Nov 14, 2006) The "thanks to GenArts" release.
   Added multi-dimensional real-optimized FFT, see tools/kiss_fftndr
   Thanks go to GenArts, Inc. for sponsoring the development.
diff --git a/Makefile b/Makefile
index 2f82a57..ae63f8d 100644
--- a/Makefile
+++ b/Makefile
@@ -1,4 +1,4 @@
-KFVER=1_2_6
+KFVER=1_2_7
 
 DISTDIR=kiss_fft_v$(KFVER)
 TARBALL=kiss_fft_v$(KFVER).tar.gz
diff --git a/tools/kiss_fftr.c b/tools/kiss_fftr.c
index 5bc669d..bfd989f 100644
--- a/tools/kiss_fftr.c
+++ b/tools/kiss_fftr.c
@@ -37,7 +37,7 @@ kiss_fftr_cfg kiss_fftr_alloc(int nfft,int inverse_fft,void * mem,size_t * lenme
     nfft >>= 1;
 
     kiss_fft_alloc (nfft, inverse_fft, NULL, &subsize);
-    memneeded = sizeof(struct kiss_fftr_state) + subsize + sizeof(kiss_fft_cpx) * ( nfft * 2);
+    memneeded = sizeof(struct kiss_fftr_state) + subsize + sizeof(kiss_fft_cpx) * ( nfft * 3 / 2);
 
     if (lenmem == NULL) {
         st = (kiss_fftr_cfg) KISS_FFT_MALLOC (memneeded);
@@ -54,9 +54,9 @@ kiss_fftr_cfg kiss_fftr_alloc(int nfft,int inverse_fft,void * mem,size_t * lenme
     st->super_twiddles = st->tmpbuf + nfft;
     kiss_fft_alloc(nfft, inverse_fft, st->substate, &subsize);
 
-    for (i = 0; i < nfft; ++i) {
+    for (i = 0; i < nfft/2; ++i) {
         double phase =
-            -3.14159265358979323846264338327 * ((double) i / nfft + .5);
+            -3.14159265358979323846264338327 * ((double) (i+1) / nfft + .5);
         if (inverse_fft)
             phase *= -1;
         kf_cexp (st->super_twiddles+i,phase);
@@ -111,7 +111,7 @@ void kiss_fftr(kiss_fftr_cfg st,const kiss_fft_scalar *timedata,kiss_fft_cpx *fr
 
         C_ADD( f1k, fpk , fpnk );
         C_SUB( f2k, fpk , fpnk );
-        C_MUL( tw , f2k , st->super_twiddles[k]);
+        C_MUL( tw , f2k , st->super_twiddles[k-1]);
 
         freqdata[k].r = HALF_OF(f1k.r + tw.r);
         freqdata[k].i = HALF_OF(f1k.i + tw.i);
@@ -146,7 +146,7 @@ void kiss_fftri(kiss_fftr_cfg st,const kiss_fft_cpx *freqdata,kiss_fft_scalar *t
 
         C_ADD (fek, fk, fnkc);
         C_SUB (tmp, fk, fnkc);
-        C_MUL (fok, tmp, st->super_twiddles[k]);
+        C_MUL (fok, tmp, st->super_twiddles[k-1]);
         C_ADD (st->tmpbuf[k],     fek, fok);
         C_SUB (st->tmpbuf[ncfft - k], fek, fok);
 #ifdef USE_SIMD