From 3b5dfee9614b6a5ed90ef8c2a3b695f0dade7ef8 Mon Sep 17 00:00:00 2001
From: Mark Borgerding <mark@borgerding.net>
Date: Sat, 9 Jul 2005 03:42:07 +0000
Subject: [PATCH] added kiss_fft_cleanup

updated Makefiles for long and simd targets
---
 Makefile       |  2 +-
 TIPS           |  4 ++++
 kiss_fft.c     | 16 +++++++++++++++-
 kiss_fft.h     |  7 +++++++
 test/Makefile  |  4 ++--
 tools/Makefile | 12 +++++++++---
 6 files changed, 38 insertions(+), 7 deletions(-)

diff --git a/Makefile b/Makefile
index ef68d11..b20a743 100644
--- a/Makefile
+++ b/Makefile
@@ -6,7 +6,7 @@ ZIPFILE=kiss_fft_v$(KFVER).zip
 
 testall:
 	# The simd and long types may or may not work on your machine 
-	export DATATYPE=simd && cd test && make test
+	#export DATATYPE=simd && cd test && make test
 	export DATATYPE=long && cd test && make test
 	
 	export DATATYPE=short && cd test && make test
diff --git a/TIPS b/TIPS
index 5007c40..c02f0c4 100644
--- a/TIPS
+++ b/TIPS
@@ -10,6 +10,10 @@ Speed:
     * If the input data has no imaginary component, use the kiss_fftr code under tools/.
       Real ffts are roughly twice as fast as complex.
 
+    * If you can rearrange your code to do 4 FFTs in parallel and you are on a recent Intel or AMD machine,
+    then you might want to experiment with the USE_SIMD code.
+    
+
 Reducing code size:
     * remove some of the butterflies. There are currently butterflies optimized for radices
         2,3,4,5.  It is worth mentioning that you can still use FFT sizes that contain 
diff --git a/kiss_fft.c b/kiss_fft.c
index 0c26f66..e4c4044 100644
--- a/kiss_fft.c
+++ b/kiss_fft.c
@@ -26,7 +26,8 @@ static size_t ntmpbuf=0;
 #define CHECKBUF(buf,nbuf,n) \
     do { \
         if ( nbuf < (size_t)(n) ) {\
-            buf = (kiss_fft_cpx*)realloc(buf,sizeof(kiss_fft_cpx)*(n)); \
+            free(buf); \
+            buf = (kiss_fft_cpx*)KISS_FFT_MALLOC(sizeof(kiss_fft_cpx)*(n)); \
             nbuf = (size_t)(n); \
         } \
    }while(0)
@@ -369,3 +370,16 @@ void kiss_fft(kiss_fft_cfg cfg,const kiss_fft_cpx *fin,kiss_fft_cpx *fout)
     kiss_fft_stride(cfg,fin,fout,1);
 }
 
+
+/* not really necessary to call, but if someone is doing in-place ffts, they may want to free the 
+   buffers from CHECKBUF
+ */ 
+void kiss_fft_cleanup(void)
+{
+    free(scratchbuf);
+    scratchbuf = NULL;
+    nscratchbuf=0;
+    free(tmpbuf);
+    tmpbuf=NULL;
+    ntmpbuf=0;
+}
diff --git a/kiss_fft.h b/kiss_fft.h
index a0e232f..4919c5c 100644
--- a/kiss_fft.h
+++ b/kiss_fft.h
@@ -99,6 +99,13 @@ void kiss_fft_stride(kiss_fft_cfg cfg,const kiss_fft_cpx *fin,kiss_fft_cpx *fout
    buffer and can be simply free()d when no longer needed*/
 #define kiss_fft_free free
 
+/*
+ Cleans up some memory that gets managed internally. Not necessary to call, but it might clean up 
+ your compiler output to call this before you exit.
+*/
+void kiss_fft_cleanup(void);
+	
+
 #ifdef __cplusplus
 } 
 #endif
diff --git a/test/Makefile b/test/Makefile
index 9fe7c3b..30f5e55 100644
--- a/test/Makefile
+++ b/test/Makefile
@@ -54,9 +54,9 @@ tools:
 	cd ../tools && make all
 
 # for x86 pentium+ machines , these flags work well
-CFLAGS=-Wall -O3 -pedantic -march=pentiumpro -ffast-math -fomit-frame-pointer   -I.. -I../tools $(WARNINGS)
+CFLAGS=-O3  -march=pentiumpro -ffast-math -fomit-frame-pointer   -I.. -I../tools $(WARNINGS)
 # If the above flags do not work, try the following
-#CFLAGS=-Wall -O3 -I.. -I../tools $(WARNINGS)
+#CFLAGS=-O3 -I.. -I../tools $(WARNINGS)
 
 $(SELFTEST): $(SELFTESTSRC) $(SRCFILES)
 	$(CC) -o $@ $(CFLAGS) $(TYPEFLAGS) -lm $+
diff --git a/tools/Makefile b/tools/Makefile
index fa3fe46..0110ee4 100644
--- a/tools/Makefile
+++ b/tools/Makefile
@@ -6,17 +6,23 @@ ifeq "$(DATATYPE)" ""
  DATATYPE=float
 endif
 
+ifeq  "$(DATATYPE)" "long"
+ TYPEFLAGS=-DFIXED_POINT=32
+endif
 
 ifeq  "$(DATATYPE)" "short"
- TYPEFLAGS=-DFIXED_POINT -Dkiss_fft_scalar=short
-else
- TYPEFLAGS=-Dkiss_fft_scalar=$(DATATYPE)
+ TYPEFLAGS=-DFIXED_POINT=16
 endif
 
 ifeq  "$(DATATYPE)" "simd"
  TYPEFLAGS=-DUSE_SIMD=1 -msse
 endif
 
+ifeq "$(TYPEFLAGS)"  ""
+ TYPEFLAGS=-Dkiss_fft_scalar=$(DATATYPE)
+endif
+
+
 ifeq  "$(DATATYPE)" "float"
  FFTUTIL=fft
  FASTFILT=fastconv