some optimizations for encoder thought up mostly by Joerg

src/Makefile

 CC = gcc
-#CFLAGS = -g -Wall -DPROFILE
-CFLAGS = -g -Wall
+#CFLAGS = -g -pg -lc -O2 -Wall -fprofile-arcs -ftest-coverage -DPROFILE
+CFLAGS = -g -pg -lc -O2 -Wall -fprofile-arcs -ftest-coverage -DUNROLL_BLOCK_8
+#CFLAGS = -g -O3 -Wall -DUNROLL_BLOCK_8
 
 OBJS = aee.o aed.o sz_compat.o
 
-.PHONY : all clean test test_sz
+.PHONY : all clean check
 
 all: libae.a
 
 test_encode: test_encode.o libae.a
-	$(CC) $(CFLAGS) -o test_encode test_encode.o -L. -lae
+	$(CC) $(CFLAGS) -o $@ $< -L. -lae
 
 test_decode: test_decode.o libae.a
-	$(CC) $(CFLAGS) -o test_decode test_decode.o -L. -lae
+	$(CC) $(CFLAGS) -o $@ $< -L. -lae
 
 test_szcomp: test_szcomp.o libae.a
-	$(CC) $(CFLAGS) -o test_szcomp test_szcomp.o -L. -lae
+	$(CC) $(CFLAGS) -o $@ $< -L. -lae
 
 libae.a: $(OBJS)
 	$(AR) $(ARFLAGS) $@ $(OBJS)
@@ -25,6 +26,7 @@ aed.o: libae.h
 aee.o: libae.h
 
 install: libae.a
+	mkdir -p ../lib ../include
 	ln -sfT ../src/szlib.h ../include/szlib.h
 	ln -sfT ../src/libae.h ../include/libae.h
 	ln -sfT ../src/libae.a ../lib/libsz.a
@@ -33,17 +35,16 @@ clean:
 	rm -f $(OBJS) test_encode.o test_decode.o \
 	test_encode test_decode libae.a \
 	test_szcomp test_szcomp.o \
-	../data/test.ae ../data/test
+	../data/test.ae ../data/test \
+	*.gcno *.gcda *.gcov gmon.out
 
-test: test_encode test_decode
+check: test_encode test_decode test_szcomp
 	./test_encode 1 1 < ../data/example_data > ../data/test.ae
 	./test_decode 1 1 < ../data/test.ae > ../data/test
 	diff ../data/test ../data/example_data
 	./test_encode 99 99 < ../data/example_data > ../data/test.ae
 	./test_decode 101 101 < ../data/test.ae > ../data/test
 	diff ../data/test ../data/example_data
-
-test_sz: test_szcomp
 	 ./test_szcomp 65536 < ../data/example_data_16 > ../data/test
 	 diff ../data/test ../data/example_data_16
 

src/aed.c

@@ -9,6 +9,8 @@
 
 #include "libae.h"
 
+#define MIN(a, b) (((a) < (b))? (a): (b))
+
 #define SAFE (strm->avail_in >= state->in_blklen        \
               && strm->avail_out >= strm->block_size)
 
@@ -118,15 +120,7 @@ static inline void u_put(ae_streamp strm, int64_t sample)
     {
         d = sample;
         x = state->last_out;
-
-        if ((x - state->xmin) < (state->xmax - x))
-        {
-            th = x - state->xmin;
-        }
-        else
-        {
-            th = state->xmax - x;
-        }
+        th = MIN(x - state->xmin, state->xmax - x);
 
         if (d <= 2*th)
         {
@@ -195,14 +189,10 @@ static inline void fast_split(ae_streamp strm)
     k = state->id - 1;
 
     if (state->ref)
-    {
         u_put(strm, u_get(strm, strm->bit_per_sample));
-    }
 
     for (i = state->ref; i < strm->block_size; i++)
-    {
         state->block[i] = u_get_fs(strm) << k;
-    }
 
     for (i = state->ref; i < strm->block_size; i++)
     {