diff --git a/ChangeLog b/ChangeLog
index 0e17fb59..4b73a8f6 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -20,6 +20,7 @@
 - Python x.bandjoin(y) is now x.ibandjoin(y), sorry
 - oop, removed a DEBUG from buffer.c, vips is 30% faster
 - faster and lower-mem TIFF read
+- faster bilinear interpolator
 
 7/5/15 started 8.1.1
 - oop, vips-8.0 wrapper script should be vips-8.1, thanks Danilo
diff --git a/libvips/resample/interpolate.c b/libvips/resample/interpolate.c
index 1bc7fde7..c8b3ac19 100644
--- a/libvips/resample/interpolate.c
+++ b/libvips/resample/interpolate.c
@@ -429,12 +429,13 @@ G_DEFINE_TYPE( VipsInterpolateBilinear, vips_interpolate_bilinear,
  */
 
 #define BILINEAR_INT_INNER { \
-	tq[z] = (c1 * tp1[z] + c2 * tp2[z] + \
-		 c3 * tp3[z] + c4 * tp4[z]) >> VIPS_INTERPOLATE_SHIFT; \
+	tq[z] = (sc1 * tp1[z] + sc2 * tp2[z] + \
+		 sc3 * tp3[z] + sc4 * tp4[z]) >> VIPS_INTERPOLATE_SHIFT; \
 	z += 1; \
 }
 
 /* Fixed-point arithmetic, no tables.
+ */
 #define BILINEAR_INT( TYPE ) { \
 	TYPE * restrict tq = (TYPE *) out; \
 	\
@@ -443,10 +444,15 @@ G_DEFINE_TYPE( VipsInterpolateBilinear, vips_interpolate_bilinear,
         \
 	float Yd = 1.0f - Y; \
         \
-	int c4 = VIPS_INTERPOLATE_SCALE * (Y * X); \
-	int c2 = VIPS_INTERPOLATE_SCALE * (Yd * X); \
-	int c3 = VIPS_INTERPOLATE_SCALE * (Y - c4); \
-	int c1 = VIPS_INTERPOLATE_SCALE * (Yd - c2); \
+	float c4 = Y * X; \
+	float c2 = Yd * X; \
+	float c3 = Y - c4; \
+	float c1 = Yd - c2; \
+	\
+	int sc1 = VIPS_INTERPOLATE_SCALE * c1;\
+	int sc2 = VIPS_INTERPOLATE_SCALE * c2;\
+	int sc3 = VIPS_INTERPOLATE_SCALE * c3;\
+	int sc4 = VIPS_INTERPOLATE_SCALE * c4;\
  	\
 	const TYPE * restrict tp1 = (TYPE *) p1; \
 	const TYPE * restrict tp2 = (TYPE *) p2; \
@@ -456,32 +462,6 @@ G_DEFINE_TYPE( VipsInterpolateBilinear, vips_interpolate_bilinear,
 	z = 0; \
 	VIPS_UNROLL( b, BILINEAR_INT_INNER ); \
 }
- */
-
-/* Fixed-point arithmetic, no tables.
- */
-#define BILINEAR_INT( TYPE ) { \
-	TYPE * restrict tq = (TYPE *) out; \
-	\
-	float Y = y - iy; \
-	float X = x - ix; \
-        \
-	float Yd = 1.0f - Y; \
-        \
-	int c4 = VIPS_INTERPOLATE_SCALE * (Y * X); \
-	int c2 = VIPS_INTERPOLATE_SCALE * (Yd * X); \
-	int c3 = VIPS_INTERPOLATE_SCALE * (Y - c4); \
-	int c1 = VIPS_INTERPOLATE_SCALE * (Yd - c2); \
- 	\
-	const TYPE * restrict tp1 = (TYPE *) p1; \
-	const TYPE * restrict tp2 = (TYPE *) p2; \
-	const TYPE * restrict tp3 = (TYPE *) p3; \
-	const TYPE * restrict tp4 = (TYPE *) p4; \
-	\
-	for( z = 0; z < b; z++ ) \
-		tq[z] = (c1 * tp1[z] + c2 * tp2[z] + \
-			 c3 * tp3[z] + c4 * tp4[z]) >> VIPS_INTERPOLATE_SHIFT; \
-}
 
 #define BILINEAR_FLOAT_INNER { \
 	tq[z] = c1 * tp1[z] + c2 * tp2[z] + \
@@ -557,8 +537,36 @@ vips_interpolate_bilinear_interpolate( VipsInterpolate *interpolate,
 	g_assert( (int) x + 1 < VIPS_RECT_RIGHT( &in->valid ) );
 	g_assert( (int) y + 1 < VIPS_RECT_BOTTOM( &in->valid ) );
 
+	unsigned char * restrict tq = (unsigned char *) out; 
+	
+	float Y = y - iy; 
+	float X = x - ix; 
+        
+	float Yd = 1.0f - Y; 
+ 
+	float c4 = Y * X; 
+	float c2 = Yd * X; 
+	float c3 = Y - c4; 
+	float c1 = Yd - c2; 
+
+	int sc1 = VIPS_INTERPOLATE_SCALE * c1;
+	int sc2 = VIPS_INTERPOLATE_SCALE * c2;
+	int sc3 = VIPS_INTERPOLATE_SCALE * c3;
+	int sc4 = VIPS_INTERPOLATE_SCALE * c4;
+ 	
+	const unsigned char * restrict tp1 = (unsigned char *) p1; 
+	const unsigned char * restrict tp2 = (unsigned char *) p2; 
+	const unsigned char * restrict tp3 = (unsigned char *) p3; 
+	const unsigned char * restrict tp4 = (unsigned char *) p4; 
+	
+	for( z = 0; z < b; z++ ) 
+		tq[z] = (sc1 * tp1[z] + sc2 * tp2[z] + 
+			 sc3 * tp3[z] + sc4 * tp4[z]) >> VIPS_INTERPOLATE_SHIFT; 
+
+/*
 	SWITCH_INTERPOLATE( in->im->BandFmt,
 		BILINEAR_INT, BILINEAR_FLOAT );
+		 */
 }
 
 static void