Merge remote-tracking branch 'origin/released'

Conflicts:
	pixman/pixman/pixman-sse2.c

12 files changed, 645 insertions, 183 deletions

diff --git a/pixman/pixman/Makefile.am b/pixman/pixman/Makefile.am
index 1b232ad0f..deacf8728 100644
--- a/pixman/pixman/Makefile.am
+++ b/pixman/pixman/Makefile.am
@@ -92,6 +92,7 @@ endif
 
 # iwmmxt code
 if USE_ARM_IWMMXT
+libpixman_iwmmxt_la_SOURCES = pixman-mmx.c
 noinst_LTLIBRARIES += libpixman-iwmmxt.la
 libpixman_1_la_LIBADD += libpixman-iwmmxt.la
 
diff --git a/pixman/pixman/loongson-mmintrin.h b/pixman/pixman/loongson-mmintrin.h
index 1a114fe0f..086c6e0f1 100644
--- a/pixman/pixman/loongson-mmintrin.h
+++ b/pixman/pixman/loongson-mmintrin.h
@@ -45,6 +45,28 @@ _mm_setzero_si64 (void)
 }
 
 extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_add_pi16 (__m64 __m1, __m64 __m2)
+{
+	__m64 ret;
+	asm("paddh %0, %1, %2\n\t"
+	   : "=f" (ret)
+	   : "f" (__m1), "f" (__m2)
+	);
+	return ret;
+}
+
+extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_add_pi32 (__m64 __m1, __m64 __m2)
+{
+	__m64 ret;
+	asm("paddw %0, %1, %2\n\t"
+	   : "=f" (ret)
+	   : "f" (__m1), "f" (__m2)
+	);
+	return ret;
+}
+
+extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_adds_pu16 (__m64 __m1, __m64 __m2)
 {
 	__m64 ret;
@@ -150,6 +172,78 @@ _mm_packs_pu16 (__m64 __m1, __m64 __m2)
 }
 
 extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_packs_pi32 (__m64 __m1, __m64 __m2)
+{
+	__m64 ret;
+	asm("packsswh %0, %1, %2\n\t"
+	   : "=f" (ret)
+	   : "f" (__m1), "f" (__m2)
+	);
+	return ret;
+}
+
+#define _MM_SHUFFLE(fp3,fp2,fp1,fp0) \
+ (((fp3) << 6) | ((fp2) << 4) | ((fp1) << 2) | (fp0))
+extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_set_pi16 (uint16_t __w3, uint16_t __w2, uint16_t __w1, uint16_t __w0)
+{
+	if (__builtin_constant_p (__w3) &&
+	    __builtin_constant_p (__w2) &&
+	    __builtin_constant_p (__w1) &&
+	    __builtin_constant_p (__w0))
+	{
+		uint64_t val = ((uint64_t)__w3 << 48)
+			     | ((uint64_t)__w2 << 32)
+			     | ((uint64_t)__w1 << 16)
+			     | ((uint64_t)__w0 <<  0);
+		return *(__m64 *)&val;
+	}
+	else if (__w3 == __w2 && __w2 == __w1 && __w1 == __w0)
+	{
+		/* TODO: handle other cases */
+		uint64_t val = __w3;
+		uint64_t imm = _MM_SHUFFLE (0, 0, 0, 0);
+		__m64 ret;
+		asm("pshufh %0, %1, %2\n\t"
+		    : "=f" (ret)
+		    : "f" (*(__m64 *)&val), "f" (*(__m64 *)&imm)
+		);
+		return ret;
+	}
+	uint64_t val = ((uint64_t)__w3 << 48)
+		     | ((uint64_t)__w2 << 32)
+		     | ((uint64_t)__w1 << 16)
+		     | ((uint64_t)__w0 <<  0);
+	return *(__m64 *)&val;
+}
+
+extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_set_pi32 (unsigned __i1, unsigned __i0)
+{
+	if (__builtin_constant_p (__i1) &&
+	    __builtin_constant_p (__i0))
+	{
+		uint64_t val = ((uint64_t)__i1 << 32)
+			     | ((uint64_t)__i0 <<  0);
+		return *(__m64 *)&val;
+	}
+	else if (__i1 == __i0)
+	{
+		uint64_t imm = _MM_SHUFFLE (1, 0, 1, 0);
+		__m64 ret;
+		asm("pshufh %0, %1, %2\n\t"
+		    : "=f" (ret)
+		    : "f" (*(__m32 *)&__i1), "f" (*(__m64 *)&imm)
+		);
+		return ret;
+	}
+	uint64_t val = ((uint64_t)__i1 << 32)
+		     | ((uint64_t)__i0 <<  0);
+	return *(__m64 *)&val;
+}
+#undef _MM_SHUFFLE
+
+extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_shuffle_pi16 (__m64 __m, int64_t __n)
 {
 	__m64 ret;
@@ -193,6 +287,17 @@ _mm_srli_pi16 (__m64 __m, int64_t __count)
 }
 
 extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_srli_pi32 (__m64 __m, int64_t __count)
+{
+	__m64 ret;
+	asm("psrlw %0, %1, %2\n\t"
+	   : "=f" (ret)
+	   : "f" (__m), "f" (*(__m64 *)&__count)
+	);
+	return ret;
+}
+
+extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_srli_si64 (__m64 __m, int64_t __count)
 {
 	__m64 ret;
@@ -204,6 +309,17 @@ _mm_srli_si64 (__m64 __m, int64_t __count)
 }
 
 extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_sub_pi16 (__m64 __m1, __m64 __m2)
+{
+	__m64 ret;
+	asm("psubh %0, %1, %2\n\t"
+	   : "=f" (ret)
+	   : "f" (__m1), "f" (__m2)
+	);
+	return ret;
+}
+
+extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_unpackhi_pi8 (__m64 __m1, __m64 __m2)
 {
 	__m64 ret;
diff --git a/pixman/pixman/pixman-arm-neon-asm-bilinear.S b/pixman/pixman/pixman-arm-neon-asm-bilinear.S
index f7913adb7..e37b5c298 100644
--- a/pixman/pixman/pixman-arm-neon-asm-bilinear.S
+++ b/pixman/pixman/pixman-arm-neon-asm-bilinear.S
@@ -64,6 +64,7 @@
 .altmacro
 .p2align 2
 
+#include "pixman-private.h"
 #include "pixman-arm-neon-asm.h"
 
 /*
@@ -488,12 +489,12 @@ fname:
     vmull.u8  q1, d0, d28
     vmlal.u8  q1, d1, d29
     /* 5 cycles bubble */
-    vshll.u16 q0, d2, #8
+    vshll.u16 q0, d2, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q0, d2, d30
     vmlal.u16 q0, d3, d30
     /* 5 cycles bubble */
     bilinear_duplicate_mask mask_fmt, 1, d4
-    vshrn.u32 d0, q0, #16
+    vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
     /* 3 cycles bubble */
     vmovn.u16 d0, q0
     /* 1 cycle bubble */
@@ -514,16 +515,16 @@ fname:
                 q1, q11, d0, d1, d20, d21, d22, d23
     bilinear_load_mask mask_fmt, 2, d4
     bilinear_load_dst dst_fmt, op, 2, d18, d19, q9
-    vshll.u16 q0, d2, #8
+    vshll.u16 q0, d2, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q0, d2, d30
     vmlal.u16 q0, d3, d30
-    vshll.u16 q10, d22, #8
+    vshll.u16 q10, d22, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q10, d22, d31
     vmlal.u16 q10, d23, d31
-    vshrn.u32 d0, q0, #16
-    vshrn.u32 d1, q10, #16
+    vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d1, q10, #(2 * BILINEAR_INTERPOLATION_BITS)
     bilinear_duplicate_mask mask_fmt, 2, d4
-    vshr.u16  q15, q12, #8
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vadd.u16  q12, q12, q13
     vmovn.u16 d0, q0
     bilinear_interleave_src_dst \
@@ -544,29 +545,29 @@ fname:
                 q3, q9,  d4, d5, d16, d17, d18, d19
     pld       [TMP1, PF_OFFS]
     sub       TMP1, TMP1, STRIDE
-    vshll.u16 q0, d2, #8
+    vshll.u16 q0, d2, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q0, d2, d30
     vmlal.u16 q0, d3, d30
-    vshll.u16 q10, d22, #8
+    vshll.u16 q10, d22, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q10, d22, d31
     vmlal.u16 q10, d23, d31
-    vshr.u16  q15, q12, #8
-    vshll.u16 q2, d6, #8
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
+    vshll.u16 q2, d6, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q2, d6, d30
     vmlal.u16 q2, d7, d30
-    vshll.u16 q8, d18, #8
+    vshll.u16 q8, d18, #BILINEAR_INTERPOLATION_BITS
     bilinear_load_mask mask_fmt, 4, d22
     bilinear_load_dst dst_fmt, op, 4, d2, d3, q1
     pld       [TMP1, PF_OFFS]
     vmlsl.u16 q8, d18, d31
     vmlal.u16 q8, d19, d31
     vadd.u16  q12, q12, q13
-    vshrn.u32 d0, q0, #16
-    vshrn.u32 d1, q10, #16
-    vshrn.u32 d4, q2, #16
-    vshrn.u32 d5, q8, #16
+    vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d1, q10, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d4, q2, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d5, q8, #(2 * BILINEAR_INTERPOLATION_BITS)
     bilinear_duplicate_mask mask_fmt, 4, d22
-    vshr.u16  q15, q12, #8
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vmovn.u16 d0, q0
     vmovn.u16 d1, q2
     vadd.u16  q12, q12, q13
@@ -694,13 +695,13 @@ pixman_asm_function fname
     blt       0f
     tst       OUT, #(1 << dst_bpp_shift)
     beq       0f
-    vshr.u16  q15, q12, #8
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vadd.u16  q12, q12, q13
     bilinear_process_last_pixel
     sub       WIDTH, WIDTH, #1
 0:
     vadd.u16  q13, q13, q13
-    vshr.u16  q15, q12, #8
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vadd.u16  q12, q12, q13
 
     cmp       WIDTH, #2
@@ -921,7 +922,7 @@ pixman_asm_function fname
     vmull.u8    q10, d22, d28
     vmlal.u8    q10, d23, d29
 
-    vshll.u16   q0, d16, #8
+    vshll.u16   q0, d16, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16   q0, d16, d30
     vmlal.u16   q0, d17, d30
 
@@ -932,27 +933,27 @@ pixman_asm_function fname
     vmull.u8    q11, d16, d28
     vmlal.u8    q11, d17, d29
 
-    vshll.u16   q1, d18, #8
+    vshll.u16   q1, d18, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16   q1, d18, d31
     vmlal.u16   q1, d19, d31
-    vshr.u16    q15, q12, #8
+    vshr.u16    q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vadd.u16    q12, q12, q13
 .endm
 
 .macro bilinear_over_8888_8888_process_pixblock_tail
-    vshll.u16   q2, d20, #8
+    vshll.u16   q2, d20, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16   q2, d20, d30
     vmlal.u16   q2, d21, d30
-    vshll.u16   q3, d22, #8
+    vshll.u16   q3, d22, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16   q3, d22, d31
     vmlal.u16   q3, d23, d31
-    vshrn.u32   d0, q0, #16
-    vshrn.u32   d1, q1, #16
+    vshrn.u32   d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32   d1, q1, #(2 * BILINEAR_INTERPOLATION_BITS)
     vld1.32     {d2, d3}, [OUT, :128]
     pld         [OUT, #(prefetch_offset * 4)]
-    vshrn.u32   d4, q2, #16
-    vshr.u16    q15, q12, #8
-    vshrn.u32   d5, q3, #16
+    vshrn.u32   d4, q2, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshr.u16    q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32   d5, q3, #(2 * BILINEAR_INTERPOLATION_BITS)
     vmovn.u16   d6, q0
     vmovn.u16   d7, q2
     vuzp.8      d6, d7
@@ -975,7 +976,7 @@ pixman_asm_function fname
 .endm
 
 .macro bilinear_over_8888_8888_process_pixblock_tail_head
-                                            vshll.u16   q2, d20, #8
+                                            vshll.u16   q2, d20, #BILINEAR_INTERPOLATION_BITS
     mov         TMP1, X, asr #16
     add         X, X, UX
     add         TMP1, TOP, TMP1, asl #2
@@ -984,21 +985,21 @@ pixman_asm_function fname
     add         X, X, UX
     add         TMP2, TOP, TMP2, asl #2
                                             vmlal.u16   q2, d21, d30
-                                            vshll.u16   q3, d22, #8
+                                            vshll.u16   q3, d22, #BILINEAR_INTERPOLATION_BITS
     vld1.32     {d20}, [TMP1], STRIDE
                                             vmlsl.u16   q3, d22, d31
                                             vmlal.u16   q3, d23, d31
     vld1.32     {d21}, [TMP1]
     vmull.u8    q8, d20, d28
     vmlal.u8    q8, d21, d29
-                                            vshrn.u32   d0, q0, #16
-                                            vshrn.u32   d1, q1, #16
+                                            vshrn.u32   d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+                                            vshrn.u32   d1, q1, #(2 * BILINEAR_INTERPOLATION_BITS)
                                             vld1.32     {d2, d3}, [OUT, :128]
                                             pld         [OUT, PF_OFFS]
-                                            vshrn.u32   d4, q2, #16
-                                            vshr.u16    q15, q12, #8
+                                            vshrn.u32   d4, q2, #(2 * BILINEAR_INTERPOLATION_BITS)
+                                            vshr.u16    q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vld1.32     {d22}, [TMP2], STRIDE
-                                            vshrn.u32   d5, q3, #16
+                                            vshrn.u32   d5, q3, #(2 * BILINEAR_INTERPOLATION_BITS)
                                             vmovn.u16   d6, q0
     vld1.32     {d23}, [TMP2]
     vmull.u8    q9, d22, d28
@@ -1022,7 +1023,7 @@ pixman_asm_function fname
     vmlal.u8    q10, d23, d29
                                             vmull.u8    q11, d2, d4
                                             vmull.u8    q2, d3, d4
-    vshll.u16   q0, d16, #8
+    vshll.u16   q0, d16, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16   q0, d16, d30
                                             vrshr.u16   q1, q11, #8
     vmlal.u16   q0, d17, d30
@@ -1037,12 +1038,12 @@ pixman_asm_function fname
     vmull.u8    q11, d16, d28
     vmlal.u8    q11, d17, d29
                                             vuzp.8      d6, d7
-    vshll.u16   q1, d18, #8
+    vshll.u16   q1, d18, #BILINEAR_INTERPOLATION_BITS
                                             vuzp.8      d6, d7
     vmlsl.u16   q1, d18, d31
                                             vadd.u16    q12, q12, q13
     vmlal.u16   q1, d19, d31
-    vshr.u16    q15, q12, #8
+    vshr.u16    q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vadd.u16    q12, q12, q13
                                             vst1.32     {d6, d7}, [OUT, :128]!
 .endm
@@ -1081,14 +1082,14 @@ pixman_asm_function fname
     vmull.u8    q3, d2, d28
     vmlal.u8    q2, d1, d29
     vmlal.u8    q3, d3, d29
-    vshll.u16   q0, d4, #8
-    vshll.u16   q1, d6, #8
+    vshll.u16   q0, d4, #BILINEAR_INTERPOLATION_BITS
+    vshll.u16   q1, d6, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16   q0, d4, d30
     vmlsl.u16   q1, d6, d31
     vmlal.u16   q0, d5, d30
     vmlal.u16   q1, d7, d31
-    vshrn.u32   d0, q0, #16
-    vshrn.u32   d1, q1, #16
+    vshrn.u32   d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32   d1, q1, #(2 * BILINEAR_INTERPOLATION_BITS)
     vld1.32     {d2}, [TMP3], STRIDE
     vld1.32     {d3}, [TMP3]
     pld         [TMP4, PF_OFFS]
@@ -1099,7 +1100,7 @@ pixman_asm_function fname
     vmlal.u8    q3, d3, d29
     vmull.u8    q1, d4, d28
     vmlal.u8    q1, d5, d29
-    vshr.u16    q15, q12, #8
+    vshr.u16    q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vld1.32     {d22[0]}, [MASK]!
     pld         [MASK, #prefetch_offset]
     vadd.u16    q12, q12, q13
@@ -1107,17 +1108,17 @@ pixman_asm_function fname
 .endm
 
 .macro bilinear_over_8888_8_8888_process_pixblock_tail
-    vshll.u16   q9, d6, #8
-    vshll.u16   q10, d2, #8
+    vshll.u16   q9, d6, #BILINEAR_INTERPOLATION_BITS
+    vshll.u16   q10, d2, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16   q9, d6, d30
     vmlsl.u16   q10, d2, d31
     vmlal.u16   q9, d7, d30
     vmlal.u16   q10, d3, d31
-    vshr.u16    q15, q12, #8
+    vshr.u16    q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vadd.u16    q12, q12, q13
     vdup.32     d22, d22[0]
-    vshrn.u32   d18, q9, #16
-    vshrn.u32   d19, q10, #16
+    vshrn.u32   d18, q9, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32   d19, q10, #(2 * BILINEAR_INTERPOLATION_BITS)
     vmovn.u16   d17, q9
     vld1.32     {d18, d19}, [OUT, :128]
     pld         [OUT, PF_OFFS]
@@ -1146,11 +1147,11 @@ pixman_asm_function fname
 .endm
 
 .macro bilinear_over_8888_8_8888_process_pixblock_tail_head
-                                            vshll.u16   q9, d6, #8
+                                            vshll.u16   q9, d6, #BILINEAR_INTERPOLATION_BITS
     mov         TMP1, X, asr #16
     add         X, X, UX
     add         TMP1, TOP, TMP1, asl #2
-                                            vshll.u16   q10, d2, #8
+                                            vshll.u16   q10, d2, #BILINEAR_INTERPOLATION_BITS
     vld1.32     {d0}, [TMP1], STRIDE
     mov         TMP2, X, asr #16
     add         X, X, UX
@@ -1167,12 +1168,12 @@ pixman_asm_function fname
     mov         TMP4, X, asr #16
     add         X, X, UX
     add         TMP4, TOP, TMP4, asl #2
-                                            vshr.u16    q15, q12, #8
+                                            vshr.u16    q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
                                             vadd.u16    q12, q12, q13
     vld1.32     {d3}, [TMP2]
                                             vdup.32     d22, d22[0]
-                                            vshrn.u32   d18, q9, #16
-                                            vshrn.u32   d19, q10, #16
+                                            vshrn.u32   d18, q9, #(2 * BILINEAR_INTERPOLATION_BITS)
+                                            vshrn.u32   d19, q10, #(2 * BILINEAR_INTERPOLATION_BITS)
     vmull.u8    q2, d0, d28
     vmull.u8    q3, d2, d28
                                             vmovn.u16   d17, q9
@@ -1182,8 +1183,8 @@ pixman_asm_function fname
     vmlal.u8    q3, d3, d29
                                             vuzp.8      d16, d17
                                             vuzp.8      d18, d19
-    vshll.u16   q0, d4, #8
-    vshll.u16   q1, d6, #8
+    vshll.u16   q0, d4, #BILINEAR_INTERPOLATION_BITS
+    vshll.u16   q1, d6, #BILINEAR_INTERPOLATION_BITS
                                             vuzp.8      d16, d17
                                             vuzp.8      d18, d19
     vmlsl.u16   q0, d4, d30
@@ -1194,8 +1195,8 @@ pixman_asm_function fname
     vmlal.u16   q1, d7, d31
                                             vrsra.u16   q10, q10, #8
                                             vrsra.u16   q11, q11, #8
-    vshrn.u32   d0, q0, #16
-    vshrn.u32   d1, q1, #16
+    vshrn.u32   d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32   d1, q1, #(2 * BILINEAR_INTERPOLATION_BITS)
                                             vrshrn.u16  d16, q10, #8
                                             vrshrn.u16  d17, q11, #8
     vld1.32     {d2}, [TMP3], STRIDE
@@ -1216,7 +1217,7 @@ pixman_asm_function fname
                                             vraddhn.u16 d18, q9, q10
                                             vraddhn.u16 d19, q15, q11
     vmlal.u8    q1, d5, d29
-    vshr.u16    q15, q12, #8
+    vshr.u16    q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
                                             vqadd.u8    q9, q8, q9
     vld1.32     {d22[0]}, [MASK]!
                                             vuzp.8      d18, d19
diff --git a/pixman/pixman/pixman-arm-neon-asm.S b/pixman/pixman/pixman-arm-neon-asm.S
index 87aae1d55..187197dc3 100644
--- a/pixman/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman/pixman-arm-neon-asm.S
@@ -49,6 +49,7 @@
     .altmacro
     .p2align 2
 
+#include "pixman-private.h"
 #include "pixman-arm-neon-asm.h"
 
 /* Global configuration options and preferences */
@@ -2986,11 +2987,11 @@ fname:
     vmull.u8  q1, d0, d28
     vmlal.u8  q1, d1, d29
     /* 5 cycles bubble */
-    vshll.u16 q0, d2, #8
+    vshll.u16 q0, d2, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q0, d2, d30
     vmlal.u16 q0, d3, d30
     /* 5 cycles bubble */
-    vshrn.u32 d0, q0, #16
+    vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
     /* 3 cycles bubble */
     vmovn.u16 d0, q0
     /* 1 cycle bubble */
@@ -3000,15 +3001,15 @@ fname:
 .macro bilinear_interpolate_two_pixels src_fmt, dst_fmt
     bilinear_load_and_vertical_interpolate_two_&src_fmt \
                 q1, q11, d0, d1, d20, d21, d22, d23
-    vshll.u16 q0, d2, #8
+    vshll.u16 q0, d2, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q0, d2, d30
     vmlal.u16 q0, d3, d30
-    vshll.u16 q10, d22, #8
+    vshll.u16 q10, d22, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q10, d22, d31
     vmlal.u16 q10, d23, d31
-    vshrn.u32 d0, q0, #16
-    vshrn.u32 d1, q10, #16
-    vshr.u16  q15, q12, #8
+    vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d1, q10, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vadd.u16  q12, q12, q13
     vmovn.u16 d0, q0
     bilinear_store_&dst_fmt 2, q2, q3
@@ -3020,26 +3021,26 @@ fname:
                 q3, q9,  d4, d5, d16, d17, d18, d19
     pld       [TMP1, PF_OFFS]
     sub       TMP1, TMP1, STRIDE
-    vshll.u16 q0, d2, #8
+    vshll.u16 q0, d2, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q0, d2, d30
     vmlal.u16 q0, d3, d30
-    vshll.u16 q10, d22, #8
+    vshll.u16 q10, d22, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q10, d22, d31
     vmlal.u16 q10, d23, d31
-    vshr.u16  q15, q12, #8
-    vshll.u16 q2, d6, #8
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
+    vshll.u16 q2, d6, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q2, d6, d30
     vmlal.u16 q2, d7, d30
-    vshll.u16 q8, d18, #8
+    vshll.u16 q8, d18, #BILINEAR_INTERPOLATION_BITS
     pld       [TMP2, PF_OFFS]
     vmlsl.u16 q8, d18, d31
     vmlal.u16 q8, d19, d31
     vadd.u16  q12, q12, q13
-    vshrn.u32 d0, q0, #16
-    vshrn.u32 d1, q10, #16
-    vshrn.u32 d4, q2, #16
-    vshrn.u32 d5, q8, #16
-    vshr.u16  q15, q12, #8
+    vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d1, q10, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d4, q2, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d5, q8, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vmovn.u16 d0, q0
     vmovn.u16 d1, q2
     vadd.u16  q12, q12, q13
@@ -3158,13 +3159,13 @@ pixman_asm_function fname
     blt       0f
     tst       OUT, #(1 << dst_bpp_shift)
     beq       0f
-    vshr.u16  q15, q12, #8
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vadd.u16  q12, q12, q13
     bilinear_interpolate_last_pixel src_fmt, dst_fmt
     sub       WIDTH, WIDTH, #1
 0:
     vadd.u16  q13, q13, q13
-    vshr.u16  q15, q12, #8
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vadd.u16  q12, q12, q13
 
     cmp       WIDTH, #2
@@ -3282,7 +3283,7 @@ pixman_asm_function fname
     vmull.u8  q10, d22, d28
     vmlal.u8  q10, d23, d29
 
-    vshll.u16 q0, d16, #8
+    vshll.u16 q0, d16, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q0, d16, d30
     vmlal.u16 q0, d17, d30
 
@@ -3293,25 +3294,25 @@ pixman_asm_function fname
     vmull.u8  q11, d16, d28
     vmlal.u8  q11, d17, d29
 
-    vshll.u16 q1, d18, #8
+    vshll.u16 q1, d18, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q1, d18, d31
 .endm
 
 .macro bilinear_interpolate_four_pixels_8888_8888_tail
     vmlal.u16 q1, d19, d31
-    vshr.u16  q15, q12, #8
-    vshll.u16 q2, d20, #8
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
+    vshll.u16 q2, d20, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q2, d20, d30
     vmlal.u16 q2, d21, d30
-    vshll.u16 q3, d22, #8
+    vshll.u16 q3, d22, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q3, d22, d31
     vmlal.u16 q3, d23, d31
     vadd.u16  q12, q12, q13
-    vshrn.u32 d0, q0, #16
-    vshrn.u32 d1, q1, #16
-    vshrn.u32 d4, q2, #16
-    vshr.u16  q15, q12, #8
-    vshrn.u32 d5, q3, #16
+    vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d1, q1, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d4, q2, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d5, q3, #(2 * BILINEAR_INTERPOLATION_BITS)
     vmovn.u16 d6, q0
     vmovn.u16 d7, q2
     vadd.u16  q12, q12, q13
@@ -3326,22 +3327,22 @@ pixman_asm_function fname
     add       X, X, UX
     add       TMP2, TOP, TMP2, asl #2
         vmlal.u16 q1, d19, d31
-        vshr.u16  q15, q12, #8
-        vshll.u16 q2, d20, #8
+        vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
+        vshll.u16 q2, d20, #BILINEAR_INTERPOLATION_BITS
         vmlsl.u16 q2, d20, d30
         vmlal.u16 q2, d21, d30
-        vshll.u16 q3, d22, #8
+        vshll.u16 q3, d22, #BILINEAR_INTERPOLATION_BITS
     vld1.32   {d20}, [TMP1], STRIDE
         vmlsl.u16 q3, d22, d31
         vmlal.u16 q3, d23, d31
     vld1.32   {d21}, [TMP1]
     vmull.u8  q8, d20, d28
     vmlal.u8  q8, d21, d29
-        vshrn.u32 d0, q0, #16
-        vshrn.u32 d1, q1, #16
-        vshrn.u32 d4, q2, #16
+        vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+        vshrn.u32 d1, q1, #(2 * BILINEAR_INTERPOLATION_BITS)
+        vshrn.u32 d4, q2, #(2 * BILINEAR_INTERPOLATION_BITS)
     vld1.32   {d22}, [TMP2], STRIDE
-        vshrn.u32 d5, q3, #16
+        vshrn.u32 d5, q3, #(2 * BILINEAR_INTERPOLATION_BITS)
         vadd.u16  q12, q12, q13
     vld1.32   {d23}, [TMP2]
     vmull.u8  q9, d22, d28
@@ -3353,12 +3354,12 @@ pixman_asm_function fname
     add       TMP4, TOP, TMP4, asl #2
     vmlal.u8  q9, d23, d29
     vld1.32   {d22}, [TMP3], STRIDE
-        vshr.u16  q15, q12, #8
+        vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vld1.32   {d23}, [TMP3]
     vmull.u8  q10, d22, d28
     vmlal.u8  q10, d23, d29
         vmovn.u16 d6, q0
-    vshll.u16 q0, d16, #8
+    vshll.u16 q0, d16, #BILINEAR_INTERPOLATION_BITS
         vmovn.u16 d7, q2
     vmlsl.u16 q0, d16, d30
     vmlal.u16 q0, d17, d30
@@ -3370,7 +3371,7 @@ pixman_asm_function fname
     vmull.u8  q11, d16, d28
     vmlal.u8  q11, d17, d29
         vst1.32   {d6, d7}, [OUT, :128]!
-    vshll.u16 q1, d18, #8
+    vshll.u16 q1, d18, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q1, d18, d31
 .endm
 
@@ -3403,7 +3404,7 @@ pixman_asm_function fname
     vld1.32   {d23}, [TMP3]
     vmull.u8  q10, d22, d28
     vmlal.u8  q10, d23, d29
-    vshll.u16 q0, d16, #8
+    vshll.u16 q0, d16, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q0, d16, d30
     vmlal.u16 q0, d17, d30
     pld       [TMP4, PF_OFFS]
@@ -3412,7 +3413,7 @@ pixman_asm_function fname
     pld       [TMP4, PF_OFFS]
     vmull.u8  q11, d16, d28
     vmlal.u8  q11, d17, d29
-    vshll.u16 q1, d18, #8
+    vshll.u16 q1, d18, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q1, d18, d31
 
     mov       TMP1, X, asr #16
@@ -3422,22 +3423,22 @@ pixman_asm_function fname
     add       X, X, UX
     add       TMP2, TOP, TMP2, asl #2
         vmlal.u16 q1, d19, d31
-        vshr.u16  q15, q12, #8
-        vshll.u16 q2, d20, #8
+        vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
+        vshll.u16 q2, d20, #BILINEAR_INTERPOLATION_BITS
         vmlsl.u16 q2, d20, d30
         vmlal.u16 q2, d21, d30
-        vshll.u16 q3, d22, #8
+        vshll.u16 q3, d22, #BILINEAR_INTERPOLATION_BITS
     vld1.32   {d20}, [TMP1], STRIDE
         vmlsl.u16 q3, d22, d31
         vmlal.u16 q3, d23, d31
     vld1.32   {d21}, [TMP1]
     vmull.u8  q8, d20, d28
     vmlal.u8  q8, d21, d29
-        vshrn.u32 d0, q0, #16
-        vshrn.u32 d1, q1, #16
-        vshrn.u32 d4, q2, #16
+        vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+        vshrn.u32 d1, q1, #(2 * BILINEAR_INTERPOLATION_BITS)
+        vshrn.u32 d4, q2, #(2 * BILINEAR_INTERPOLATION_BITS)
     vld1.32   {d22}, [TMP2], STRIDE
-        vshrn.u32 d5, q3, #16
+        vshrn.u32 d5, q3, #(2 * BILINEAR_INTERPOLATION_BITS)
         vadd.u16  q12, q12, q13
     vld1.32   {d23}, [TMP2]
     vmull.u8  q9, d22, d28
@@ -3449,12 +3450,12 @@ pixman_asm_function fname
     add       TMP4, TOP, TMP4, asl #2
     vmlal.u8  q9, d23, d29
     vld1.32   {d22}, [TMP3], STRIDE
-        vshr.u16  q15, q12, #8
+        vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vld1.32   {d23}, [TMP3]
     vmull.u8  q10, d22, d28
     vmlal.u8  q10, d23, d29
         vmovn.u16 d8, q0
-    vshll.u16 q0, d16, #8
+    vshll.u16 q0, d16, #BILINEAR_INTERPOLATION_BITS
         vmovn.u16 d9, q2
     vmlsl.u16 q0, d16, d30
     vmlal.u16 q0, d17, d30
@@ -3465,25 +3466,25 @@ pixman_asm_function fname
     pld       [TMP4, PF_OFFS]
     vmull.u8  q11, d16, d28
     vmlal.u8  q11, d17, d29
-    vshll.u16 q1, d18, #8
+    vshll.u16 q1, d18, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q1, d18, d31
 .endm
 
 .macro bilinear_interpolate_eight_pixels_8888_0565_tail
     vmlal.u16 q1, d19, d31
-    vshr.u16  q15, q12, #8
-    vshll.u16 q2, d20, #8
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
+    vshll.u16 q2, d20, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q2, d20, d30
     vmlal.u16 q2, d21, d30
-    vshll.u16 q3, d22, #8
+    vshll.u16 q3, d22, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q3, d22, d31
     vmlal.u16 q3, d23, d31
     vadd.u16  q12, q12, q13
-    vshrn.u32 d0, q0, #16
-    vshrn.u32 d1, q1, #16
-    vshrn.u32 d4, q2, #16
-    vshr.u16  q15, q12, #8
-    vshrn.u32 d5, q3, #16
+    vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d1, q1, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d4, q2, #(2 * BILINEAR_INTERPOLATION_BITS)
+    vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
+    vshrn.u32 d5, q3, #(2 * BILINEAR_INTERPOLATION_BITS)
     vmovn.u16 d10, q0
     vmovn.u16 d11, q2
     vadd.u16  q12, q12, q13
@@ -3508,23 +3509,23 @@ pixman_asm_function fname
     add       X, X, UX
     add       TMP2, TOP, TMP2, asl #2
         vmlal.u16 q1, d19, d31
-        vshr.u16  q15, q12, #8
+        vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
             vuzp.u8 d8, d9
-        vshll.u16 q2, d20, #8
+        vshll.u16 q2, d20, #BILINEAR_INTERPOLATION_BITS
         vmlsl.u16 q2, d20, d30
         vmlal.u16 q2, d21, d30
-        vshll.u16 q3, d22, #8
+        vshll.u16 q3, d22, #BILINEAR_INTERPOLATION_BITS
     vld1.32   {d20}, [TMP1], STRIDE
         vmlsl.u16 q3, d22, d31
         vmlal.u16 q3, d23, d31
     vld1.32   {d21}, [TMP1]
     vmull.u8  q8, d20, d28
     vmlal.u8  q8, d21, d29
-        vshrn.u32 d0, q0, #16
-        vshrn.u32 d1, q1, #16
-        vshrn.u32 d4, q2, #16
+        vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
+        vshrn.u32 d1, q1, #(2 * BILINEAR_INTERPOLATION_BITS)
+        vshrn.u32 d4, q2, #(2 * BILINEAR_INTERPOLATION_BITS)
     vld1.32   {d22}, [TMP2], STRIDE
-        vshrn.u32 d5, q3, #16
+        vshrn.u32 d5, q3, #(2 * BILINEAR_INTERPOLATION_BITS)
         vadd.u16  q12, q12, q13
     vld1.32   {d23}, [TMP2]
     vmull.u8  q9, d22, d28
@@ -3536,12 +3537,12 @@ pixman_asm_function fname
     add       TMP4, TOP, TMP4, asl #2
     vmlal.u8  q9, d23, d29
     vld1.32   {d22}, [TMP3], STRIDE
-        vshr.u16  q15, q12, #8
+        vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vld1.32   {d23}, [TMP3]
     vmull.u8  q10, d22, d28
     vmlal.u8  q10, d23, d29
         vmovn.u16 d10, q0
-    vshll.u16 q0, d16, #8
+    vshll.u16 q0, d16, #BILINEAR_INTERPOLATION_BITS
         vmovn.u16 d11, q2
     vmlsl.u16 q0, d16, d30
     vmlal.u16 q0, d17, d30
@@ -3553,7 +3554,7 @@ pixman_asm_function fname
     vmull.u8  q11, d16, d28
     vmlal.u8  q11, d17, d29
             vuzp.u8 d10, d11
-    vshll.u16 q1, d18, #8
+    vshll.u16 q1, d18, #BILINEAR_INTERPOLATION_BITS
     vmlsl.u16 q1, d18, d31
 
     mov       TMP1, X, asr #16
@@ -3564,12 +3565,12 @@ pixman_asm_function fname
     add       TMP2, TOP, TMP2, asl #2
         vmlal.u16 q1, d19, d31
             vuzp.u8 d9, d11
-        vshr.u16  q15, q12, #8
-        vshll.u16 q2, d20, #8
+        vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
+        vshll.u16 q2, d20, #BILINEAR_INTERPOLATION_BITS
             vuzp.u8 d8, d10
         vmlsl.u16 q2, d20, d30
         vmlal.u16 q2, d21, d30
-        vshll.u16 q3, d22, #8
+        vshll.u16 q3, d22, #BILINEAR_INTERPOLATION_BITS
     vld1.32   {d20}, [TMP1], STRIDE
         vmlsl.u16 q3, d22, d31
         vmlal.u16 q3, d23, d31
@@ -3579,13 +3580,13 @@ pixman_asm_function fname
             vshll.u8  q6, d9, #8
             vshll.u8  q5, d10, #8
             vshll.u8  q7, d8, #8
-        vshrn.u32 d0, q0, #16
+        vshrn.u32 d0, q0, #(2 * BILINEAR_INTERPOLATION_BITS)
             vsri.u16  q5, q6, #5
-        vshrn.u32 d1, q1, #16
+        vshrn.u32 d1, q1, #(2 * BILINEAR_INTERPOLATION_BITS)
             vsri.u16  q5, q7, #11
-        vshrn.u32 d4, q2, #16
+        vshrn.u32 d4, q2, #(2 * BILINEAR_INTERPOLATION_BITS)
     vld1.32   {d22}, [TMP2], STRIDE
-        vshrn.u32 d5, q3, #16
+        vshrn.u32 d5, q3, #(2 * BILINEAR_INTERPOLATION_BITS)
         vadd.u16  q12, q12, q13
     vld1.32   {d23}, [TMP2]
     vmull.u8  q9, d22, d28
@@ -3597,12 +3598,12 @@ pixman_asm_function fname
     add       TMP4, TOP, TMP4, asl #2
     vmlal.u8  q9, d23, d29
     vld1.32   {d22}, [TMP3], STRIDE
-        vshr.u16  q15, q12, #8
+        vshr.u16  q15, q12, #(16 - BILINEAR_INTERPOLATION_BITS)
     vld1.32   {d23}, [TMP3]
     vmull.u8  q10, d22, d28
     vmlal.u8  q10, d23, d29
         vmovn.u16 d8, q0
-    vshll.u16 q0, d16, #8
+    vshll.u16 q0, d16, #BILINEAR_INTERPOLATION_BITS
         vmovn.u16 d9, q2
     vmlsl.u16 q0, d16, d30
     vmlal.u16 q0, d17, d30
@@ -3613,7 +3614,7 @@ pixman_asm_function fname
     pld       [TMP4, PF_OFFS]
     vmull.u8  q11, d16, d28
     vmlal.u8  q11, d17, d29
-    vshll.u16 q1, d18, #8
+    vshll.u16 q1, d18, #BILINEAR_INTERPOLATION_BITS
             vst1.32   {d10, d11}, [OUT, :128]!
     vmlsl.u16 q1, d18, d31
 .endm
diff --git a/pixman/pixman/pixman-bits-image.c b/pixman/pixman/pixman-bits-image.c
index 05eab9634..b6c8630f4 100644
--- a/pixman/pixman/pixman-bits-image.c
+++ b/pixman/pixman/pixman-bits-image.c
@@ -131,8 +131,8 @@ bits_image_fetch_pixel_bilinear (bits_image_t   *image,
     x1 = x - pixman_fixed_1 / 2;
     y1 = y - pixman_fixed_1 / 2;
 
-    distx = (x1 >> 8) & 0xff;
-    disty = (y1 >> 8) & 0xff;
+    distx = pixman_fixed_to_bilinear_weight (x1);
+    disty = pixman_fixed_to_bilinear_weight (y1);
 
     x1 = pixman_fixed_to_int (x1);
     y1 = pixman_fixed_to_int (y1);
@@ -200,7 +200,7 @@ bits_image_fetch_bilinear_no_repeat_8888 (pixman_iter_t *iter,
     x = x_top = x_bottom = v.vector[0] - pixman_fixed_1/2;
 
     y = v.vector[1] - pixman_fixed_1/2;
-    disty = (y >> 8) & 0xff;
+    disty = pixman_fixed_to_bilinear_weight (y);
 
     /* Load the pointers to the first and second lines from the source
      * image that bilinear code must read.
@@ -309,7 +309,7 @@ bits_image_fetch_bilinear_no_repeat_8888 (pixman_iter_t *iter,
 	tr = top_row[pixman_fixed_to_int (x_top) + 1] | top_mask;
 	br = bottom_row[pixman_fixed_to_int (x_bottom) + 1] | bottom_mask;
 
-	distx = (x >> 8) & 0xff;
+	distx = pixman_fixed_to_bilinear_weight (x);
 
 	*buffer++ = bilinear_interpolation (0, tr, 0, br, distx, disty);
 
@@ -334,7 +334,7 @@ bits_image_fetch_bilinear_no_repeat_8888 (pixman_iter_t *iter,
 	    bl = bottom_row [pixman_fixed_to_int (x_bottom)] | bottom_mask;
 	    br = bottom_row [pixman_fixed_to_int (x_bottom) + 1] | bottom_mask;
 
-	    distx = (x >> 8) & 0xff;
+	    distx = pixman_fixed_to_bilinear_weight (x);
 
 	    *buffer = bilinear_interpolation (tl, tr, bl, br, distx, disty);
 	}
@@ -358,7 +358,7 @@ bits_image_fetch_bilinear_no_repeat_8888 (pixman_iter_t *iter,
 	    tl = top_row [pixman_fixed_to_int (x_top)] | top_mask;
 	    bl = bottom_row [pixman_fixed_to_int (x_bottom)] | bottom_mask;
 
-	    distx = (x >> 8) & 0xff;
+	    distx = pixman_fixed_to_bilinear_weight (x);
 
 	    *buffer = bilinear_interpolation (tl, 0, bl, 0, distx, disty);
 	}
@@ -695,8 +695,8 @@ bits_image_fetch_bilinear_affine (pixman_image_t * image,
 	x1 = x - pixman_fixed_1 / 2;
 	y1 = y - pixman_fixed_1 / 2;
 
-	distx = (x1 >> 8) & 0xff;
-	disty = (y1 >> 8) & 0xff;
+	distx = pixman_fixed_to_bilinear_weight (x1);
+	disty = pixman_fixed_to_bilinear_weight (y1);
 
 	y1 = pixman_fixed_to_int (y1);
 	y2 = y1 + 1;
diff --git a/pixman/pixman/pixman-inlines.h b/pixman/pixman/pixman-inlines.h
index 3532867a4..5517de5a5 100644
--- a/pixman/pixman/pixman-inlines.h
+++ b/pixman/pixman/pixman-inlines.h
@@ -81,6 +81,13 @@ repeat (pixman_repeat_t repeat, int *c, int size)
     return TRUE;
 }
 
+static force_inline int
+pixman_fixed_to_bilinear_weight (pixman_fixed_t x)
+{
+    return (x >> (16 - BILINEAR_INTERPOLATION_BITS)) &
+	   ((1 << BILINEAR_INTERPOLATION_BITS) - 1);
+}
+
 #if SIZEOF_LONG > 4
 
 static force_inline uint32_t
@@ -92,6 +99,9 @@ bilinear_interpolation (uint32_t tl, uint32_t tr,
     uint64_t tl64, tr64, bl64, br64;
     uint64_t f, r;
 
+    distx <<= (8 - BILINEAR_INTERPOLATION_BITS);
+    disty <<= (8 - BILINEAR_INTERPOLATION_BITS);
+
     distxy = distx * disty;
     distxiy = distx * (256 - disty);
     distixy = (256 - distx) * disty;
@@ -135,6 +145,9 @@ bilinear_interpolation (uint32_t tl, uint32_t tr,
     int distxy, distxiy, distixy, distixiy;
     uint32_t f, r;
 
+    distx <<= (8 - BILINEAR_INTERPOLATION_BITS);
+    disty <<= (8 - BILINEAR_INTERPOLATION_BITS);
+
     distxy = distx * disty;
     distxiy = (distx << 8) - distxy;	/* distx * (256 - disty) */
     distixy = (disty << 8) - distxy;	/* disty * (256 - distx) */
@@ -758,12 +771,14 @@ bilinear_pad_repeat_get_scanline_bounds (int32_t         source_image_width,
  *                        all source pixels are fetched from zero padding
  *                        zone for NONE repeat
  *
- * Note: normally the sum of 'weight_top' and 'weight_bottom' is equal to 256,
- *       but sometimes it may be less than that for NONE repeat when handling
- *       fuzzy antialiased top or bottom image edges. Also both top and
- *       bottom weight variables are guaranteed to have value in 0-255
- *       range and can fit into unsigned byte or be used with 8-bit SIMD
- *       multiplication instructions.
+ * Note: normally the sum of 'weight_top' and 'weight_bottom' is equal to
+ *       BILINEAR_INTERPOLATION_RANGE, but sometimes it may be less than that
+ *       for NONE repeat when handling fuzzy antialiased top or bottom image
+ *       edges. Also both top and bottom weight variables are guaranteed to
+ *       have value, which is less than BILINEAR_INTERPOLATION_RANGE.
+ *       For example, the weights can fit into unsigned byte or be used
+ *       with 8-bit SIMD multiplication instructions for 8-bit interpolation
+ *       precision.
  */
 #define FAST_BILINEAR_MAINLOOP_INT(scale_func_name, scanline_func, src_type_t, mask_type_t,	\
 				  dst_type_t, repeat_mode, flags)				\
@@ -877,18 +892,18 @@ fast_composite_scaled_bilinear ## scale_func_name (pixman_implementation_t *imp,
 	}											\
 												\
 	y1 = pixman_fixed_to_int (vy);								\
-	weight2 = (vy >> 8) & 0xff;								\
+	weight2 = pixman_fixed_to_bilinear_weight (vy);						\
 	if (weight2)										\
 	{											\
-	    /* normal case, both row weights are in 0-255 range and fit unsigned byte */	\
+	    /* both weight1 and weight2 are smaller than BILINEAR_INTERPOLATION_RANGE */	\
 	    y2 = y1 + 1;									\
-	    weight1 = 256 - weight2;								\
+	    weight1 = BILINEAR_INTERPOLATION_RANGE - weight2;					\
 	}											\
 	else											\
 	{											\
-	    /* set both top and bottom row to the same scanline, and weights to 128+128 */	\
+	    /* set both top and bottom row to the same scanline and tweak weights */		\
 	    y2 = y1;										\
-	    weight1 = weight2 = 128;								\
+	    weight1 = weight2 = BILINEAR_INTERPOLATION_RANGE / 2;				\
 	}											\
 	vy += unit_y;										\
 	if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_PAD)					\
diff --git a/pixman/pixman/pixman-mips-dspr2-asm.S b/pixman/pixman/pixman-mips-dspr2-asm.S
index 87558f032..48f108ed9 100644
--- a/pixman/pixman/pixman-mips-dspr2-asm.S
+++ b/pixman/pixman/pixman-mips-dspr2-asm.S
@@ -29,6 +29,7 @@
  * Author:  Nemanja Lukic (nlukic@mips.com)
  */
 
+#include "pixman-private.h"
 #include "pixman-mips-dspr2-asm.h"
 
 LEAF_MIPS_DSPR2(pixman_fill_buff16_mips)
@@ -771,11 +772,15 @@ LEAF_MIPS_DSPR2(pixman_scaled_bilinear_scanline_8888_8_8888_OVER_asm_mips)
     lw       s1, 48(sp)        /* s1 = wb */
     lw       s2, 52(sp)        /* s2 = vx */
     lw       s3, 56(sp)        /* s3 = unit_x */
-    li       v0, 256
+    li       v0, BILINEAR_INTERPOLATION_RANGE
     li       s8, 0x00ff00ff
+
+    sll      s0, s0, (2 * (8 - BILINEAR_INTERPOLATION_BITS))
+    sll      s1, s1, (2 * (8 - BILINEAR_INTERPOLATION_BITS))
+
 0:
     andi     t4, s2, 0xffff    /* t4 = (short)vx */
-    srl      t4, t4, 8         /* t4 = vx >> 8 */
+    srl      t4, t4, (16 - BILINEAR_INTERPOLATION_BITS) /* t4 = vx >> 8 */
     subu     t5, v0, t4        /* t5 = ( 256 - (vx>>8)) */
 
     mul      s4, s0, t5        /* s4 = wt*(256-(vx>>8)) */
diff --git a/pixman/pixman/pixman-mmx.c b/pixman/pixman/pixman-mmx.c
index 7c1f4fe24..b3ef2495b 100644
--- a/pixman/pixman/pixman-mmx.c
+++ b/pixman/pixman/pixman-mmx.c
@@ -42,6 +42,7 @@
 #endif
 #include "pixman-private.h"
 #include "pixman-combine32.h"
+#include "pixman-inlines.h"
 
 #define no_vERBOSE
 
@@ -718,6 +719,24 @@ combine (const uint32_t *src, const uint32_t *mask)
     return vsrc;
 }
 
+static force_inline __m64
+core_combine_over_u_pixel_mmx (__m64 vsrc, __m64 vdst)
+{
+    vsrc = _mm_unpacklo_pi8 (vsrc, _mm_setzero_si64 ());
+
+    if (is_opaque (vsrc))
+    {
+	return vsrc;
+    }
+    else if (!is_zero (vsrc))
+    {
+	return over (vsrc, expand_alpha (vsrc),
+		     _mm_unpacklo_pi8 (vdst, _mm_setzero_si64 ()));
+    }
+
+    return _mm_unpacklo_pi8 (vdst, _mm_setzero_si64 ());
+}
+
 static void
 mmx_combine_over_u (pixman_implementation_t *imp,
                     pixman_op_t              op,
@@ -1623,9 +1642,7 @@ mmx_composite_over_8888_n_8888 (pixman_implementation_t *imp,
     PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
 
     mask = _pixman_image_get_solid (imp, mask_image, dest_image->bits.format);
-    mask &= 0xff000000;
-    mask = mask | mask >> 8 | mask >> 16 | mask >> 24;
-    vmask = load8888 (&mask);
+    vmask = expand_alpha (load8888 (&mask));
 
     while (height--)
     {
@@ -1694,9 +1711,7 @@ mmx_composite_over_x888_n_8888 (pixman_implementation_t *imp,
     PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
     mask = _pixman_image_get_solid (imp, mask_image, dest_image->bits.format);
 
-    mask &= 0xff000000;
-    mask = mask | mask >> 8 | mask >> 16 | mask >> 24;
-    vmask = load8888 (&mask);
+    vmask = expand_alpha (load8888 (&mask));
     srca = MC (4x00ff);
 
     while (height--)
@@ -3532,6 +3547,242 @@ mmx_composite_over_reverse_n_8888 (pixman_implementation_t *imp,
     _mm_empty ();
 }
 
+#define BSHIFT ((1 << BILINEAR_INTERPOLATION_BITS))
+#define BMSK (BSHIFT - 1)
+
+#define BILINEAR_DECLARE_VARIABLES						\
+    const __m64 mm_wt = _mm_set_pi16 (wt, wt, wt, wt);				\
+    const __m64 mm_wb = _mm_set_pi16 (wb, wb, wb, wb);				\
+    const __m64 mm_BSHIFT = _mm_set_pi16 (BSHIFT, BSHIFT, BSHIFT, BSHIFT);	\
+    const __m64 mm_addc7 = _mm_set_pi16 (0, 1, 0, 1);				\
+    const __m64 mm_xorc7 = _mm_set_pi16 (0, BMSK, 0, BMSK);			\
+    const __m64 mm_ux = _mm_set_pi16 (unit_x, unit_x, unit_x, unit_x);		\
+    const __m64 mm_zero = _mm_setzero_si64 ();					\
+    __m64 mm_x = _mm_set_pi16 (vx, vx, vx, vx)
+
+#define BILINEAR_INTERPOLATE_ONE_PIXEL(pix)					\
+do {										\
+    /* fetch 2x2 pixel block into 2 mmx registers */				\
+    __m64 t = ldq_u ((__m64 *)&src_top [pixman_fixed_to_int (vx)]);		\
+    __m64 b = ldq_u ((__m64 *)&src_bottom [pixman_fixed_to_int (vx)]);		\
+    vx += unit_x;								\
+    /* vertical interpolation */						\
+    __m64 t_hi = _mm_mullo_pi16 (_mm_unpackhi_pi8 (t, mm_zero), mm_wt);		\
+    __m64 t_lo = _mm_mullo_pi16 (_mm_unpacklo_pi8 (t, mm_zero), mm_wt);		\
+    __m64 b_hi = _mm_mullo_pi16 (_mm_unpackhi_pi8 (b, mm_zero), mm_wb);		\
+    __m64 b_lo = _mm_mullo_pi16 (_mm_unpacklo_pi8 (b, mm_zero), mm_wb);		\
+    __m64 hi = _mm_add_pi16 (t_hi, b_hi);					\
+    __m64 lo = _mm_add_pi16 (t_lo, b_lo);					\
+    if (BILINEAR_INTERPOLATION_BITS < 8)					\
+    {										\
+	/* calculate horizontal weights */					\
+	__m64 mm_wh = _mm_add_pi16 (mm_addc7, _mm_xor_si64 (mm_xorc7,		\
+			  _mm_srli_pi16 (mm_x,					\
+					 16 - BILINEAR_INTERPOLATION_BITS)));	\
+	mm_x = _mm_add_pi16 (mm_x, mm_ux);					\
+	/* horizontal interpolation */						\
+	__m64 p = _mm_unpacklo_pi16 (lo, hi);					\
+	__m64 q = _mm_unpackhi_pi16 (lo, hi);					\
+	lo = _mm_madd_pi16 (p, mm_wh);						\
+	hi = _mm_madd_pi16 (q, mm_wh);						\
+    }										\
+    else									\
+    {										\
+	/* calculate horizontal weights */					\
+	__m64 mm_wh_lo = _mm_sub_pi16 (mm_BSHIFT, _mm_srli_pi16 (mm_x,		\
+					16 - BILINEAR_INTERPOLATION_BITS));	\
+	__m64 mm_wh_hi = _mm_srli_pi16 (mm_x,					\
+					16 - BILINEAR_INTERPOLATION_BITS);	\
+	mm_x = _mm_add_pi16 (mm_x, mm_ux);					\
+	/* horizontal interpolation */						\
+	__m64 mm_lo_lo = _mm_mullo_pi16 (lo, mm_wh_lo);				\
+	__m64 mm_lo_hi = _mm_mullo_pi16 (hi, mm_wh_hi);				\
+	__m64 mm_hi_lo = _mm_mulhi_pu16 (lo, mm_wh_lo);				\
+	__m64 mm_hi_hi = _mm_mulhi_pu16 (hi, mm_wh_hi);				\
+	lo = _mm_add_pi32 (_mm_unpacklo_pi16 (mm_lo_lo, mm_hi_lo),		\
+			   _mm_unpacklo_pi16 (mm_lo_hi, mm_hi_hi));		\
+	hi = _mm_add_pi32 (_mm_unpackhi_pi16 (mm_lo_lo, mm_hi_lo),		\
+			   _mm_unpackhi_pi16 (mm_lo_hi, mm_hi_hi));		\
+    }										\
+    /* shift and pack the result */						\
+    hi = _mm_srli_pi32 (hi, BILINEAR_INTERPOLATION_BITS * 2);			\
+    lo = _mm_srli_pi32 (lo, BILINEAR_INTERPOLATION_BITS * 2);			\
+    lo = _mm_packs_pi32 (lo, hi);						\
+    lo = _mm_packs_pu16 (lo, lo);						\
+    pix = lo;									\
+} while (0)
+
+#define BILINEAR_SKIP_ONE_PIXEL()						\
+do {										\
+    vx += unit_x;								\
+    mm_x = _mm_add_pi16 (mm_x, mm_ux);						\
+} while(0)
+
+static force_inline void
+scaled_bilinear_scanline_mmx_8888_8888_SRC (uint32_t *       dst,
+					    const uint32_t * mask,
+					    const uint32_t * src_top,
+					    const uint32_t * src_bottom,
+					    int32_t          w,
+					    int              wt,
+					    int              wb,
+					    pixman_fixed_t   vx,
+					    pixman_fixed_t   unit_x,
+					    pixman_fixed_t   max_vx,
+					    pixman_bool_t    zero_src)
+{
+    BILINEAR_DECLARE_VARIABLES;
+    __m64 pix;
+
+    while (w--)
+    {
+	BILINEAR_INTERPOLATE_ONE_PIXEL (pix);
+	store (dst, pix);
+	dst++;
+    }
+
+    _mm_empty ();
+}
+
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8888_cover_SRC,
+			       scaled_bilinear_scanline_mmx_8888_8888_SRC,
+			       uint32_t, uint32_t, uint32_t,
+			       COVER, FLAG_NONE)
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8888_pad_SRC,
+			       scaled_bilinear_scanline_mmx_8888_8888_SRC,
+			       uint32_t, uint32_t, uint32_t,
+			       PAD, FLAG_NONE)
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8888_none_SRC,
+			       scaled_bilinear_scanline_mmx_8888_8888_SRC,
+			       uint32_t, uint32_t, uint32_t,
+			       NONE, FLAG_NONE)
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8888_normal_SRC,
+			       scaled_bilinear_scanline_mmx_8888_8888_SRC,
+			       uint32_t, uint32_t, uint32_t,
+			       NORMAL, FLAG_NONE)
+
+static force_inline void
+scaled_bilinear_scanline_mmx_8888_8888_OVER (uint32_t *       dst,
+					     const uint32_t * mask,
+					     const uint32_t * src_top,
+					     const uint32_t * src_bottom,
+					     int32_t          w,
+					     int              wt,
+					     int              wb,
+					     pixman_fixed_t   vx,
+					     pixman_fixed_t   unit_x,
+					     pixman_fixed_t   max_vx,
+					     pixman_bool_t    zero_src)
+{
+    BILINEAR_DECLARE_VARIABLES;
+    __m64 pix1, pix2;
+
+    while (w)
+    {
+	BILINEAR_INTERPOLATE_ONE_PIXEL (pix1);
+
+	if (!is_zero (pix1))
+	{
+	    pix2 = load (dst);
+	    store8888 (dst, core_combine_over_u_pixel_mmx (pix1, pix2));
+	}
+
+	w--;
+	dst++;
+    }
+
+    _mm_empty ();
+}
+
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8888_cover_OVER,
+			       scaled_bilinear_scanline_mmx_8888_8888_OVER,
+			       uint32_t, uint32_t, uint32_t,
+			       COVER, FLAG_NONE)
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8888_pad_OVER,
+			       scaled_bilinear_scanline_mmx_8888_8888_OVER,
+			       uint32_t, uint32_t, uint32_t,
+			       PAD, FLAG_NONE)
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8888_none_OVER,
+			       scaled_bilinear_scanline_mmx_8888_8888_OVER,
+			       uint32_t, uint32_t, uint32_t,
+			       NONE, FLAG_NONE)
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8888_normal_OVER,
+			       scaled_bilinear_scanline_mmx_8888_8888_OVER,
+			       uint32_t, uint32_t, uint32_t,
+			       NORMAL, FLAG_NONE)
+
+static force_inline void
+scaled_bilinear_scanline_mmx_8888_8_8888_OVER (uint32_t *       dst,
+					       const uint8_t  * mask,
+					       const uint32_t * src_top,
+					       const uint32_t * src_bottom,
+					       int32_t          w,
+					       int              wt,
+					       int              wb,
+					       pixman_fixed_t   vx,
+					       pixman_fixed_t   unit_x,
+					       pixman_fixed_t   max_vx,
+					       pixman_bool_t    zero_src)
+{
+    BILINEAR_DECLARE_VARIABLES;
+    __m64 pix1, pix2;
+    uint32_t m;
+
+    while (w)
+    {
+	m = (uint32_t) *mask++;
+
+	if (m)
+	{
+	    BILINEAR_INTERPOLATE_ONE_PIXEL (pix1);
+
+	    if (m == 0xff && is_opaque (pix1))
+	    {
+		store (dst, pix1);
+	    }
+	    else
+	    {
+		__m64 ms, md, ma, msa;
+
+		pix2 = load (dst);
+		ma = expand_alpha_rev (to_m64 (m));
+		ms = _mm_unpacklo_pi8 (pix1, _mm_setzero_si64 ());
+		md = _mm_unpacklo_pi8 (pix2, _mm_setzero_si64 ());
+
+		msa = expand_alpha (ms);
+
+		store8888 (dst, (in_over (ms, msa, ma, md)));
+	    }
+	}
+	else
+	{
+	    BILINEAR_SKIP_ONE_PIXEL ();
+	}
+
+	w--;
+	dst++;
+    }
+
+    _mm_empty ();
+}
+
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8_8888_cover_OVER,
+			       scaled_bilinear_scanline_mmx_8888_8_8888_OVER,
+			       uint32_t, uint8_t, uint32_t,
+			       COVER, FLAG_HAVE_NON_SOLID_MASK)
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8_8888_pad_OVER,
+			       scaled_bilinear_scanline_mmx_8888_8_8888_OVER,
+			       uint32_t, uint8_t, uint32_t,
+			       PAD, FLAG_HAVE_NON_SOLID_MASK)
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8_8888_none_OVER,
+			       scaled_bilinear_scanline_mmx_8888_8_8888_OVER,
+			       uint32_t, uint8_t, uint32_t,
+			       NONE, FLAG_HAVE_NON_SOLID_MASK)
+FAST_BILINEAR_MAINLOOP_COMMON (mmx_8888_8_8888_normal_OVER,
+			       scaled_bilinear_scanline_mmx_8888_8_8888_OVER,
+			       uint32_t, uint8_t, uint32_t,
+			       NORMAL, FLAG_HAVE_NON_SOLID_MASK)
+
 static uint32_t *
 mmx_fetch_x8r8g8b8 (pixman_iter_t *iter, const uint32_t *mask)
 {
@@ -3787,6 +4038,23 @@ static const pixman_fast_path_t mmx_fast_paths[] =
     PIXMAN_STD_FAST_PATH    (IN,   a8,       null,     a8,       mmx_composite_in_8_8              ),
     PIXMAN_STD_FAST_PATH    (IN,   solid,    a8,       a8,       mmx_composite_in_n_8_8            ),
 
+    SIMPLE_BILINEAR_FAST_PATH (SRC, a8r8g8b8,          a8r8g8b8, mmx_8888_8888                     ),
+    SIMPLE_BILINEAR_FAST_PATH (SRC, a8r8g8b8,          x8r8g8b8, mmx_8888_8888                     ),
+    SIMPLE_BILINEAR_FAST_PATH (SRC, x8r8g8b8,          x8r8g8b8, mmx_8888_8888                     ),
+    SIMPLE_BILINEAR_FAST_PATH (SRC, a8b8g8r8,          a8b8g8r8, mmx_8888_8888                     ),
+    SIMPLE_BILINEAR_FAST_PATH (SRC, a8b8g8r8,          x8b8g8r8, mmx_8888_8888                     ),
+    SIMPLE_BILINEAR_FAST_PATH (SRC, x8b8g8r8,          x8b8g8r8, mmx_8888_8888                     ),
+
+    SIMPLE_BILINEAR_FAST_PATH (OVER, a8r8g8b8,         x8r8g8b8, mmx_8888_8888                     ),
+    SIMPLE_BILINEAR_FAST_PATH (OVER, a8b8g8r8,         x8b8g8r8, mmx_8888_8888                     ),
+    SIMPLE_BILINEAR_FAST_PATH (OVER, a8r8g8b8,         a8r8g8b8, mmx_8888_8888                     ),
+    SIMPLE_BILINEAR_FAST_PATH (OVER, a8b8g8r8,         a8b8g8r8, mmx_8888_8888                     ),
+
+    SIMPLE_BILINEAR_A8_MASK_FAST_PATH (OVER, a8r8g8b8, x8r8g8b8, mmx_8888_8_8888                   ),
+    SIMPLE_BILINEAR_A8_MASK_FAST_PATH (OVER, a8b8g8r8, x8b8g8r8, mmx_8888_8_8888                   ),
+    SIMPLE_BILINEAR_A8_MASK_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, mmx_8888_8_8888                   ),
+    SIMPLE_BILINEAR_A8_MASK_FAST_PATH (OVER, a8b8g8r8, a8b8g8r8, mmx_8888_8_8888                   ),
+
     { PIXMAN_OP_NONE },
 };
 
diff --git a/pixman/pixman/pixman-private.h b/pixman/pixman/pixman-private.h
index 72e3b4f6d..0c27798b0 100644
--- a/pixman/pixman/pixman-private.h
+++ b/pixman/pixman/pixman-private.h
@@ -1,10 +1,24 @@
+#ifndef PIXMAN_PRIVATE_H
+#define PIXMAN_PRIVATE_H
+
+/*
+ * The defines which are shared between C and assembly code
+ */
+
+/* bilinear interpolation precision (must be <= 8) */
+#define BILINEAR_INTERPOLATION_BITS 7
+#define BILINEAR_INTERPOLATION_RANGE (1 << BILINEAR_INTERPOLATION_BITS)
+
+/*
+ * C specific part
+ */
+
+#ifndef __ASSEMBLER__
+
 #ifndef PACKAGE
 #  error config.h must be included before pixman-private.h
 #endif
 
-#ifndef PIXMAN_PRIVATE_H
-#define PIXMAN_PRIVATE_H
-
 #define PIXMAN_DISABLE_DEPRECATED
 #define PIXMAN_USE_INTERNAL_API
 
@@ -1052,4 +1066,6 @@ void pixman_timer_register (pixman_timer_t *timer);
 
 #endif /* PIXMAN_TIMERS */
 
+#endif /* __ASSEMBLER__ */
+
 #endif /* PIXMAN_PRIVATE_H */
diff --git a/pixman/pixman/pixman-sse2.c b/pixman/pixman/pixman-sse2.c
index b656d17d4..ba067bc31 100644
--- a/pixman/pixman/pixman-sse2.c
+++ b/pixman/pixman/pixman-sse2.c
@@ -5364,11 +5364,15 @@ FAST_NEAREST_MAINLOOP_COMMON (sse2_8888_n_8888_none_OVER,
 			      scaled_nearest_scanline_sse2_8888_n_8888_OVER,
 			      uint32_t, uint32_t, uint32_t, NONE, TRUE, TRUE)
 
+#define BMSK ((1 << BILINEAR_INTERPOLATION_BITS) - 1)
+
 #define BILINEAR_DECLARE_VARIABLES						\
     const __m128i xmm_wt = _mm_set_epi16 (wt, wt, wt, wt, wt, wt, wt, wt);	\
     const __m128i xmm_wb = _mm_set_epi16 (wb, wb, wb, wb, wb, wb, wb, wb);	\
-    const __m128i xmm_xorc = _mm_set_epi16 (0, 0, 0, 0, 0xff, 0xff, 0xff, 0xff);\
-    const __m128i xmm_addc = _mm_set_epi16 (0, 0, 0, 0, 1, 1, 1, 1);		\
+    const __m128i xmm_xorc8 = _mm_set_epi16 (0, 0, 0, 0, BMSK, BMSK, BMSK, BMSK);\
+    const __m128i xmm_addc8 = _mm_set_epi16 (0, 0, 0, 0, 1, 1, 1, 1);		\
+    const __m128i xmm_xorc7 = _mm_set_epi16 (0, BMSK, 0, BMSK, 0, BMSK, 0, BMSK);\
+    const __m128i xmm_addc7 = _mm_set_epi16 (0, 1, 0, 1, 0, 1, 0, 1);		\
     const __m128i xmm_ux = _mm_set_epi16 (unit_x&0xffff, unit_x&0xffff, unit_x&0xffff, unit_x&0xffff,	\
 					  unit_x&0xffff, unit_x&0xffff, unit_x&0xffff, unit_x&0xffff);	\
     const __m128i xmm_zero = _mm_setzero_si128 ();				\
@@ -5388,18 +5392,30 @@ do {										\
 					xmm_wt),				\
 		       _mm_mullo_epi16 (_mm_unpacklo_epi8 (blbr, xmm_zero),	\
 					xmm_wb));				\
-    /* calculate horizontal weights */						\
-    xmm_wh = _mm_add_epi16 (xmm_addc,						\
-			    _mm_xor_si128 (xmm_xorc,				\
-					   _mm_srli_epi16 (xmm_x, 8)));		\
-    xmm_x = _mm_add_epi16 (xmm_x, xmm_ux);					\
-    /* horizontal interpolation */						\
-    xmm_lo = _mm_mullo_epi16 (a, xmm_wh);					\
-    xmm_hi = _mm_mulhi_epu16 (a, xmm_wh);					\
-    a = _mm_add_epi32 (_mm_unpacklo_epi16 (xmm_lo, xmm_hi),			\
-		       _mm_unpackhi_epi16 (xmm_lo, xmm_hi));			\
+    if (BILINEAR_INTERPOLATION_BITS < 8)					\
+    {										\
+	/* calculate horizontal weights */					\
+	xmm_wh = _mm_add_epi16 (xmm_addc7, _mm_xor_si128 (xmm_xorc7,		\
+		   _mm_srli_epi16 (xmm_x, 16 - BILINEAR_INTERPOLATION_BITS)));	\
+	xmm_x = _mm_add_epi16 (xmm_x, xmm_ux);					\
+	/* horizontal interpolation */						\
+	a = _mm_madd_epi16 (_mm_unpackhi_epi16 (_mm_shuffle_epi32 (		\
+		a, _MM_SHUFFLE (1, 0, 3, 2)), a), xmm_wh);			\
+    }										\
+    else									\
+    {										\
+	/* calculate horizontal weights */					\
+	xmm_wh = _mm_add_epi16 (xmm_addc8, _mm_xor_si128 (xmm_xorc8,		\
+		_mm_srli_epi16 (xmm_x, 16 - BILINEAR_INTERPOLATION_BITS)));	\
+	xmm_x = _mm_add_epi16 (xmm_x, xmm_ux);					\
+	/* horizontal interpolation */						\
+	xmm_lo = _mm_mullo_epi16 (a, xmm_wh);					\
+	xmm_hi = _mm_mulhi_epu16 (a, xmm_wh);					\
+	a = _mm_add_epi32 (_mm_unpacklo_epi16 (xmm_lo, xmm_hi),			\
+			   _mm_unpackhi_epi16 (xmm_lo, xmm_hi));		\
+    }										\
     /* shift and pack the result */						\
-    a = _mm_srli_epi32 (a, 16);							\
+    a = _mm_srli_epi32 (a, BILINEAR_INTERPOLATION_BITS * 2);			\
     a = _mm_packs_epi32 (a, a);							\
     a = _mm_packus_epi16 (a, a);						\
     pix = _mm_cvtsi128_si32 (a);						\
@@ -5845,6 +5861,9 @@ static const pixman_fast_path_t sse2_fast_paths[] =
     SIMPLE_BILINEAR_FAST_PATH (SRC, a8r8g8b8, a8r8g8b8, sse2_8888_8888),
     SIMPLE_BILINEAR_FAST_PATH (SRC, a8r8g8b8, x8r8g8b8, sse2_8888_8888),
     SIMPLE_BILINEAR_FAST_PATH (SRC, x8r8g8b8, x8r8g8b8, sse2_8888_8888),
+    SIMPLE_BILINEAR_FAST_PATH (SRC, a8b8g8r8, a8b8g8r8, sse2_8888_8888),
+    SIMPLE_BILINEAR_FAST_PATH (SRC, a8b8g8r8, x8b8g8r8, sse2_8888_8888),
+    SIMPLE_BILINEAR_FAST_PATH (SRC, x8b8g8r8, x8b8g8r8, sse2_8888_8888),
 
     SIMPLE_BILINEAR_FAST_PATH (OVER, a8r8g8b8, x8r8g8b8, sse2_8888_8888),
     SIMPLE_BILINEAR_FAST_PATH (OVER, a8b8g8r8, x8b8g8r8, sse2_8888_8888),
diff --git a/pixman/test/affine-test.c b/pixman/test/affine-test.c
index a4ceed3da..6827cc3a8 100644
--- a/pixman/test/affine-test.c
+++ b/pixman/test/affine-test.c
@@ -301,11 +301,21 @@ test_composite (int      testnum,
     return crc32;
 }
 
+#if BILINEAR_INTERPOLATION_BITS == 8
+#define CHECKSUM 0x1EF2175A
+#elif BILINEAR_INTERPOLATION_BITS == 7
+#define CHECKSUM 0x74050F50
+#elif BILINEAR_INTERPOLATION_BITS == 4
+#define CHECKSUM 0x4362EAE8
+#else
+#define CHECKSUM 0x00000000
+#endif
+
 int
 main (int argc, const char *argv[])
 {
     pixman_disable_out_of_bounds_workaround ();
 
-    return fuzzer_test_main ("affine", 8000000, 0x1EF2175A,
+    return fuzzer_test_main ("affine", 8000000, CHECKSUM,
 			     test_composite, argc, argv);
 }
diff --git a/pixman/test/scaling-test.c b/pixman/test/scaling-test.c
index 6f2da1432..44c4f3de4 100644
--- a/pixman/test/scaling-test.c
+++ b/pixman/test/scaling-test.c
@@ -357,11 +357,21 @@ test_composite (int      testnum,
     return crc32;
 }
 
+#if BILINEAR_INTERPOLATION_BITS == 8
+#define CHECKSUM 0x80DF1CB2
+#elif BILINEAR_INTERPOLATION_BITS == 7
+#define CHECKSUM 0x2818D5FB
+#elif BILINEAR_INTERPOLATION_BITS == 4
+#define CHECKSUM 0x387540A5
+#else
+#define CHECKSUM 0x00000000
+#endif
+
 int
 main (int argc, const char *argv[])
 {
     pixman_disable_out_of_bounds_workaround ();
 
-    return fuzzer_test_main("scaling", 8000000, 0x80DF1CB2,
+    return fuzzer_test_main("scaling", 8000000, CHECKSUM,
 			    test_composite, argc, argv);
 }