video_chroma: added I420_ABGR32 support (mostly for opengl), some clean up as well
[vlc.git] / modules / video_chroma / i420_rgb_mmx.h
index 85aa909..3200a1f 100644 (file)
@@ -24,6 +24,8 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  *****************************************************************************/
 
+#ifdef MODULE_NAME_IS_i420_rgb_mmx
+
 /* hope these constant values are cache line aligned */
 #if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 3)
 #define USED_U64(foo) \
@@ -46,6 +48,22 @@ USED_U64(mmx_mask_f8) = 0xf8f8f8f8f8f8f8f8ULL;
 USED_U64(mmx_mask_fc) = 0xfcfcfcfcfcfcfcfcULL;
 #undef USED_U64
 
+#if defined(CAN_COMPILE_MMX)
+
+/* MMX assembly */
+#define MMX_CALL(MMX_INSTRUCTIONS)      \
+    do {                                \
+    __asm__ __volatile__(               \
+        ".p2align 3 \n\t"               \
+        MMX_INSTRUCTIONS                \
+        :                               \
+        : "r" (p_y), "r" (p_u),         \
+          "r" (p_v), "r" (p_buffer) );  \
+    } while(0)
+
+#define MMX_END __asm__ __volatile__ ( "emms" )
+
 /* Use RIP-relative code in PIC mode on amd64 */
 #if defined(__x86_64__) && defined(__PIC__)
 #   define G "(%%rip)"
@@ -60,42 +78,6 @@ pxor      %%mm4, %%mm4      # zero mm4                                      \n\
 movq       (%0), %%mm6      # Load 8 Y        Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0       \n\
 "
 
-#define SSE2_INIT_16_ALIGNED "                                              \n\
-movq        (%1), %%xmm0    # Load 8 Cb       00 00 00 00 u3 u2 u1 u0       \n\
-movq        (%2), %%xmm1    # Load 8 Cr       00 00 00 00 v3 v2 v1 v0       \n\
-pxor      %%xmm4, %%xmm4    # zero mm4                                      \n\
-movdqa      (%0), %%xmm6    # Load 16 Y       Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0       \n\
-"
-
-#define SSE2_INIT_16_UNALIGNED "                                            \n\
-movq        (%1), %%xmm0    # Load 8 Cb       00 00 00 00 u3 u2 u1 u0       \n\
-movq        (%2), %%xmm1    # Load 8 Cr       00 00 00 00 v3 v2 v1 v0       \n\
-pxor      %%xmm4, %%xmm4    # zero mm4                                      \n\
-movdqu      (%0), %%xmm6    # Load 16 Y       Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0       \n\
-prefetchnta (%3)            # Tell CPU not to cache output RGB data         \n\
-"
-
-#define MMX_INTRINSICS_INIT_16      \
-    tmp64 = *(uint32_t *)p_u;       \
-    mm0 = (__m64)tmp64;             \
-    tmp64 = *(uint32_t *)p_v;       \
-    mm1 = (__m64)tmp64;             \
-    mm4 = _mm_setzero_si64();       \
-    mm6 = (__m64)*(uint64_t *)p_y;  \
-
-#define SSE2_INTRINSICS_INIT_16_ALIGNED     \
-    xmm0 = _mm_loadl_epi64((__m128i *)p_u); \
-    xmm1 = _mm_loadl_epi64((__m128i *)p_v); \
-    xmm4 = _mm_setzero_si128();             \
-    xmm6 = _mm_load_si128((__m128i *)p_y);  \
-
-#define SSE2_INTRINSICS_INIT_16_UNALIGNED   \
-    xmm0 = _mm_loadl_epi64((__m128i *)p_u); \
-    xmm1 = _mm_loadl_epi64((__m128i *)p_v); \
-    xmm4 = _mm_setzero_si128();             \
-    xmm6 = _mm_loadu_si128((__m128i *)p_y); \
-    _mm_prefetch(p_buffer, _MM_HINT_NTA);    \
-
 #define MMX_INIT_16_GRAY "                                                  \n\
 movq      (%0), %%mm6       # Load 8 Y        Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0       \n\
 #movl      $0, (%3)         # cache preload for image                       \n\
@@ -109,43 +91,6 @@ pxor     %%mm4, %%mm4       # zero mm4                                      \n\
 movq      (%0), %%mm6       # Load 8 Y        Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0       \n\
 "
 
-#define SSE2_INIT_32_ALIGNED "                                              \n\
-movq        (%1), %%xmm0    # Load 8 Cb       00 00 00 00 u3 u2 u1 u0       \n\
-movq        (%2), %%xmm1    # Load 8 Cr       00 00 00 00 v3 v2 v1 v0       \n\
-pxor      %%xmm4, %%xmm4    # zero mm4                                      \n\
-movdqa      (%0), %%xmm6    # Load 16 Y       Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0       \n\
-"
-
-#define SSE2_INIT_32_UNALIGNED "                                            \n\
-movq        (%1), %%xmm0    # Load 8 Cb       00 00 00 00 u3 u2 u1 u0       \n\
-movq        (%2), %%xmm1    # Load 8 Cr       00 00 00 00 v3 v2 v1 v0       \n\
-pxor      %%xmm4, %%xmm4    # zero mm4                                      \n\
-movdqu      (%0), %%xmm6    # Load 16 Y       Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0       \n\
-prefetchnta (%3)            # Tell CPU not to cache output RGB data         \n\
-"
-
-#define MMX_INTRINSICS_INIT_32      \
-    tmp64 = *(uint32_t *)p_u;       \
-    mm0 = (__m64)tmp64;             \
-    *(uint16_t *)p_buffer = 0;      \
-    tmp64 = *(uint32_t *)p_v;       \
-    mm1 = (__m64)tmp64;             \
-    mm4 = _mm_setzero_si64();       \
-    mm6 = (__m64)*(uint64_t *)p_y;
-
-#define SSE2_INTRINSICS_INIT_32_ALIGNED      \
-    xmm0 = _mm_loadl_epi64((__m128i *)p_u); \
-    xmm1 = _mm_loadl_epi64((__m128i *)p_v); \
-    xmm4 = _mm_setzero_si128();             \
-    xmm6 = _mm_load_si128((__m128i *)p_y);  \
-
-#define SSE2_INTRINSICS_INIT_32_UNALIGNED    \
-    xmm0 = _mm_loadl_epi64((__m128i *)p_u); \
-    xmm1 = _mm_loadl_epi64((__m128i *)p_v); \
-    xmm4 = _mm_setzero_si128();             \
-    xmm6 = _mm_loadu_si128((__m128i *)p_y); \
-    _mm_prefetch(p_buffer, _MM_HINT_NTA); \
-
 /*
  * Do the multiply part of the conversion for even and odd pixels,
  * register usage:
@@ -181,113 +126,6 @@ pmulhw    mmx_Y_coeff"G", %%mm6 # Mul 4 Y even    00 y6 00 y4 00 y2 00 y0   \n\
 pmulhw    mmx_Y_coeff"G", %%mm7 # Mul 4 Y odd     00 y7 00 y5 00 y3 00 y1   \n\
 "
 
-#define SSE2_YUV_MUL "                                                      \n\
-# convert the chroma part                                                   \n\
-punpcklbw %%xmm4, %%xmm0        # scatter 8 Cb    00 u3 00 u2 00 u1 00 u0   \n\
-punpcklbw %%xmm4, %%xmm1        # scatter 8 Cr    00 v3 00 v2 00 v1 00 v0   \n\
-movl      $0x00800080, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to     0080 0080 ... 0080 0080   \n\
-psubsw    %%xmm5, %%xmm0        # Cb -= 128                                 \n\
-psubsw    %%xmm5, %%xmm1        # Cr -= 128                                 \n\
-psllw     $3, %%xmm0            # Promote precision                         \n\
-psllw     $3, %%xmm1            # Promote precision                         \n\
-movdqa    %%xmm0, %%xmm2        # Copy 8 Cb       00 u3 00 u2 00 u1 00 u0   \n\
-movdqa    %%xmm1, %%xmm3        # Copy 8 Cr       00 v3 00 v2 00 v1 00 v0   \n\
-movl      $0xf37df37d, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to     f37d f37d ... f37d f37d   \n\
-pmulhw    %%xmm5, %%xmm2        # Mul Cb with green coeff -> Cb green       \n\
-movl      $0xe5fce5fc, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to     e5fc e5fc ... e5fc e5fc   \n\
-pmulhw    %%xmm5, %%xmm3        # Mul Cr with green coeff -> Cr green       \n\
-movl      $0x40934093, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to     4093 4093 ... 4093 4093   \n\
-pmulhw    %%xmm5, %%xmm0        # Mul Cb -> Cblue 00 b3 00 b2 00 b1 00 b0   \n\
-movl      $0x33123312, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to     3312 3312 ... 3312 3312   \n\
-pmulhw    %%xmm5, %%xmm1        # Mul Cr -> Cred  00 r3 00 r2 00 r1 00 r0   \n\
-paddsw    %%xmm3, %%xmm2        # Cb green + Cr green -> Cgreen             \n\
-                                                                            \n\
-# convert the luma part                                                     \n\
-movl      $0x10101010, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to   1010 1010 ... 1010 1010     \n\
-psubusb   %%xmm5, %%xmm6        # Y -= 16                                   \n\
-movdqa    %%xmm6, %%xmm7        # Copy 16 Y       Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0   \n\
-movl      $0x00ff00ff, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     00ff 00ff ... 00ff 00ff   \n\
-pand      %%xmm5, %%xmm6        # get Y even      00 Y6 00 Y4 00 Y2 00 Y0   \n\
-psrlw     $8, %%xmm7            # get Y odd       00 Y7 00 Y5 00 Y3 00 Y1   \n\
-psllw     $3, %%xmm6            # Promote precision                         \n\
-psllw     $3, %%xmm7            # Promote precision                         \n\
-movl      $0x253f253f, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     253f 253f ... 253f 253f   \n\
-pmulhw    %%xmm5, %%xmm6        # Mul 8 Y even    00 y6 00 y4 00 y2 00 y0   \n\
-pmulhw    %%xmm5, %%xmm7        # Mul 8 Y odd     00 y7 00 y5 00 y3 00 y1   \n\
-"
-
-#define MMX_INTRINSICS_YUV_MUL \
-    mm0 = _mm_unpacklo_pi8(mm0, mm4); \
-    mm1 = _mm_unpacklo_pi8(mm1, mm4); \
-    mm0 = _mm_subs_pi16(mm0, (__m64)mmx_80w); \
-    mm1 = _mm_subs_pi16(mm1, (__m64)mmx_80w); \
-    mm0 = _mm_slli_pi16(mm0, 3); \
-    mm1 = _mm_slli_pi16(mm1, 3); \
-    mm2 = mm0; \
-    mm3 = mm1; \
-    mm2 = _mm_mulhi_pi16(mm2, (__m64)mmx_U_green); \
-    mm3 = _mm_mulhi_pi16(mm3, (__m64)mmx_V_green); \
-    mm0 = _mm_mulhi_pi16(mm0, (__m64)mmx_U_blue); \
-    mm1 = _mm_mulhi_pi16(mm1, (__m64)mmx_V_red); \
-    mm2 = _mm_adds_pi16(mm2, mm3); \
-    \
-    mm6 = _mm_subs_pu8(mm6, (__m64)mmx_10w); \
-    mm7 = mm6; \
-    mm6 = _mm_and_si64(mm6, (__m64)mmx_00ffw); \
-    mm7 = _mm_srli_pi16(mm7, 8); \
-    mm6 = _mm_slli_pi16(mm6, 3); \
-    mm7 = _mm_slli_pi16(mm7, 3); \
-    mm6 = _mm_mulhi_pi16(mm6, (__m64)mmx_Y_coeff); \
-    mm7 = _mm_mulhi_pi16(mm7, (__m64)mmx_Y_coeff);
-
-#define SSE2_INTRINSICS_YUV_MUL \
-    xmm0 = _mm_unpacklo_epi8(xmm0, xmm4); \
-    xmm1 = _mm_unpacklo_epi8(xmm1, xmm4); \
-    xmm5 = _mm_set1_epi32(0x00800080UL); \
-    xmm0 = _mm_subs_epi16(xmm0, xmm5); \
-    xmm1 = _mm_subs_epi16(xmm1, xmm5); \
-    xmm0 = _mm_slli_epi16(xmm0, 3); \
-    xmm1 = _mm_slli_epi16(xmm1, 3); \
-    xmm2 = xmm0; \
-    xmm3 = xmm1; \
-    xmm5 = _mm_set1_epi32(0xf37df37dUL); \
-    xmm2 = _mm_mulhi_epi16(xmm2, xmm5); \
-    xmm5 = _mm_set1_epi32(0xe5fce5fcUL); \
-    xmm3 = _mm_mulhi_epi16(xmm3, xmm5); \
-    xmm5 = _mm_set1_epi32(0x40934093UL); \
-    xmm0 = _mm_mulhi_epi16(xmm0, xmm5); \
-    xmm5 = _mm_set1_epi32(0x33123312UL); \
-    xmm1 = _mm_mulhi_epi16(xmm1, xmm5); \
-    xmm2 = _mm_adds_epi16(xmm2, xmm3); \
-    \
-    xmm5 = _mm_set1_epi32(0x10101010UL); \
-    xmm6 = _mm_subs_epu8(xmm6, xmm5); \
-    xmm7 = xmm6; \
-    xmm5 = _mm_set1_epi32(0x00ff00ffUL); \
-    xmm6 = _mm_and_si128(xmm6, xmm5); \
-    xmm7 = _mm_srli_epi16(xmm7, 8); \
-    xmm6 = _mm_slli_epi16(xmm6, 3); \
-    xmm7 = _mm_slli_epi16(xmm7, 3); \
-    xmm5 = _mm_set1_epi32(0x253f253fUL); \
-    xmm6 = _mm_mulhi_epi16(xmm6, xmm5); \
-    xmm7 = _mm_mulhi_epi16(xmm7, xmm5);
-
 /*
  * Do the addition part of the conversion for even and odd pixels,
  * register usage:
@@ -324,80 +162,6 @@ punpcklbw %%mm4, %%mm1          #                 R7 R6 R5 R4 R3 R2 R1 R0   \n\
 punpcklbw %%mm5, %%mm2          #                 G7 G6 G5 G4 G3 G2 G1 G0   \n\
 "
 
-#define SSE2_YUV_ADD "                                                      \n\
-# Do horizontal and vertical scaling                                        \n\
-movdqa    %%xmm0, %%xmm3        # Copy Cblue                                \n\
-movdqa    %%xmm1, %%xmm4        # Copy Cred                                 \n\
-movdqa    %%xmm2, %%xmm5        # Copy Cgreen                               \n\
-paddsw    %%xmm6, %%xmm0        # Y even + Cblue  00 B6 00 B4 00 B2 00 B0   \n\
-paddsw    %%xmm7, %%xmm3        # Y odd  + Cblue  00 B7 00 B5 00 B3 00 B1   \n\
-paddsw    %%xmm6, %%xmm1        # Y even + Cred   00 R6 00 R4 00 R2 00 R0   \n\
-paddsw    %%xmm7, %%xmm4        # Y odd  + Cred   00 R7 00 R5 00 R3 00 R1   \n\
-paddsw    %%xmm6, %%xmm2        # Y even + Cgreen 00 G6 00 G4 00 G2 00 G0   \n\
-paddsw    %%xmm7, %%xmm5        # Y odd  + Cgreen 00 G7 00 G5 00 G3 00 G1   \n\
-                                                                            \n\
-# Limit RGB even to 0..255                                                  \n\
-packuswb  %%xmm0, %%xmm0        # B6 B4 B2 B0 / B6 B4 B2 B0                 \n\
-packuswb  %%xmm1, %%xmm1        # R6 R4 R2 R0 / R6 R4 R2 R0                 \n\
-packuswb  %%xmm2, %%xmm2        # G6 G4 G2 G0 / G6 G4 G2 G0                 \n\
-                                                                            \n\
-# Limit RGB odd to 0..255                                                   \n\
-packuswb  %%xmm3, %%xmm3        # B7 B5 B3 B1 / B7 B5 B3 B1                 \n\
-packuswb  %%xmm4, %%xmm4        # R7 R5 R3 R1 / R7 R5 R3 R1                 \n\
-packuswb  %%xmm5, %%xmm5        # G7 G5 G3 G1 / G7 G5 G3 G1                 \n\
-                                                                            \n\
-# Interleave RGB even and odd                                               \n\
-punpcklbw %%xmm3, %%xmm0        #                 B7 B6 B5 B4 B3 B2 B1 B0   \n\
-punpcklbw %%xmm4, %%xmm1        #                 R7 R6 R5 R4 R3 R2 R1 R0   \n\
-punpcklbw %%xmm5, %%xmm2        #                 G7 G6 G5 G4 G3 G2 G1 G0   \n\
-"
-
-#define MMX_INTRINSICS_YUV_ADD \
-    mm3 = mm0; \
-    mm4 = mm1; \
-    mm5 = mm2; \
-    mm0 = _mm_adds_pi16(mm0, mm6); \
-    mm3 = _mm_adds_pi16(mm3, mm7); \
-    mm1 = _mm_adds_pi16(mm1, mm6); \
-    mm4 = _mm_adds_pi16(mm4, mm7); \
-    mm2 = _mm_adds_pi16(mm2, mm6); \
-    mm5 = _mm_adds_pi16(mm5, mm7); \
-    \
-    mm0 = _mm_packs_pu16(mm0, mm0); \
-    mm1 = _mm_packs_pu16(mm1, mm1); \
-    mm2 = _mm_packs_pu16(mm2, mm2); \
-    \
-    mm3 = _mm_packs_pu16(mm3, mm3); \
-    mm4 = _mm_packs_pu16(mm4, mm4); \
-    mm5 = _mm_packs_pu16(mm5, mm5); \
-    \
-    mm0 = _mm_unpacklo_pi8(mm0, mm3); \
-    mm1 = _mm_unpacklo_pi8(mm1, mm4); \
-    mm2 = _mm_unpacklo_pi8(mm2, mm5);
-
-#define SSE2_INTRINSICS_YUV_ADD \
-    xmm3 = xmm0; \
-    xmm4 = xmm1; \
-    xmm5 = xmm2; \
-    xmm0 = _mm_adds_epi16(xmm0, xmm6); \
-    xmm3 = _mm_adds_epi16(xmm3, xmm7); \
-    xmm1 = _mm_adds_epi16(xmm1, xmm6); \
-    xmm4 = _mm_adds_epi16(xmm4, xmm7); \
-    xmm2 = _mm_adds_epi16(xmm2, xmm6); \
-    xmm5 = _mm_adds_epi16(xmm5, xmm7); \
-    \
-    xmm0 = _mm_packus_epi16(xmm0, xmm0); \
-    xmm1 = _mm_packus_epi16(xmm1, xmm1); \
-    xmm2 = _mm_packus_epi16(xmm2, xmm2); \
-    \
-    xmm3 = _mm_packus_epi16(xmm3, xmm3); \
-    xmm4 = _mm_packus_epi16(xmm4, xmm4); \
-    xmm5 = _mm_packus_epi16(xmm5, xmm5); \
-    \
-    xmm0 = _mm_unpacklo_epi8(xmm0, xmm3); \
-    xmm1 = _mm_unpacklo_epi8(xmm1, xmm4); \
-    xmm2 = _mm_unpacklo_epi8(xmm2, xmm5);
-
 /*
  * Grayscale case, only use Y
  */
@@ -476,138 +240,6 @@ movd      4(%2), %%mm1          # Load 4 Cr       __ __ __ __ v3 v2 v1 v0   \n\
 movq      %%mm5, 8(%3)          # store pixel 4-7                           \n\
 "
 
-#define SSE2_UNPACK_15_ALIGNED "                                            \n\
-# mask unneeded bits off                                                    \n\
-movl      $0xf8f8f8f8, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     f8f8 f8f8 ... f8f8 f8f8   \n\
-pand      %%xmm5, %%xmm0        # b7b6b5b4 b3______ b7b6b5b4 b3______       \n\
-psrlw     $3,%%xmm0             # ______b7 b6b5b4b3 ______b7 b6b5b4b3       \n\
-pand      %%xmm5, %%xmm2        # g7g6g5g4 g3______ g7g6g5g4 g3______       \n\
-pand      %%xmm5, %%xmm1        # r7r6r5r4 r3______ r7r6r5r4 r3______       \n\
-psrlw     $1,%%xmm1             # __r7r6r5 r4r3____ __r7r6r5 r4r3____       \n\
-pxor      %%xmm4, %%xmm4        # zero mm4                                  \n\
-movdqa    %%xmm0, %%xmm5        # Copy B15-B0                               \n\
-movdqa    %%xmm2, %%xmm7        # Copy G15-G0                               \n\
-                                                                            \n\
-# convert rgb24 plane to rgb15 pack for pixel 0-7                           \n\
-punpcklbw %%xmm4, %%xmm2        # ________ ________ g7g6g5g4 g3______       \n\
-punpcklbw %%xmm1, %%xmm0        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
-psllw     $2,%%xmm2             # ________ ____g7g6 g5g4g3__ ________       \n\
-por       %%xmm2, %%xmm0        # r7r6r5r4 r3__g7g6 g5g4g3b7 b6b5b4b3       \n\
-movntdq   %%xmm0, (%3)          # store pixel 0-7                           \n\
-                                                                            \n\
-# convert rgb24 plane to rgb15 pack for pixel 8-15                          \n\
-punpckhbw %%xmm4, %%xmm7        # ________ ________ g7g6g5g4 g3______       \n\
-punpckhbw %%xmm1, %%xmm5        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
-psllw     $2,%%xmm7             # ________ ____g7g6 g5g4g3__ ________       \n\
-por       %%xmm7, %%xmm5        # r7r6r5r4 r3__g7g6 g5g4g3b7 b6b5b4b3       \n\
-movntdq   %%xmm5, 16(%3)        # store pixel 4-7                           \n\
-"
-
-#define SSE2_UNPACK_15_UNALIGNED "                                          \n\
-# mask unneeded bits off                                                    \n\
-movl      $0xf8f8f8f8, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     f8f8 f8f8 ... f8f8 f8f8   \n\
-pand      %%xmm5, %%xmm0        # b7b6b5b4 b3______ b7b6b5b4 b3______       \n\
-psrlw     $3,%%xmm0             # ______b7 b6b5b4b3 ______b7 b6b5b4b3       \n\
-pand      %%xmm5, %%xmm2        # g7g6g5g4 g3______ g7g6g5g4 g3______       \n\
-pand      %%xmm5, %%xmm1        # r7r6r5r4 r3______ r7r6r5r4 r3______       \n\
-psrlw     $1,%%xmm1             # __r7r6r5 r4r3____ __r7r6r5 r4r3____       \n\
-pxor      %%xmm4, %%xmm4        # zero mm4                                  \n\
-movdqa    %%xmm0, %%xmm5        # Copy B15-B0                               \n\
-movdqa    %%xmm2, %%xmm7        # Copy G15-G0                               \n\
-                                                                            \n\
-# convert rgb24 plane to rgb15 pack for pixel 0-7                           \n\
-punpcklbw %%xmm4, %%xmm2        # ________ ________ g7g6g5g4 g3______       \n\
-punpcklbw %%xmm1, %%xmm0        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
-psllw     $2,%%xmm2             # ________ ____g7g6 g5g4g3__ ________       \n\
-por       %%xmm2, %%xmm0        # r7r6r5r4 r3__g7g6 g5g4g3b7 b6b5b4b3       \n\
-movdqu    %%xmm0, (%3)          # store pixel 0-7                           \n\
-                                                                            \n\
-# convert rgb24 plane to rgb15 pack for pixel 8-15                          \n\
-punpckhbw %%xmm4, %%xmm7        # ________ ________ g7g6g5g4 g3______       \n\
-punpckhbw %%xmm1, %%xmm5        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
-psllw     $2,%%xmm7             # ________ ____g7g6 g5g4g3__ ________       \n\
-por       %%xmm7, %%xmm5        # r7r6r5r4 r3__g7g6 g5g4g3b7 b6b5b4b3       \n\
-movdqu    %%xmm5, 16(%3)        # store pixel 4-7                           \n\
-"
-
-#define MMX_INTRINSICS_UNPACK_15 \
-    mm0 = _mm_and_si64(mm0, (__m64)mmx_mask_f8); \
-    mm0 = _mm_srli_pi16(mm0, 3); \
-    mm2 = _mm_and_si64(mm2, (__m64)mmx_mask_f8); \
-    mm1 = _mm_and_si64(mm1, (__m64)mmx_mask_f8); \
-    mm1 = _mm_srli_pi16(mm1, 1); \
-    mm4 = _mm_setzero_si64(); \
-    mm5 = mm0; \
-    mm7 = mm2; \
-    \
-    mm2 = _mm_unpacklo_pi8(mm2, mm4); \
-    mm0 = _mm_unpacklo_pi8(mm0, mm1); \
-    mm2 = _mm_slli_pi16(mm2, 2); \
-    mm0 = _mm_or_si64(mm0, mm2); \
-    tmp64 = *(uint64_t *)(p_y + 8); \
-    mm6 = (__m64)tmp64; \
-    *(uint64_t *)p_buffer = (uint64_t)mm0; \
-    \
-    mm7 = _mm_unpackhi_pi8(mm7, mm4); \
-    mm5 = _mm_unpackhi_pi8(mm5, mm1); \
-    mm7 = _mm_slli_pi16(mm7, 2); \
-    tmp64 = (uint64_t)*(uint32_t *)(p_u + 4); \
-    mm0 = (__m64)tmp64; \
-    mm5 = _mm_or_si64(mm5, mm7); \
-    tmp64 = (uint64_t)*(uint32_t *)(p_v + 4); \
-    mm1 = (__m64)tmp64; \
-    *(uint64_t *)(p_buffer + 4) = (uint64_t)mm5;
-
-#define SSE2_INTRINSICS_UNPACK_15_ALIGNED \
-    xmm5 = _mm_set1_epi32(0xf8f8f8f8UL); \
-    xmm0 = _mm_and_si128(xmm0, xmm5); \
-    xmm0 = _mm_srli_epi16(xmm0, 3); \
-    xmm2 = _mm_and_si128(xmm2, xmm5); \
-    xmm1 = _mm_and_si128(xmm1, xmm5); \
-    xmm1 = _mm_srli_epi16(xmm1, 1); \
-    xmm4 = _mm_setzero_si128();          \
-    xmm5 = xmm0; \
-    xmm7 = xmm2; \
-    \
-    xmm2 = _mm_unpacklo_epi8(xmm2, xmm4); \
-    xmm0 = _mm_unpacklo_epi8(xmm0, xmm1); \
-    xmm2 = _mm_slli_epi16(xmm2, 2); \
-    xmm0 = _mm_or_si128(xmm0, xmm2); \
-    _mm_stream_si128((__m128i*)p_buffer, xmm0); \
-    \
-    xmm7 = _mm_unpackhi_epi8(xmm7, xmm4); \
-    xmm5 = _mm_unpackhi_epi8(xmm5, xmm1); \
-    xmm7 = _mm_slli_epi16(xmm7, 2); \
-    xmm5 = _mm_or_si128(xmm5, xmm7); \
-    _mm_stream_si128((__m128i*)(p_buffer+8), xmm5);
-
-#define SSE2_INTRINSICS_UNPACK_15_UNALIGNED \
-    xmm5 = _mm_set1_epi32(0xf8f8f8f8UL); \
-    xmm0 = _mm_and_si128(xmm0, xmm5); \
-    xmm0 = _mm_srli_epi16(xmm0, 3); \
-    xmm2 = _mm_and_si128(xmm2, xmm5); \
-    xmm1 = _mm_and_si128(xmm1, xmm5); \
-    xmm1 = _mm_srli_epi16(xmm1, 1); \
-    xmm4 = _mm_setzero_si128();          \
-    xmm5 = xmm0; \
-    xmm7 = xmm2; \
-    \
-    xmm2 = _mm_unpacklo_epi8(xmm2, xmm4); \
-    xmm0 = _mm_unpacklo_epi8(xmm0, xmm1); \
-    xmm2 = _mm_slli_epi16(xmm2, 2); \
-    xmm0 = _mm_or_si128(xmm0, xmm2); \
-    _mm_storeu_si128((__m128i*)p_buffer, xmm0); \
-    \
-    xmm7 = _mm_unpackhi_epi8(xmm7, xmm4); \
-    xmm5 = _mm_unpackhi_epi8(xmm5, xmm1); \
-    xmm7 = _mm_slli_epi16(xmm7, 2); \
-    xmm5 = _mm_or_si128(xmm5, xmm7); \
-    _mm_storeu_si128((__m128i*)(p_buffer+16), xmm5);
-
 /*
  * convert RGB plane to RGB 16 bits,
  * mm0 -> B, mm1 -> R, mm2 -> G,
@@ -643,36 +275,454 @@ movd      4(%2), %%mm1          # Load 4 Cr       __ __ __ __ v3 v2 v1 v0   \n\
 movq      %%mm5, 8(%3)          # store pixel 4-7                           \n\
 "
 
-#define SSE2_UNPACK_16_ALIGNED "                                            \n\
-# mask unneeded bits off                                                    \n\
-movl      $0xf8f8f8f8, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     f8f8 f8f8 ... f8f8 f8f8   \n\
-pand      %%xmm5, %%xmm0        # b7b6b5b4 b3______ b7b6b5b4 b3______       \n\
-pand      %%xmm5, %%xmm1        # r7r6r5r4 r3______ r7r6r5r4 r3______       \n\
-movl      $0xfcfcfcfc, %%eax    #                                           \n\
-movd      %%eax, %%xmm5         #                                           \n\
-pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     f8f8 f8f8 ... f8f8 f8f8   \n\
-pand      %%xmm5, %%xmm2        # g7g6g5g4 g3g2____ g7g6g5g4 g3g2____       \n\
-psrlw     $3,%%xmm0             # ______b7 b6b5b4b3 ______b7 b6b5b4b3       \n\
-pxor      %%xmm4, %%xmm4        # zero mm4                                  \n\
-movdqa    %%xmm0, %%xmm5        # Copy B15-B0                               \n\
-movdqa    %%xmm2, %%xmm7        # Copy G15-G0                               \n\
-                                                                            \n\
-# convert rgb24 plane to rgb16 pack for pixel 0-7                           \n\
-punpcklbw %%xmm4, %%xmm2        # ________ ________ g7g6g5g4 g3g2____       \n\
-punpcklbw %%xmm1, %%xmm0        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
-psllw     $3,%%xmm2             # ________ __g7g6g5 g4g3g2__ ________       \n\
-por       %%xmm2, %%xmm0        # r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3       \n\
-movntdq   %%xmm0, (%3)          # store pixel 0-7                           \n\
-                                                                            \n\
-# convert rgb24 plane to rgb16 pack for pixel 8-15                          \n\
-punpckhbw %%xmm4, %%xmm7        # ________ ________ g7g6g5g4 g3g2____       \n\
-punpckhbw %%xmm1, %%xmm5        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
-psllw     $3,%%xmm7             # ________ __g7g6g5 g4g3g2__ ________       \n\
-por       %%xmm7, %%xmm5        # r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3       \n\
-movntdq   %%xmm5, 16(%3)        # store pixel 4-7                           \n\
-"
+/*
+ * convert RGB plane to RGB packed format,
+ * mm0 -> B, mm1 -> R, mm2 -> G
+ */
+
+#define MMX_UNPACK_32_ARGB "                                                \n\
+pxor      %%mm3, %%mm3  # zero mm3                                          \n\
+movq      %%mm0, %%mm4  #                 B7 B6 B5 B4 B3 B2 B1 B0           \n\
+punpcklbw %%mm2, %%mm4  #                 G3 B3 G2 B2 G1 B1 G0 B0           \n\
+movq      %%mm1, %%mm5  #                 R7 R6 R5 R4 R3 R2 R1 R0           \n\
+punpcklbw %%mm3, %%mm5  #                 00 R3 00 R2 00 R1 00 R0           \n\
+movq      %%mm4, %%mm6  #                 G3 B3 G2 B2 G1 B1 G0 B0           \n\
+punpcklwd %%mm5, %%mm4  #                 00 R1 B1 G1 00 R0 B0 G0           \n\
+movq      %%mm4, (%3)   # Store ARGB1 ARGB0                                 \n\
+punpckhwd %%mm5, %%mm6  #                 00 R3 B3 G3 00 R2 B2 G2           \n\
+movq      %%mm6, 8(%3)  # Store ARGB3 ARGB2                                 \n\
+punpckhbw %%mm2, %%mm0  #                 G7 B7 G6 B6 G5 B5 G4 B4           \n\
+punpckhbw %%mm3, %%mm1  #                 00 R7 00 R6 00 R5 00 R4           \n\
+movq      %%mm0, %%mm5  #                 G7 B7 G6 B6 G5 B5 G4 B4           \n\
+punpcklwd %%mm1, %%mm5  #                 00 R5 B5 G5 00 R4 B4 G4           \n\
+movq      %%mm5, 16(%3) # Store ARGB5 ARGB4                                 \n\
+punpckhwd %%mm1, %%mm0  #                 00 R7 B7 G7 00 R6 B6 G6           \n\
+movq      %%mm0, 24(%3) # Store ARGB7 ARGB6                                 \n\
+"
+
+#define MMX_UNPACK_32_BGRA "                                                \n\
+pxor      %%mm3, %%mm3  # zero mm3                                          \n\
+movq      %%mm2, %%mm4  #                 G7 G6 G5 G4 G3 G2 G1 G0           \n\
+punpcklbw %%mm0, %%mm4  #                 B3 G3 B2 G2 B1 G1 B0 G0           \n\
+punpcklbw %%mm1, %%mm3  #                 R3 00 R2 00 R1 00 R0 00           \n\
+movq      %%mm3, %%mm5  #                 R3 00 R2 00 R1 00 R0 00           \n\
+punpcklwd %%mm4, %%mm3  #                 B1 G1 R1 00 B0 G0 R0 00           \n\
+movq      %%mm3, (%3)   # Store BGRA1 BGRA0                                 \n\
+punpckhwd %%mm4, %%mm5  #                 B3 G3 R3 00 B2 G2 R2 00           \n\
+movq      %%mm5, 8(%3)  # Store BGRA3 BGRA2                                 \n\
+pxor      %%mm6, %%mm6  # zero mm6                                          \n\
+punpckhbw %%mm0, %%mm2  #                 B7 G7 B6 G6 B5 G5 B4 G4           \n\
+punpckhbw %%mm1, %%mm6  #                 R7 00 R6 00 R5 00 R4 00           \n\
+movq      %%mm6, %%mm0  #                 R7 00 R6 00 R5 00 R4 00           \n\
+punpcklwd %%mm2, %%mm6  #                 B5 G5 R5 00 B4 G4 R4 00           \n\
+movq      %%mm6, 16(%3) # Store BGRA5 BGRA4                                 \n\
+punpckhwd %%mm2, %%mm0  #                 B7 G7 R7 00 B6 G6 R6 00           \n\
+movq      %%mm0, 24(%3) # Store BGRA7 BGRA6                                 \n\
+"
+
+#define MMX_UNPACK_32_ABGR "                                                \n\
+pxor      %%mm3, %%mm3  # zero mm3                                          \n\
+movq      %%mm1, %%mm4  #                 R7 R6 R5 R4 R3 R2 R1 R0           \n\
+punpcklbw %%mm2, %%mm4  #                 G3 R3 G2 R2 G1 R1 G0 R0           \n\
+movq      %%mm0, %%mm5  #                 B7 B6 B5 B4 B3 B2 B1 B0           \n\
+punpcklbw %%mm3, %%mm5  #                 00 B3 00 B2 00 B1 00 B0           \n\
+movq      %%mm4, %%mm6  #                 G3 R3 G2 R2 G1 R1 G0 R0           \n\
+punpcklwd %%mm5, %%mm4  #                 00 B1 G1 R1 00 B0 G0 R0           \n\
+movq      %%mm4, (%3)   # Store ABGR1 ABGR0                                 \n\
+punpckhwd %%mm5, %%mm6  #                 00 B3 G3 R3 00 B2 G2 R2           \n\
+movq      %%mm6, 8(%3)  # Store ABGR3 ABGR2                                 \n\
+punpckhbw %%mm2, %%mm1  #                 G7 R7 G6 R6 G5 R5 G4 R4           \n\
+punpckhbw %%mm3, %%mm0  #                 00 B7 00 B6 00 B5 00 B4           \n\
+movq      %%mm1, %%mm2  #                 G7 R7 G6 R6 G5 R5 G4 R4           \n\
+punpcklwd %%mm0, %%mm1  #                 00 B5 G5 R5 00 B4 G4 R4           \n\
+movq      %%mm1, 16(%3) # Store ABGR5 ABGR4                                 \n\
+punpckhwd %%mm0, %%mm2  #                 B7 G7 R7 00 B6 G6 R6 00           \n\
+movq      %%mm2, 24(%3) # Store ABGR7 ABGR6                                 \n\
+"
+
+#elif defined(HAVE_MMX_INTRINSICS)
+
+/* MMX intrinsics */
+
+#include <mmintrin.h>
+
+#define MMX_CALL(MMX_INSTRUCTIONS)  \
+    do {                            \
+        __m64 mm0, mm1, mm2, mm3,   \
+              mm4, mm5, mm6, mm7;   \
+        MMX_INSTRUCTIONS            \
+    } while(0)
+
+#define MMX_END _mm_empty()
+    
+#define MMX_INIT_16                     \
+    mm0 = _mm_cvtsi32_si64((int)*p_u);  \
+    mm1 = _mm_cvtsi32_si64((int)*p_v);  \
+    mm4 = _mm_setzero_si64();           \
+    mm6 = (__m64)*(uint64_t *)p_y
+
+#define MMX_INIT_32                     \
+    mm0 = _mm_cvtsi32_si64((int)*p_u);  \
+    *(uint16_t *)p_buffer = 0;          \
+    mm1 = _mm_cvtsi32_si64((int)*p_v);  \
+    mm4 = _mm_setzero_si64();           \
+    mm6 = (__m64)*(uint64_t *)p_y;
+
+#define MMX_YUV_MUL                                 \
+    mm0 = _mm_unpacklo_pi8(mm0, mm4);               \
+    mm1 = _mm_unpacklo_pi8(mm1, mm4);               \
+    mm0 = _mm_subs_pi16(mm0, (__m64)mmx_80w);       \
+    mm1 = _mm_subs_pi16(mm1, (__m64)mmx_80w);       \
+    mm0 = _mm_slli_pi16(mm0, 3);                    \
+    mm1 = _mm_slli_pi16(mm1, 3);                    \
+    mm2 = mm0;                                      \
+    mm3 = mm1;                                      \
+    mm2 = _mm_mulhi_pi16(mm2, (__m64)mmx_U_green);  \
+    mm3 = _mm_mulhi_pi16(mm3, (__m64)mmx_V_green);  \
+    mm0 = _mm_mulhi_pi16(mm0, (__m64)mmx_U_blue);   \
+    mm1 = _mm_mulhi_pi16(mm1, (__m64)mmx_V_red);    \
+    mm2 = _mm_adds_pi16(mm2, mm3);                  \
+    \
+    mm6 = _mm_subs_pu8(mm6, (__m64)mmx_10w);        \
+    mm7 = mm6;                                      \
+    mm6 = _mm_and_si64(mm6, (__m64)mmx_00ffw);      \
+    mm7 = _mm_srli_pi16(mm7, 8);                    \
+    mm6 = _mm_slli_pi16(mm6, 3);                    \
+    mm7 = _mm_slli_pi16(mm7, 3);                    \
+    mm6 = _mm_mulhi_pi16(mm6, (__m64)mmx_Y_coeff);  \
+    mm7 = _mm_mulhi_pi16(mm7, (__m64)mmx_Y_coeff);
+
+#define MMX_YUV_ADD                     \
+    mm3 = mm0;                          \
+    mm4 = mm1;                          \
+    mm5 = mm2;                          \
+    mm0 = _mm_adds_pi16(mm0, mm6);      \
+    mm3 = _mm_adds_pi16(mm3, mm7);      \
+    mm1 = _mm_adds_pi16(mm1, mm6);      \
+    mm4 = _mm_adds_pi16(mm4, mm7);      \
+    mm2 = _mm_adds_pi16(mm2, mm6);      \
+    mm5 = _mm_adds_pi16(mm5, mm7);      \
+    \
+    mm0 = _mm_packs_pu16(mm0, mm0);     \
+    mm1 = _mm_packs_pu16(mm1, mm1);     \
+    mm2 = _mm_packs_pu16(mm2, mm2);     \
+    \
+    mm3 = _mm_packs_pu16(mm3, mm3);     \
+    mm4 = _mm_packs_pu16(mm4, mm4);     \
+    mm5 = _mm_packs_pu16(mm5, mm5);     \
+    \
+    mm0 = _mm_unpacklo_pi8(mm0, mm3);   \
+    mm1 = _mm_unpacklo_pi8(mm1, mm4);   \
+    mm2 = _mm_unpacklo_pi8(mm2, mm5);
+
+#define MMX_UNPACK_15                               \
+    mm0 = _mm_and_si64(mm0, (__m64)mmx_mask_f8);    \
+    mm0 = _mm_srli_pi16(mm0, 3);                    \
+    mm2 = _mm_and_si64(mm2, (__m64)mmx_mask_f8);    \
+    mm1 = _mm_and_si64(mm1, (__m64)mmx_mask_f8);    \
+    mm1 = _mm_srli_pi16(mm1, 1);                    \
+    mm4 = _mm_setzero_si64();                       \
+    mm5 = mm0;                                      \
+    mm7 = mm2;                                      \
+    \
+    mm2 = _mm_unpacklo_pi8(mm2, mm4);               \
+    mm0 = _mm_unpacklo_pi8(mm0, mm1);               \
+    mm2 = _mm_slli_pi16(mm2, 2);                    \
+    mm0 = _mm_or_si64(mm0, mm2);                    \
+    mm6 = (__m64)*(uint64_t *)(p_y + 8);            \
+    *(uint64_t *)p_buffer = (uint64_t)mm0;          \
+    \
+    mm7 = _mm_unpackhi_pi8(mm7, mm4);               \
+    mm5 = _mm_unpackhi_pi8(mm5, mm1);               \
+    mm7 = _mm_slli_pi16(mm7, 2);                    \
+    mm0 = _mm_cvtsi32_si64((int)*(uint32_t *)(p_u + 4)); \
+    mm5 = _mm_or_si64(mm5, mm7);                    \
+    mm1 = _mm_cvtsi32_si64((int)*(uint32_t *)(p_v + 4)); \
+    *(uint64_t *)(p_buffer + 4) = (uint64_t)mm5;
+
+#define MMX_UNPACK_16                               \
+    mm0 = _mm_and_si64(mm0, (__m64)mmx_mask_f8);    \
+    mm2 = _mm_and_si64(mm2, (__m64)mmx_mask_fc);    \
+    mm1 = _mm_and_si64(mm1, (__m64)mmx_mask_f8);    \
+    mm0 = _mm_srli_pi16(mm0, 3);                    \
+    mm4 = _mm_setzero_si64();                       \
+    mm5 = mm0;                                      \
+    mm7 = mm2;                                      \
+    \
+    mm2 = _mm_unpacklo_pi8(mm2, mm4);               \
+    mm0 = _mm_unpacklo_pi8(mm0, mm1);               \
+    mm2 = _mm_slli_pi16(mm2, 3);                    \
+    mm0 = _mm_or_si64(mm0, mm2);                    \
+    mm6 = (__m64)*(uint64_t *)(p_y + 8);            \
+    *(uint64_t *)p_buffer = (uint64_t)mm0;          \
+    \
+    mm7 = _mm_unpackhi_pi8(mm7, mm4);               \
+    mm5 = _mm_unpackhi_pi8(mm5, mm1);               \
+    mm7 = _mm_slli_pi16(mm7, 3);                    \
+    mm0 = _mm_cvtsi32_si64((int)*(uint32_t *)(p_u + 4)); \
+    mm5 = _mm_or_si64(mm5, mm7);                    \
+    mm1 = _mm_cvtsi32_si64((int)*(uint32_t *)(p_v + 4)); \
+    *(uint64_t *)(p_buffer + 4) = (uint64_t)mm5;
+
+#define MMX_UNPACK_32_ARGB                      \
+    mm3 = _mm_setzero_si64();                   \
+    mm4 = mm0;                                  \
+    mm4 = _mm_unpacklo_pi8(mm4, mm2);           \
+    mm5 = mm1;                                  \
+    mm5 = _mm_unpacklo_pi8(mm5, mm3);           \
+    mm6 = mm4;                                  \
+    mm4 = _mm_unpacklo_pi16(mm4, mm5);          \
+    *(uint64_t *)p_buffer = (uint64_t)mm4;      \
+    mm6 = _mm_unpackhi_pi16(mm6, mm5);          \
+    *(uint64_t *)(p_buffer + 2) = (uint64_t)mm6;\
+    mm0 = _mm_unpackhi_pi8(mm0, mm2);           \
+    mm1 = _mm_unpackhi_pi8(mm1, mm3);           \
+    mm5 = mm0;                                  \
+    mm5 = _mm_unpacklo_pi16(mm5, mm1);          \
+    *(uint64_t *)(p_buffer + 4) = (uint64_t)mm5;\
+    mm0 = _mm_unpackhi_pi16(mm0, mm1);          \
+    *(uint64_t *)(p_buffer + 6) = (uint64_t)mm0;
+
+#define MMX_UNPACK_32_BGRA                      \
+    mm3 = _mm_setzero_si64();                   \
+    mm4 = mm2;                                  \
+    mm4 = _mm_unpacklo_pi8(mm4, mm0);           \
+    mm3 = _mm_unpacklo_pi8(mm3, mm1);           \
+    mm5 = mm3;                                  \
+    mm3 = _mm_unpacklo_pi16(mm3, mm4);          \
+    *(uint64_t *)p_buffer = (uint64_t)mm3;      \
+    mm5 = _mm_unpackhi_pi16(mm5, mm4);          \
+    *(uint64_t *)(p_buffer + 2) = (uint64_t)mm5;\
+    mm6 = _mm_setzero_si64();                   \
+    mm2 = _mm_unpackhi_pi8(mm2, mm0);           \
+    mm6 = _mm_unpackhi_pi8(mm6, mm1);           \
+    mm0 = mm6;                                  \
+    mm6 = _mm_unpacklo_pi16(mm6, mm2);          \
+    *(uint64_t *)(p_buffer + 4) = (uint64_t)mm6;\
+    mm0 = _mm_unpackhi_pi16(mm0, mm2);          \
+    *(uint64_t *)(p_buffer + 6) = (uint64_t)mm0;
+
+#define MMX_UNPACK_32_ABGR                      \
+    ;
+
+#endif
+
+#elif defined( MODULE_NAME_IS_i420_rgb_sse2 )
+
+#if defined(CAN_COMPILE_SSE2)
+
+/* SSE2 assembly */
+
+#define SSE2_CALL(SSE2_INSTRUCTIONS)    \
+    do {                                \
+    __asm__ __volatile__(               \
+        ".p2align 3 \n\t"               \
+        SSE2_INSTRUCTIONS               \
+        :                               \
+        : "r" (p_y), "r" (p_u),         \
+          "r" (p_v), "r" (p_buffer)     \
+        : "eax" );                      \
+    } while(0)
+
+#define SSE2_END  __asm__ __volatile__ ( "sfence" ::: "memory" )
+
+#define SSE2_INIT_16_ALIGNED "                                              \n\
+movq        (%1), %%xmm0    # Load 8 Cb       00 00 00 00 u3 u2 u1 u0       \n\
+movq        (%2), %%xmm1    # Load 8 Cr       00 00 00 00 v3 v2 v1 v0       \n\
+pxor      %%xmm4, %%xmm4    # zero mm4                                      \n\
+movdqa      (%0), %%xmm6    # Load 16 Y       Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0       \n\
+"
+
+#define SSE2_INIT_16_UNALIGNED "                                            \n\
+movq        (%1), %%xmm0    # Load 8 Cb       00 00 00 00 u3 u2 u1 u0       \n\
+movq        (%2), %%xmm1    # Load 8 Cr       00 00 00 00 v3 v2 v1 v0       \n\
+pxor      %%xmm4, %%xmm4    # zero mm4                                      \n\
+movdqu      (%0), %%xmm6    # Load 16 Y       Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0       \n\
+prefetchnta (%3)            # Tell CPU not to cache output RGB data         \n\
+"
+
+#define SSE2_INIT_32_ALIGNED "                                              \n\
+movq        (%1), %%xmm0    # Load 8 Cb       00 00 00 00 u3 u2 u1 u0       \n\
+movq        (%2), %%xmm1    # Load 8 Cr       00 00 00 00 v3 v2 v1 v0       \n\
+pxor      %%xmm4, %%xmm4    # zero mm4                                      \n\
+movdqa      (%0), %%xmm6    # Load 16 Y       Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0       \n\
+"
+
+#define SSE2_INIT_32_UNALIGNED "                                            \n\
+movq        (%1), %%xmm0    # Load 8 Cb       00 00 00 00 u3 u2 u1 u0       \n\
+movq        (%2), %%xmm1    # Load 8 Cr       00 00 00 00 v3 v2 v1 v0       \n\
+pxor      %%xmm4, %%xmm4    # zero mm4                                      \n\
+movdqu      (%0), %%xmm6    # Load 16 Y       Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0       \n\
+prefetchnta (%3)            # Tell CPU not to cache output RGB data         \n\
+"
+
+#define SSE2_YUV_MUL "                                                      \n\
+# convert the chroma part                                                   \n\
+punpcklbw %%xmm4, %%xmm0        # scatter 8 Cb    00 u3 00 u2 00 u1 00 u0   \n\
+punpcklbw %%xmm4, %%xmm1        # scatter 8 Cr    00 v3 00 v2 00 v1 00 v0   \n\
+movl      $0x00800080, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to     0080 0080 ... 0080 0080   \n\
+psubsw    %%xmm5, %%xmm0        # Cb -= 128                                 \n\
+psubsw    %%xmm5, %%xmm1        # Cr -= 128                                 \n\
+psllw     $3, %%xmm0            # Promote precision                         \n\
+psllw     $3, %%xmm1            # Promote precision                         \n\
+movdqa    %%xmm0, %%xmm2        # Copy 8 Cb       00 u3 00 u2 00 u1 00 u0   \n\
+movdqa    %%xmm1, %%xmm3        # Copy 8 Cr       00 v3 00 v2 00 v1 00 v0   \n\
+movl      $0xf37df37d, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to     f37d f37d ... f37d f37d   \n\
+pmulhw    %%xmm5, %%xmm2        # Mul Cb with green coeff -> Cb green       \n\
+movl      $0xe5fce5fc, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to     e5fc e5fc ... e5fc e5fc   \n\
+pmulhw    %%xmm5, %%xmm3        # Mul Cr with green coeff -> Cr green       \n\
+movl      $0x40934093, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to     4093 4093 ... 4093 4093   \n\
+pmulhw    %%xmm5, %%xmm0        # Mul Cb -> Cblue 00 b3 00 b2 00 b1 00 b0   \n\
+movl      $0x33123312, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to     3312 3312 ... 3312 3312   \n\
+pmulhw    %%xmm5, %%xmm1        # Mul Cr -> Cred  00 r3 00 r2 00 r1 00 r0   \n\
+paddsw    %%xmm3, %%xmm2        # Cb green + Cr green -> Cgreen             \n\
+                                                                            \n\
+# convert the luma part                                                     \n\
+movl      $0x10101010, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # Set xmm5 to   1010 1010 ... 1010 1010     \n\
+psubusb   %%xmm5, %%xmm6        # Y -= 16                                   \n\
+movdqa    %%xmm6, %%xmm7        # Copy 16 Y       Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0   \n\
+movl      $0x00ff00ff, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     00ff 00ff ... 00ff 00ff   \n\
+pand      %%xmm5, %%xmm6        # get Y even      00 Y6 00 Y4 00 Y2 00 Y0   \n\
+psrlw     $8, %%xmm7            # get Y odd       00 Y7 00 Y5 00 Y3 00 Y1   \n\
+psllw     $3, %%xmm6            # Promote precision                         \n\
+psllw     $3, %%xmm7            # Promote precision                         \n\
+movl      $0x253f253f, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     253f 253f ... 253f 253f   \n\
+pmulhw    %%xmm5, %%xmm6        # Mul 8 Y even    00 y6 00 y4 00 y2 00 y0   \n\
+pmulhw    %%xmm5, %%xmm7        # Mul 8 Y odd     00 y7 00 y5 00 y3 00 y1   \n\
+"
+
+#define SSE2_YUV_ADD "                                                      \n\
+# Do horizontal and vertical scaling                                        \n\
+movdqa    %%xmm0, %%xmm3        # Copy Cblue                                \n\
+movdqa    %%xmm1, %%xmm4        # Copy Cred                                 \n\
+movdqa    %%xmm2, %%xmm5        # Copy Cgreen                               \n\
+paddsw    %%xmm6, %%xmm0        # Y even + Cblue  00 B6 00 B4 00 B2 00 B0   \n\
+paddsw    %%xmm7, %%xmm3        # Y odd  + Cblue  00 B7 00 B5 00 B3 00 B1   \n\
+paddsw    %%xmm6, %%xmm1        # Y even + Cred   00 R6 00 R4 00 R2 00 R0   \n\
+paddsw    %%xmm7, %%xmm4        # Y odd  + Cred   00 R7 00 R5 00 R3 00 R1   \n\
+paddsw    %%xmm6, %%xmm2        # Y even + Cgreen 00 G6 00 G4 00 G2 00 G0   \n\
+paddsw    %%xmm7, %%xmm5        # Y odd  + Cgreen 00 G7 00 G5 00 G3 00 G1   \n\
+                                                                            \n\
+# Limit RGB even to 0..255                                                  \n\
+packuswb  %%xmm0, %%xmm0        # B6 B4 B2 B0 / B6 B4 B2 B0                 \n\
+packuswb  %%xmm1, %%xmm1        # R6 R4 R2 R0 / R6 R4 R2 R0                 \n\
+packuswb  %%xmm2, %%xmm2        # G6 G4 G2 G0 / G6 G4 G2 G0                 \n\
+                                                                            \n\
+# Limit RGB odd to 0..255                                                   \n\
+packuswb  %%xmm3, %%xmm3        # B7 B5 B3 B1 / B7 B5 B3 B1                 \n\
+packuswb  %%xmm4, %%xmm4        # R7 R5 R3 R1 / R7 R5 R3 R1                 \n\
+packuswb  %%xmm5, %%xmm5        # G7 G5 G3 G1 / G7 G5 G3 G1                 \n\
+                                                                            \n\
+# Interleave RGB even and odd                                               \n\
+punpcklbw %%xmm3, %%xmm0        #                 B7 B6 B5 B4 B3 B2 B1 B0   \n\
+punpcklbw %%xmm4, %%xmm1        #                 R7 R6 R5 R4 R3 R2 R1 R0   \n\
+punpcklbw %%xmm5, %%xmm2        #                 G7 G6 G5 G4 G3 G2 G1 G0   \n\
+"
+
+#define SSE2_UNPACK_15_ALIGNED "                                            \n\
+# mask unneeded bits off                                                    \n\
+movl      $0xf8f8f8f8, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     f8f8 f8f8 ... f8f8 f8f8   \n\
+pand      %%xmm5, %%xmm0        # b7b6b5b4 b3______ b7b6b5b4 b3______       \n\
+psrlw     $3,%%xmm0             # ______b7 b6b5b4b3 ______b7 b6b5b4b3       \n\
+pand      %%xmm5, %%xmm2        # g7g6g5g4 g3______ g7g6g5g4 g3______       \n\
+pand      %%xmm5, %%xmm1        # r7r6r5r4 r3______ r7r6r5r4 r3______       \n\
+psrlw     $1,%%xmm1             # __r7r6r5 r4r3____ __r7r6r5 r4r3____       \n\
+pxor      %%xmm4, %%xmm4        # zero mm4                                  \n\
+movdqa    %%xmm0, %%xmm5        # Copy B15-B0                               \n\
+movdqa    %%xmm2, %%xmm7        # Copy G15-G0                               \n\
+                                                                            \n\
+# convert rgb24 plane to rgb15 pack for pixel 0-7                           \n\
+punpcklbw %%xmm4, %%xmm2        # ________ ________ g7g6g5g4 g3______       \n\
+punpcklbw %%xmm1, %%xmm0        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
+psllw     $2,%%xmm2             # ________ ____g7g6 g5g4g3__ ________       \n\
+por       %%xmm2, %%xmm0        # r7r6r5r4 r3__g7g6 g5g4g3b7 b6b5b4b3       \n\
+movntdq   %%xmm0, (%3)          # store pixel 0-7                           \n\
+                                                                            \n\
+# convert rgb24 plane to rgb15 pack for pixel 8-15                          \n\
+punpckhbw %%xmm4, %%xmm7        # ________ ________ g7g6g5g4 g3______       \n\
+punpckhbw %%xmm1, %%xmm5        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
+psllw     $2,%%xmm7             # ________ ____g7g6 g5g4g3__ ________       \n\
+por       %%xmm7, %%xmm5        # r7r6r5r4 r3__g7g6 g5g4g3b7 b6b5b4b3       \n\
+movntdq   %%xmm5, 16(%3)        # store pixel 4-7                           \n\
+"
+
+#define SSE2_UNPACK_15_UNALIGNED "                                          \n\
+# mask unneeded bits off                                                    \n\
+movl      $0xf8f8f8f8, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     f8f8 f8f8 ... f8f8 f8f8   \n\
+pand      %%xmm5, %%xmm0        # b7b6b5b4 b3______ b7b6b5b4 b3______       \n\
+psrlw     $3,%%xmm0             # ______b7 b6b5b4b3 ______b7 b6b5b4b3       \n\
+pand      %%xmm5, %%xmm2        # g7g6g5g4 g3______ g7g6g5g4 g3______       \n\
+pand      %%xmm5, %%xmm1        # r7r6r5r4 r3______ r7r6r5r4 r3______       \n\
+psrlw     $1,%%xmm1             # __r7r6r5 r4r3____ __r7r6r5 r4r3____       \n\
+pxor      %%xmm4, %%xmm4        # zero mm4                                  \n\
+movdqa    %%xmm0, %%xmm5        # Copy B15-B0                               \n\
+movdqa    %%xmm2, %%xmm7        # Copy G15-G0                               \n\
+                                                                            \n\
+# convert rgb24 plane to rgb15 pack for pixel 0-7                           \n\
+punpcklbw %%xmm4, %%xmm2        # ________ ________ g7g6g5g4 g3______       \n\
+punpcklbw %%xmm1, %%xmm0        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
+psllw     $2,%%xmm2             # ________ ____g7g6 g5g4g3__ ________       \n\
+por       %%xmm2, %%xmm0        # r7r6r5r4 r3__g7g6 g5g4g3b7 b6b5b4b3       \n\
+movdqu    %%xmm0, (%3)          # store pixel 0-7                           \n\
+                                                                            \n\
+# convert rgb24 plane to rgb15 pack for pixel 8-15                          \n\
+punpckhbw %%xmm4, %%xmm7        # ________ ________ g7g6g5g4 g3______       \n\
+punpckhbw %%xmm1, %%xmm5        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
+psllw     $2,%%xmm7             # ________ ____g7g6 g5g4g3__ ________       \n\
+por       %%xmm7, %%xmm5        # r7r6r5r4 r3__g7g6 g5g4g3b7 b6b5b4b3       \n\
+movdqu    %%xmm5, 16(%3)        # store pixel 4-7                           \n\
+"
+
+#define SSE2_UNPACK_16_ALIGNED "                                            \n\
+# mask unneeded bits off                                                    \n\
+movl      $0xf8f8f8f8, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     f8f8 f8f8 ... f8f8 f8f8   \n\
+pand      %%xmm5, %%xmm0        # b7b6b5b4 b3______ b7b6b5b4 b3______       \n\
+pand      %%xmm5, %%xmm1        # r7r6r5r4 r3______ r7r6r5r4 r3______       \n\
+movl      $0xfcfcfcfc, %%eax    #                                           \n\
+movd      %%eax, %%xmm5         #                                           \n\
+pshufd    $0, %%xmm5, %%xmm5    # set xmm5 to     f8f8 f8f8 ... f8f8 f8f8   \n\
+pand      %%xmm5, %%xmm2        # g7g6g5g4 g3g2____ g7g6g5g4 g3g2____       \n\
+psrlw     $3,%%xmm0             # ______b7 b6b5b4b3 ______b7 b6b5b4b3       \n\
+pxor      %%xmm4, %%xmm4        # zero mm4                                  \n\
+movdqa    %%xmm0, %%xmm5        # Copy B15-B0                               \n\
+movdqa    %%xmm2, %%xmm7        # Copy G15-G0                               \n\
+                                                                            \n\
+# convert rgb24 plane to rgb16 pack for pixel 0-7                           \n\
+punpcklbw %%xmm4, %%xmm2        # ________ ________ g7g6g5g4 g3g2____       \n\
+punpcklbw %%xmm1, %%xmm0        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
+psllw     $3,%%xmm2             # ________ __g7g6g5 g4g3g2__ ________       \n\
+por       %%xmm2, %%xmm0        # r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3       \n\
+movntdq   %%xmm0, (%3)          # store pixel 0-7                           \n\
+                                                                            \n\
+# convert rgb24 plane to rgb16 pack for pixel 8-15                          \n\
+punpckhbw %%xmm4, %%xmm7        # ________ ________ g7g6g5g4 g3g2____       \n\
+punpckhbw %%xmm1, %%xmm5        # r7r6r5r4 r3______ ______b7 b6b5b4b3       \n\
+psllw     $3,%%xmm7             # ________ __g7g6g5 g4g3g2__ ________       \n\
+por       %%xmm7, %%xmm5        # r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3       \n\
+movntdq   %%xmm5, 16(%3)        # store pixel 4-7                           \n\
+"
 
 #define SSE2_UNPACK_16_UNALIGNED "                                          \n\
 # mask unneeded bits off                                                    \n\
@@ -705,104 +755,6 @@ por       %%xmm7, %%xmm5        # r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3       \n\
 movdqu    %%xmm5, 16(%3)        # store pixel 4-7                           \n\
 "
 
-#define MMX_INTRINSICS_UNPACK_16 \
-    mm0 = _mm_and_si64(mm0, (__m64)mmx_mask_f8); \
-    mm2 = _mm_and_si64(mm2, (__m64)mmx_mask_fc); \
-    mm1 = _mm_and_si64(mm1, (__m64)mmx_mask_f8); \
-    mm0 = _mm_srli_pi16(mm0, 3); \
-    mm4 = _mm_setzero_si64(); \
-    mm5 = mm0; \
-    mm7 = mm2; \
-    \
-    mm2 = _mm_unpacklo_pi8(mm2, mm4); \
-    mm0 = _mm_unpacklo_pi8(mm0, mm1); \
-    mm2 = _mm_slli_pi16(mm2, 3); \
-    mm0 = _mm_or_si64(mm0, mm2); \
-    tmp64 = *(uint64_t *)(p_y + 8); \
-    mm6 = (__m64)tmp64; \
-    *(uint64_t *)p_buffer = (uint64_t)mm0; \
-    \
-    mm7 = _mm_unpackhi_pi8(mm7, mm4); \
-    mm5 = _mm_unpackhi_pi8(mm5, mm1); \
-    mm7 = _mm_slli_pi16(mm7, 3); \
-    tmp64 = (uint64_t)*(uint32_t *)(p_u + 4); \
-    mm0 = (__m64)tmp64; \
-    mm5 = _mm_or_si64(mm5, mm7); \
-    tmp64 = (uint64_t)*(uint32_t *)(p_v + 4); \
-    mm1 = (__m64)tmp64; \
-    *(uint64_t *)(p_buffer + 4) = (uint64_t)mm5;
-
-#define SSE2_INTRINSICS_UNPACK_16_ALIGNED \
-    xmm5 = _mm_set1_epi32(0xf8f8f8f8UL); \
-    xmm0 = _mm_and_si128(xmm0, xmm5); \
-    xmm1 = _mm_and_si128(xmm1, xmm5); \
-    xmm5 = _mm_set1_epi32(0xfcfcfcfcUL); \
-    xmm2 = _mm_and_si128(xmm2, xmm5); \
-    xmm0 = _mm_srli_epi16(xmm0, 3); \
-    xmm4 = _mm_setzero_si128();          \
-    xmm5 = xmm0; \
-    xmm7 = xmm2; \
-    \
-    xmm2 = _mm_unpacklo_epi8(xmm2, xmm4); \
-    xmm0 = _mm_unpacklo_epi8(xmm0, xmm1); \
-    xmm2 = _mm_slli_epi16(xmm2, 3); \
-    xmm0 = _mm_or_si128(xmm0, xmm2); \
-    _mm_stream_si128((__m128i*)p_buffer, xmm0); \
-    \
-    xmm7 = _mm_unpackhi_epi8(xmm7, xmm4); \
-    xmm5 = _mm_unpackhi_epi8(xmm5, xmm1); \
-    xmm7 = _mm_slli_epi16(xmm7, 3); \
-    xmm5 = _mm_or_si128(xmm5, xmm7); \
-    _mm_stream_si128((__m128i*)(p_buffer+8), xmm5);
-
-#define SSE2_INTRINSICS_UNPACK_16_UNALIGNED \
-    xmm5 = _mm_set1_epi32(0xf8f8f8f8UL); \
-    xmm0 = _mm_and_si128(xmm0, xmm5); \
-    xmm1 = _mm_and_si128(xmm1, xmm5); \
-    xmm5 = _mm_set1_epi32(0xfcfcfcfcUL); \
-    xmm2 = _mm_and_si128(xmm2, xmm5); \
-    xmm0 = _mm_srli_epi16(xmm0, 3); \
-    xmm4 = _mm_setzero_si128();          \
-    xmm5 = xmm0; \
-    xmm7 = xmm2; \
-    \
-    xmm2 = _mm_unpacklo_epi8(xmm2, xmm4); \
-    xmm0 = _mm_unpacklo_epi8(xmm0, xmm1); \
-    xmm2 = _mm_slli_epi16(xmm2, 3); \
-    xmm0 = _mm_or_si128(xmm0, xmm2); \
-    _mm_storeu_si128((__m128i*)p_buffer, xmm0); \
-    \
-    xmm7 = _mm_unpackhi_epi8(xmm7, xmm4); \
-    xmm5 = _mm_unpackhi_epi8(xmm5, xmm1); \
-    xmm7 = _mm_slli_epi16(xmm7, 3); \
-    xmm5 = _mm_or_si128(xmm5, xmm7); \
-    _mm_storeu_si128((__m128i*)(p_buffer+8), xmm5);
-
-/*
- * convert RGB plane to RGB packed format,
- * mm0 -> B, mm1 -> R, mm2 -> G
- */
-
-#define MMX_UNPACK_32_ARGB "                                                \n\
-pxor      %%mm3, %%mm3  # zero mm3                                          \n\
-movq      %%mm0, %%mm4  #                 B7 B6 B5 B4 B3 B2 B1 B0           \n\
-punpcklbw %%mm2, %%mm4  #                 G3 B3 G2 B2 G1 B1 G0 B0           \n\
-movq      %%mm1, %%mm5  #                 R7 R6 R5 R4 R3 R2 R1 R0           \n\
-punpcklbw %%mm3, %%mm5  #                 00 R3 00 R2 00 R1 00 R0           \n\
-movq      %%mm4, %%mm6  #                 G3 B3 G2 B2 G1 B1 G0 B0           \n\
-punpcklwd %%mm5, %%mm4  #                 00 R1 B1 G1 00 R0 B0 G0           \n\
-movq      %%mm4, (%3)   # Store ARGB1 ARGB0                                 \n\
-punpckhwd %%mm5, %%mm6  #                 00 R3 B3 G3 00 R2 B2 G2           \n\
-movq      %%mm6, 8(%3)  # Store ARGB3 ARGB2                                 \n\
-punpckhbw %%mm2, %%mm0  #                 G7 B7 G6 B6 G5 B5 G4 B4           \n\
-punpckhbw %%mm3, %%mm1  #                 00 R7 00 R6 00 R5 00 R4           \n\
-movq      %%mm0, %%mm5  #                 G7 B7 G6 B6 G5 B5 G4 B4           \n\
-punpcklwd %%mm1, %%mm5  #                 00 R5 B5 G5 00 R4 B4 G4           \n\
-movq      %%mm5, 16(%3) # Store ARGB5 ARGB4                                 \n\
-punpckhwd %%mm1, %%mm0  #                 00 R7 B7 G7 00 R6 B6 G6           \n\
-movq      %%mm0, 24(%3) # Store ARGB7 ARGB6                                 \n\
-"
-
 #define SSE2_UNPACK_32_ARGB_ALIGNED "                                       \n\
 pxor      %%xmm3, %%xmm3  # zero xmm3                                       \n\
 movdqa    %%xmm0, %%xmm4  #               B7 B6 B5 B4 B3 B2 B1 B0           \n\
@@ -843,84 +795,6 @@ punpckhwd %%xmm1, %%xmm0  #               00 R7 B7 G7 00 R6 B6 G6           \n\
 movdqu    %%xmm0, 48(%3)  # Store ARGB15 ARGB14 ARGB13 ARGB12               \n\
 "
 
-#define MMX_INTRINSICS_UNPACK_32_ARGB \
-    mm3 = _mm_setzero_si64(); \
-    mm4 = mm0; \
-    mm4 = _mm_unpacklo_pi8(mm4, mm2); \
-    mm5 = mm1; \
-    mm5 = _mm_unpacklo_pi8(mm5, mm3); \
-    mm6 = mm4; \
-    mm4 = _mm_unpacklo_pi16(mm4, mm5); \
-    *(uint64_t *)p_buffer = (uint64_t)mm4; \
-    mm6 = _mm_unpackhi_pi16(mm6, mm5); \
-    *(uint64_t *)(p_buffer + 2) = (uint64_t)mm6; \
-    mm0 = _mm_unpackhi_pi8(mm0, mm2); \
-    mm1 = _mm_unpackhi_pi8(mm1, mm3); \
-    mm5 = mm0; \
-    mm5 = _mm_unpacklo_pi16(mm5, mm1); \
-    *(uint64_t *)(p_buffer + 4) = (uint64_t)mm5; \
-    mm0 = _mm_unpackhi_pi16(mm0, mm1); \
-    *(uint64_t *)(p_buffer + 6) = (uint64_t)mm0;
-
-#define SSE2_INTRINSICS_UNPACK_32_ARGB_ALIGNED \
-    xmm3 = _mm_setzero_si128();          \
-    xmm4 = xmm0; \
-    xmm4 = _mm_unpacklo_epi8(xmm4, xmm2); \
-    xmm5 = xmm1; \
-    xmm5 = _mm_unpacklo_epi8(xmm5, xmm3); \
-    xmm6 = xmm4; \
-    xmm4 = _mm_unpacklo_epi16(xmm4, xmm5); \
-    _mm_stream_si128((__m128i*)(p_buffer), xmm4); \
-    xmm6 = _mm_unpackhi_epi16(xmm6, xmm5); \
-    _mm_stream_si128((__m128i*)(p_buffer+4), xmm6); \
-    xmm0 = _mm_unpackhi_epi8(xmm0, xmm2); \
-    xmm1 = _mm_unpackhi_epi8(xmm1, xmm3); \
-    xmm5 = xmm0; \
-    xmm5 = _mm_unpacklo_epi16(xmm5, xmm1); \
-    _mm_stream_si128((__m128i*)(p_buffer+8), xmm5); \
-    xmm0 = _mm_unpackhi_epi16(xmm0, xmm1); \
-    _mm_stream_si128((__m128i*)(p_buffer+12), xmm0);
-
-#define SSE2_INTRINSICS_UNPACK_32_ARGB_UNALIGNED \
-    xmm3 = _mm_setzero_si128();          \
-    xmm4 = xmm0; \
-    xmm4 = _mm_unpacklo_epi8(xmm4, xmm2); \
-    xmm5 = xmm1; \
-    xmm5 = _mm_unpacklo_epi8(xmm5, xmm3); \
-    xmm6 = xmm4; \
-    xmm4 = _mm_unpacklo_epi16(xmm4, xmm5); \
-    _mm_storeu_si128((__m128i*)(p_buffer), xmm4); \
-    xmm6 = _mm_unpackhi_epi16(xmm6, xmm5); \
-    _mm_storeu_si128((__m128i*)(p_buffer+4), xmm6); \
-    xmm0 = _mm_unpackhi_epi8(xmm0, xmm2); \
-    xmm1 = _mm_unpackhi_epi8(xmm1, xmm3); \
-    xmm5 = xmm0; \
-    xmm5 = _mm_unpacklo_epi16(xmm5, xmm1); \
-    _mm_storeu_si128((__m128i*)(p_buffer+8), xmm5); \
-    xmm0 = _mm_unpackhi_epi16(xmm0, xmm1); \
-    _mm_storeu_si128((__m128i*)(p_buffer+12), xmm0);
-
-#define MMX_UNPACK_32_BGRA "                                                \n\
-pxor      %%mm3, %%mm3  # zero mm3                                          \n\
-movq      %%mm2, %%mm4  #                 G7 G6 G5 G4 G3 G2 G1 G0           \n\
-punpcklbw %%mm0, %%mm4  #                 B3 G3 B2 G2 B1 G1 B0 G0           \n\
-punpcklbw %%mm1, %%mm3  #                 R3 00 R2 00 R1 00 R0 00           \n\
-movq      %%mm3, %%mm5  #                 R3 00 R2 00 R1 00 R0 00           \n\
-punpcklwd %%mm4, %%mm3  #                 B1 G1 R1 00 B0 G0 R0 00           \n\
-movq      %%mm3, (%3)   # Store BGRA1 BGRA0                                 \n\
-punpckhwd %%mm4, %%mm5  #                 B3 G3 R3 00 B2 G2 R2 00           \n\
-movq      %%mm5, 8(%3)  # Store BGRA3 BGRA2                                 \n\
-pxor      %%mm3, %%mm3  # zero mm3                                          \n\
-movq      %%mm2, %%mm4  #                 G7 G6 G5 G4 G3 G2 G1 G0           \n\
-punpckhbw %%mm0, %%mm4  #                 B7 G7 B6 G6 B5 G5 B4 G4           \n\
-punpckhbw %%mm1, %%mm3  #                 R7 00 R6 00 R5 00 R4 00           \n\
-movq      %%mm3, %%mm5  #                 R7 00 R6 00 R5 00 R4 00           \n\
-punpcklwd %%mm1, %%mm3  #                 B5 G5 R5 00 B4 G4 R4 00           \n\
-movq      %%mm3, 16(%3) # Store BGRA5 BGRA4                                 \n\
-punpckhwd %%mm4, %%mm5  #                 B7 G7 R7 00 B6 G6 R6 00           \n\
-movq      %%mm5, 24(%3) # Store BGRA7 BGRA6                                 \n\
-"
-
 #define SSE2_UNPACK_32_BGRA_ALIGNED "                                       \n\
 pxor      %%xmm3, %%xmm3  # zero mm3                                        \n\
 movdqa    %%xmm2, %%xmm4  #                 G7 G6 G5 G4 G3 G2 G1 G0         \n\
@@ -930,16 +804,15 @@ movdqa    %%xmm3, %%xmm5  #                 R3 00 R2 00 R1 00 R0 00         \n\
 punpcklwd %%xmm4, %%xmm3  #                 B1 G1 R1 00 B0 G0 R0 00         \n\
 movntdq   %%xmm3, (%3)    # Store BGRA3 BGRA2 BGRA1 BGRA0                   \n\
 punpckhwd %%xmm4, %%xmm5  #                 B3 G3 R3 00 B2 G2 R2 00         \n\
-movntdq   %%xmm5, 8(%3)   # Store BGRA7 BGRA6 BGRA5 BGRA4                   \n\
-pxor      %%xmm3, %%xmm3  # zero mm3                                        \n\
-movdqa    %%xmm2, %%xmm4  #                 G7 G6 G5 G4 G3 G2 G1 G0         \n\
-punpckhbw %%xmm0, %%xmm4  #                 B7 G7 B6 G6 B5 G5 B4 G4         \n\
-punpckhbw %%xmm1, %%xmm3  #                 R7 00 R6 00 R5 00 R4 00         \n\
-movdqa    %%xmm3, %%xmm5  #                 R7 00 R6 00 R5 00 R4 00         \n\
-punpcklwd %%xmm1, %%xmm3  #                 B5 G5 R5 00 B4 G4 R4 00         \n\
-movntdq   %%xmm3, 16(%3)  # Store BGRA11 BGRA10 BGRA9 BGRA8                 \n\
-punpckhwd %%xmm4, %%xmm5  #                 B7 G7 R7 00 B6 G6 R6 00         \n\
-movntdq   %%xmm5, 24(%3)  # Store BGRA15 BGRA14 BGRA13 BGRA12               \n\
+movntdq   %%xmm5, 16(%3)  # Store BGRA7 BGRA6 BGRA5 BGRA4                   \n\
+pxor      %%xmm6, %%xmm6  # zero mm6                                        \n\
+punpckhbw %%xmm0, %%xmm2  #                 B7 G7 B6 G6 B5 G5 B4 G4         \n\
+punpckhbw %%xmm1, %%xmm6  #                 R7 00 R6 00 R5 00 R4 00         \n\
+movdqa    %%xmm6, %%xmm0  #                 R7 00 R6 00 R5 00 R4 00         \n\
+punpcklwd %%xmm2, %%xmm6  #                 B5 G5 R5 00 B4 G4 R4 00         \n\
+movntdq   %%xmm6, 32(%3)  # Store BGRA11 BGRA10 BGRA9 BGRA8                 \n\
+punpckhwd %%xmm2, %%xmm0  #                 B7 G7 R7 00 B6 G6 R6 00         \n\
+movntdq   %%xmm0, 48(%3)  # Store BGRA15 BGRA14 BGRA13 BGRA12               \n\
 "
 
 #define SSE2_UNPACK_32_BGRA_UNALIGNED "                                     \n\
@@ -951,75 +824,327 @@ movdqa    %%xmm3, %%xmm5  #                 R3 00 R2 00 R1 00 R0 00         \n\
 punpcklwd %%xmm4, %%xmm3  #                 B1 G1 R1 00 B0 G0 R0 00         \n\
 movdqu    %%xmm3, (%3)    # Store BGRA3 BGRA2 BGRA1 BGRA0                   \n\
 punpckhwd %%xmm4, %%xmm5  #                 B3 G3 R3 00 B2 G2 R2 00         \n\
-movdqu    %%xmm5, 8(%3)   # Store BGRA7 BGRA6 BGRA5 BGRA4                   \n\
+movdqu    %%xmm5, 16(%3)  # Store BGRA7 BGRA6 BGRA5 BGRA4                   \n\
+pxor      %%xmm6, %%xmm6  # zero mm6                                        \n\
+punpckhbw %%xmm0, %%xmm2  #                 B7 G7 B6 G6 B5 G5 B4 G4         \n\
+punpckhbw %%xmm1, %%xmm6  #                 R7 00 R6 00 R5 00 R4 00         \n\
+movdqa    %%xmm6, %%xmm0  #                 R7 00 R6 00 R5 00 R4 00         \n\
+punpcklwd %%xmm2, %%xmm6  #                 B5 G5 R5 00 B4 G4 R4 00         \n\
+movdqu    %%xmm6, 32(%3)  # Store BGRA11 BGRA10 BGRA9 BGRA8                 \n\
+punpckhwd %%xmm2, %%xmm0  #                 B7 G7 R7 00 B6 G6 R6 00         \n\
+movdqu    %%xmm0, 48(%3)  # Store BGRA15 BGRA14 BGRA13 BGRA12               \n\
+"
+
+#define SSE2_UNPACK_32_ABGR_ALIGNED "                                       \n\
 pxor      %%xmm3, %%xmm3  # zero mm3                                        \n\
-movdqa    %%xmm2, %%xmm4  #                 G7 G6 G5 G4 G3 G2 G1 G0         \n\
-punpckhbw %%xmm0, %%xmm4  #                 B7 G7 B6 G6 B5 G5 B4 G4         \n\
-punpckhbw %%xmm1, %%xmm3  #                 R7 00 R6 00 R5 00 R4 00         \n\
-movdqa    %%xmm3, %%xmm5  #                 R7 00 R6 00 R5 00 R4 00         \n\
-punpcklwd %%xmm1, %%xmm3  #                 B5 G5 R5 00 B4 G4 R4 00         \n\
-movdqu    %%xmm3, 16(%3)  # Store BGRA11 BGRA10 BGRA9 BGRA8                 \n\
-punpckhwd %%xmm4, %%xmm5  #                 B7 G7 R7 00 B6 G6 R6 00         \n\
-movdqu    %%xmm5, 24(%3)  # Store BGRA15 BGRA14 BGRA13 BGRA12               \n\
+movdqa    %%xmm1, %%xmm4  #                 R7 R6 R5 R4 R3 R2 R1 R0         \n\
+punpcklbw %%xmm2, %%xmm4  #                 G3 R3 G2 R2 G1 R1 G0 R0         \n\
+movdqa    %%xmm0, %%xmm5  #                 B7 B6 B5 B4 B3 B2 B1 B0         \n\
+punpcklbw %%xmm3, %%xmm5  #                 00 B3 00 B2 00 B1 00 B0         \n\
+movdqa    %%xmm4, %%xmm6  #                 G3 R3 G2 R2 G1 R1 G0 R0         \n\
+punpcklwd %%xmm5, %%xmm4  #                 00 B1 G1 R1 00 B0 G0 R0         \n\
+movntdq   %%xmm4, (%3)    # Store ABGR3 ABGR2 ABGR1 ABGR0                   \n\
+punpckhwd %%xmm5, %%xmm6  #                 00 B3 G3 R3 00 B2 G2 R2         \n\
+movntdq   %%xmm6, 16(%3)  # Store ABGR7 ABGR6 ABGR5 ABGR4                   \n\
+punpckhbw %%xmm2, %%xmm1  #                 G7 R7 G6 R6 G5 R5 G4 R4         \n\
+punpckhbw %%xmm3, %%xmm0  #                 00 B7 00 B6 00 B5 00 B4         \n\
+movdqa    %%xmm1, %%xmm2  #                 G7 R7 G6 R6 G5 R5 G4 R4         \n\
+punpcklwd %%xmm0, %%xmm1  #                 00 B5 G5 R5 00 B4 G4 R4         \n\
+movntdq   %%xmm1, 32(%3)  # Store ABGR11 ABGR10 ABGR9 ABGR8                 \n\
+punpckhwd %%xmm0, %%xmm2  #                 B7 G7 R7 00 B6 G6 R6 00         \n\
+movntdq   %%xmm2, 48(%3)  # Store ABGR15 ABGR14 ABGR13 ABGR12               \n\
+"
+
+#define SSE2_UNPACK_32_ABGR_UNALIGNED "                                     \n\
+pxor      %%xmm3, %%xmm3  # zero mm3                                        \n\
+movdqa    %%xmm1, %%xmm4  #                 R7 R6 R5 R4 R3 R2 R1 R0         \n\
+punpcklbw %%xmm2, %%xmm4  #                 G3 R3 G2 R2 G1 R1 G0 R0         \n\
+movdqa    %%xmm0, %%xmm5  #                 B7 B6 B5 B4 B3 B2 B1 B0         \n\
+punpcklbw %%xmm3, %%xmm5  #                 00 B3 00 B2 00 B1 00 B0         \n\
+movdqa    %%xmm4, %%xmm6  #                 G3 R3 G2 R2 G1 R1 G0 R0         \n\
+punpcklwd %%xmm5, %%xmm4  #                 00 B1 G1 R1 00 B0 G0 R0         \n\
+movdqu    %%xmm4, (%3)    # Store ABGR3 ABGR2 ABGR1 ABGR0                   \n\
+punpckhwd %%xmm5, %%xmm6  #                 00 B3 G3 R3 00 B2 G2 R2         \n\
+movdqu    %%xmm6, 16(%3)  # Store ABGR7 ABGR6 ABGR5 ABGR4                   \n\
+punpckhbw %%xmm2, %%xmm1  #                 G7 R7 G6 R6 G5 R5 G4 R4         \n\
+punpckhbw %%xmm3, %%xmm0  #                 00 B7 00 B6 00 B5 00 B4         \n\
+movdqa    %%xmm1, %%xmm2  #                 R7 00 R6 00 R5 00 R4 00         \n\
+punpcklwd %%xmm0, %%xmm1  #                 00 B5 G5 R5 00 B4 G4 R4         \n\
+movdqu    %%xmm1, 32(%3)  # Store ABGR11 ABGR10 ABGR9 ABGR8                 \n\
+punpckhwd %%xmm0, %%xmm2  #                 B7 G7 R7 00 B6 G6 R6 00         \n\
+movdqu    %%xmm2, 48(%3)  # Store ABGR15 ABGR14 ABGR13 ABGR12               \n\
 "
 
-#define MMX_INTRINSICS_UNPACK_32_BGRA \
-    mm3 = _mm_setzero_si64(); \
-    mm4 = mm2; \
-    mm4 = _mm_unpacklo_pi8(mm4, mm0); \
-    mm1 = _mm_unpacklo_pi8(mm1, mm3); \
-    mm5 = mm3; \
-    mm3 = _mm_unpacklo_pi16(mm3, mm4); \
-    *(uint64_t *)p_buffer = (uint64_t)mm3; \
-    mm5 = _mm_unpackhi_pi16(mm5, mm4); \
-    *(uint64_t *)(p_buffer + 2) = (uint64_t)mm5; \
-    mm3 = _mm_setzero_si64(); \
-    mm4 = mm2; \
-    mm0 = _mm_unpackhi_pi8(mm0, mm4); \
-    mm1 = _mm_unpackhi_pi8(mm1, mm3); \
-    mm5 = mm3; \
-    mm3 = _mm_unpacklo_pi16(mm3, mm1); \
-    *(uint64_t *)(p_buffer + 4) = (uint64_t)mm3; \
-    mm5 = _mm_unpackhi_pi16(mm5, mm4); \
-    *(uint64_t *)(p_buffer + 6) = (uint64_t)mm5; \
-
-#define SSE2_INTRINSICS_UNPACK_32_BGRA_ALIGNED \
-    xmm3 = _mm_setzero_si128(); \
-    xmm4 = xmm2; \
-    xmm4 = _mm_unpacklo_epi8(xmm4, xmm0); \
-    xmm1 = _mm_unpacklo_epi8(xmm1, xmm3); \
-    xmm5 = xmm3; \
-    xmm3 = _mm_unpacklo_epi16(xmm3, xmm4); \
+#elif defined(HAVE_SSE2_INTRINSICS)
+
+/* SSE2 intrinsics */
+
+#include <emmintrin.h>
+
+#define SSE2_CALL(SSE2_INSTRUCTIONS)            \
+    do {                                        \
+        __m128i xmm0, xmm1, xmm2, xmm3,         \
+                xmm4, xmm5, xmm6, xmm7;         \
+        SSE2_INSTRUCTIONS                       \
+    } while(0)
+
+#define SSE2_END  _mm_sfence()
+
+#define SSE2_INIT_16_ALIGNED                \
+    xmm0 = _mm_loadl_epi64((__m128i *)p_u); \
+    xmm1 = _mm_loadl_epi64((__m128i *)p_v); \
+    xmm4 = _mm_setzero_si128();             \
+    xmm6 = _mm_load_si128((__m128i *)p_y);
+
+#define SSE2_INIT_16_UNALIGNED              \
+    xmm0 = _mm_loadl_epi64((__m128i *)p_u); \
+    xmm1 = _mm_loadl_epi64((__m128i *)p_v); \
+    xmm4 = _mm_setzero_si128();             \
+    xmm6 = _mm_loadu_si128((__m128i *)p_y); \
+    _mm_prefetch(p_buffer, _MM_HINT_NTA);
+
+#define SSE2_INIT_32_ALIGNED                \
+    xmm0 = _mm_loadl_epi64((__m128i *)p_u); \
+    xmm1 = _mm_loadl_epi64((__m128i *)p_v); \
+    xmm4 = _mm_setzero_si128();             \
+    xmm6 = _mm_load_si128((__m128i *)p_y);
+
+#define SSE2_INIT_32_UNALIGNED              \
+    xmm0 = _mm_loadl_epi64((__m128i *)p_u); \
+    xmm1 = _mm_loadl_epi64((__m128i *)p_v); \
+    xmm4 = _mm_setzero_si128();             \
+    xmm6 = _mm_loadu_si128((__m128i *)p_y); \
+    _mm_prefetch(p_buffer, _MM_HINT_NTA);
+
+#define SSE2_YUV_MUL                        \
+    xmm0 = _mm_unpacklo_epi8(xmm0, xmm4);   \
+    xmm1 = _mm_unpacklo_epi8(xmm1, xmm4);   \
+    xmm5 = _mm_set1_epi32(0x00800080UL);    \
+    xmm0 = _mm_subs_epi16(xmm0, xmm5);      \
+    xmm1 = _mm_subs_epi16(xmm1, xmm5);      \
+    xmm0 = _mm_slli_epi16(xmm0, 3);         \
+    xmm1 = _mm_slli_epi16(xmm1, 3);         \
+    xmm2 = xmm0;                            \
+    xmm3 = xmm1;                            \
+    xmm5 = _mm_set1_epi32(0xf37df37dUL);    \
+    xmm2 = _mm_mulhi_epi16(xmm2, xmm5);     \
+    xmm5 = _mm_set1_epi32(0xe5fce5fcUL);    \
+    xmm3 = _mm_mulhi_epi16(xmm3, xmm5);     \
+    xmm5 = _mm_set1_epi32(0x40934093UL);    \
+    xmm0 = _mm_mulhi_epi16(xmm0, xmm5);     \
+    xmm5 = _mm_set1_epi32(0x33123312UL);    \
+    xmm1 = _mm_mulhi_epi16(xmm1, xmm5);     \
+    xmm2 = _mm_adds_epi16(xmm2, xmm3);      \
+    \
+    xmm5 = _mm_set1_epi32(0x10101010UL);    \
+    xmm6 = _mm_subs_epu8(xmm6, xmm5);       \
+    xmm7 = xmm6;                            \
+    xmm5 = _mm_set1_epi32(0x00ff00ffUL);    \
+    xmm6 = _mm_and_si128(xmm6, xmm5);       \
+    xmm7 = _mm_srli_epi16(xmm7, 8);         \
+    xmm6 = _mm_slli_epi16(xmm6, 3);         \
+    xmm7 = _mm_slli_epi16(xmm7, 3);         \
+    xmm5 = _mm_set1_epi32(0x253f253fUL);    \
+    xmm6 = _mm_mulhi_epi16(xmm6, xmm5);     \
+    xmm7 = _mm_mulhi_epi16(xmm7, xmm5);
+
+#define SSE2_YUV_ADD                        \
+    xmm3 = xmm0;                            \
+    xmm4 = xmm1;                            \
+    xmm5 = xmm2;                            \
+    xmm0 = _mm_adds_epi16(xmm0, xmm6);      \
+    xmm3 = _mm_adds_epi16(xmm3, xmm7);      \
+    xmm1 = _mm_adds_epi16(xmm1, xmm6);      \
+    xmm4 = _mm_adds_epi16(xmm4, xmm7);      \
+    xmm2 = _mm_adds_epi16(xmm2, xmm6);      \
+    xmm5 = _mm_adds_epi16(xmm5, xmm7);      \
+    \
+    xmm0 = _mm_packus_epi16(xmm0, xmm0);    \
+    xmm1 = _mm_packus_epi16(xmm1, xmm1);    \
+    xmm2 = _mm_packus_epi16(xmm2, xmm2);    \
+    \
+    xmm3 = _mm_packus_epi16(xmm3, xmm3);    \
+    xmm4 = _mm_packus_epi16(xmm4, xmm4);    \
+    xmm5 = _mm_packus_epi16(xmm5, xmm5);    \
+    \
+    xmm0 = _mm_unpacklo_epi8(xmm0, xmm3);   \
+    xmm1 = _mm_unpacklo_epi8(xmm1, xmm4);   \
+    xmm2 = _mm_unpacklo_epi8(xmm2, xmm5);
+
+#define SSE2_UNPACK_15_ALIGNED              \
+    xmm5 = _mm_set1_epi32(0xf8f8f8f8UL);    \
+    xmm0 = _mm_and_si128(xmm0, xmm5);       \
+    xmm0 = _mm_srli_epi16(xmm0, 3);         \
+    xmm2 = _mm_and_si128(xmm2, xmm5);       \
+    xmm1 = _mm_and_si128(xmm1, xmm5);       \
+    xmm1 = _mm_srli_epi16(xmm1, 1);         \
+    xmm4 = _mm_setzero_si128();             \
+    xmm5 = xmm0;                            \
+    xmm7 = xmm2;                            \
+    \
+    xmm2 = _mm_unpacklo_epi8(xmm2, xmm4);   \
+    xmm0 = _mm_unpacklo_epi8(xmm0, xmm1);   \
+    xmm2 = _mm_slli_epi16(xmm2, 2);         \
+    xmm0 = _mm_or_si128(xmm0, xmm2);        \
+    _mm_stream_si128((__m128i*)p_buffer, xmm0); \
+    \
+    xmm7 = _mm_unpackhi_epi8(xmm7, xmm4);   \
+    xmm5 = _mm_unpackhi_epi8(xmm5, xmm1);   \
+    xmm7 = _mm_slli_epi16(xmm7, 2);         \
+    xmm5 = _mm_or_si128(xmm5, xmm7);        \
+    _mm_stream_si128((__m128i*)(p_buffer+8), xmm5);
+
+#define SSE2_UNPACK_15_UNALIGNED            \
+    xmm5 = _mm_set1_epi32(0xf8f8f8f8UL);    \
+    xmm0 = _mm_and_si128(xmm0, xmm5);       \
+    xmm0 = _mm_srli_epi16(xmm0, 3);         \
+    xmm2 = _mm_and_si128(xmm2, xmm5);       \
+    xmm1 = _mm_and_si128(xmm1, xmm5);       \
+    xmm1 = _mm_srli_epi16(xmm1, 1);         \
+    xmm4 = _mm_setzero_si128();             \
+    xmm5 = xmm0;                            \
+    xmm7 = xmm2;                            \
+    \
+    xmm2 = _mm_unpacklo_epi8(xmm2, xmm4);   \
+    xmm0 = _mm_unpacklo_epi8(xmm0, xmm1);   \
+    xmm2 = _mm_slli_epi16(xmm2, 2);         \
+    xmm0 = _mm_or_si128(xmm0, xmm2);        \
+    _mm_storeu_si128((__m128i*)p_buffer, xmm0); \
+    \
+    xmm7 = _mm_unpackhi_epi8(xmm7, xmm4);   \
+    xmm5 = _mm_unpackhi_epi8(xmm5, xmm1);   \
+    xmm7 = _mm_slli_epi16(xmm7, 2);         \
+    xmm5 = _mm_or_si128(xmm5, xmm7);        \
+    _mm_storeu_si128((__m128i*)(p_buffer+16), xmm5);
+
+#define SSE2_UNPACK_16_ALIGNED              \
+    xmm5 = _mm_set1_epi32(0xf8f8f8f8UL);    \
+    xmm0 = _mm_and_si128(xmm0, xmm5);       \
+    xmm1 = _mm_and_si128(xmm1, xmm5);       \
+    xmm5 = _mm_set1_epi32(0xfcfcfcfcUL);    \
+    xmm2 = _mm_and_si128(xmm2, xmm5);       \
+    xmm0 = _mm_srli_epi16(xmm0, 3);         \
+    xmm4 = _mm_setzero_si128();             \
+    xmm5 = xmm0;                            \
+    xmm7 = xmm2;                            \
+    \
+    xmm2 = _mm_unpacklo_epi8(xmm2, xmm4);   \
+    xmm0 = _mm_unpacklo_epi8(xmm0, xmm1);   \
+    xmm2 = _mm_slli_epi16(xmm2, 3);         \
+    xmm0 = _mm_or_si128(xmm0, xmm2);        \
+    _mm_stream_si128((__m128i*)p_buffer, xmm0); \
+    \
+    xmm7 = _mm_unpackhi_epi8(xmm7, xmm4);   \
+    xmm5 = _mm_unpackhi_epi8(xmm5, xmm1);   \
+    xmm7 = _mm_slli_epi16(xmm7, 3);         \
+    xmm5 = _mm_or_si128(xmm5, xmm7);        \
+    _mm_stream_si128((__m128i*)(p_buffer+8), xmm5);
+
+#define SSE2_UNPACK_16_UNALIGNED            \
+    xmm5 = _mm_set1_epi32(0xf8f8f8f8UL);    \
+    xmm0 = _mm_and_si128(xmm0, xmm5);       \
+    xmm1 = _mm_and_si128(xmm1, xmm5);       \
+    xmm5 = _mm_set1_epi32(0xfcfcfcfcUL);    \
+    xmm2 = _mm_and_si128(xmm2, xmm5);       \
+    xmm0 = _mm_srli_epi16(xmm0, 3);         \
+    xmm4 = _mm_setzero_si128();             \
+    xmm5 = xmm0;                            \
+    xmm7 = xmm2;                            \
+    \
+    xmm2 = _mm_unpacklo_epi8(xmm2, xmm4);   \
+    xmm0 = _mm_unpacklo_epi8(xmm0, xmm1);   \
+    xmm2 = _mm_slli_epi16(xmm2, 3);         \
+    xmm0 = _mm_or_si128(xmm0, xmm2);        \
+    _mm_storeu_si128((__m128i*)p_buffer, xmm0); \
+    \
+    xmm7 = _mm_unpackhi_epi8(xmm7, xmm4);   \
+    xmm5 = _mm_unpackhi_epi8(xmm5, xmm1);   \
+    xmm7 = _mm_slli_epi16(xmm7, 3);         \
+    xmm5 = _mm_or_si128(xmm5, xmm7);        \
+    _mm_storeu_si128((__m128i*)(p_buffer+8), xmm5);
+
+#define SSE2_UNPACK_32_ARGB_ALIGNED         \
+    xmm3 = _mm_setzero_si128();             \
+    xmm4 = xmm0;                            \
+    xmm4 = _mm_unpacklo_epi8(xmm4, xmm2);   \
+    xmm5 = xmm1;                            \
+    xmm5 = _mm_unpacklo_epi8(xmm5, xmm3);   \
+    xmm6 = xmm4;                            \
+    xmm4 = _mm_unpacklo_epi16(xmm4, xmm5);  \
+    _mm_stream_si128((__m128i*)(p_buffer), xmm4); \
+    xmm6 = _mm_unpackhi_epi16(xmm6, xmm5);  \
+    _mm_stream_si128((__m128i*)(p_buffer+4), xmm6); \
+    xmm0 = _mm_unpackhi_epi8(xmm0, xmm2);   \
+    xmm1 = _mm_unpackhi_epi8(xmm1, xmm3);   \
+    xmm5 = xmm0;                            \
+    xmm5 = _mm_unpacklo_epi16(xmm5, xmm1);  \
+    _mm_stream_si128((__m128i*)(p_buffer+8), xmm5); \
+    xmm0 = _mm_unpackhi_epi16(xmm0, xmm1);  \
+    _mm_stream_si128((__m128i*)(p_buffer+12), xmm0);
+
+#define SSE2_UNPACK_32_ARGB_UNALIGNED       \
+    xmm3 = _mm_setzero_si128();             \
+    xmm4 = xmm0;                            \
+    xmm4 = _mm_unpacklo_epi8(xmm4, xmm2);   \
+    xmm5 = xmm1;                            \
+    xmm5 = _mm_unpacklo_epi8(xmm5, xmm3);   \
+    xmm6 = xmm4;                            \
+    xmm4 = _mm_unpacklo_epi16(xmm4, xmm5);  \
+    _mm_storeu_si128((__m128i*)(p_buffer), xmm4); \
+    xmm6 = _mm_unpackhi_epi16(xmm6, xmm5);  \
+    _mm_storeu_si128((__m128i*)(p_buffer+4), xmm6); \
+    xmm0 = _mm_unpackhi_epi8(xmm0, xmm2);   \
+    xmm1 = _mm_unpackhi_epi8(xmm1, xmm3);   \
+    xmm5 = xmm0;                            \
+    xmm5 = _mm_unpacklo_epi16(xmm5, xmm1);  \
+    _mm_storeu_si128((__m128i*)(p_buffer+8), xmm5); \
+    xmm0 = _mm_unpackhi_epi16(xmm0, xmm1);  \
+    _mm_storeu_si128((__m128i*)(p_buffer+12), xmm0);
+
+#define SSE2_UNPACK_32_BGRA_ALIGNED         \
+    xmm3 = _mm_setzero_si128();             \
+    xmm4 = xmm2;                            \
+    xmm4 = _mm_unpacklo_epi8(xmm4, xmm0);   \
+    xmm3 = _mm_unpacklo_epi8(xmm3, xmm1);   \
+    xmm5 = xmm3;                            \
+    xmm3 = _mm_unpacklo_epi16(xmm3, xmm4);  \
     _mm_stream_si128((__m128i*)(p_buffer), xmm3); \
-    xmm5 = _mm_unpackhi_epi16(xmm5, xmm4); \
+    xmm5 = _mm_unpackhi_epi16(xmm5, xmm4);  \
     _mm_stream_si128((__m128i*)(p_buffer+4), xmm5); \
-    xmm3 = _mm_setzero_si128(); \
-    xmm4 = xmm2; \
-    xmm0 = _mm_unpackhi_epi8(xmm0, xmm4); \
-    xmm1 = _mm_unpackhi_epi8(xmm1, xmm3); \
-    xmm5 = xmm3; \
-    xmm3 = _mm_unpacklo_epi16(xmm3, xmm1); \
-    _mm_stream_si128((__m128i*)(p_buffer+8), xmm3); \
-    xmm5 = _mm_unpackhi_epi16(xmm5, xmm4); \
-    _mm_stream_si128((__m128i*)(p_buffer+12), xmm5); \
-
-#define SSE2_INTRINSICS_UNPACK_32_BGRA_UNALIGNED \
-    xmm3 = _mm_setzero_si128(); \
-    xmm4 = xmm2; \
-    xmm4 = _mm_unpacklo_epi8(xmm4, xmm0); \
-    xmm1 = _mm_unpacklo_epi8(xmm1, xmm3); \
-    xmm5 = xmm3; \
-    xmm3 = _mm_unpacklo_epi16(xmm3, xmm4); \
+    xmm6 = _mm_setzero_si128();             \
+    xmm2 = _mm_unpackhi_epi8(xmm2, xmm0);   \
+    xmm6 = _mm_unpackhi_epi8(xmm6, xmm1);   \
+    xmm0 = xmm6;                            \
+    xmm6 = _mm_unpacklo_epi16(xmm6, xmm2);  \
+    _mm_stream_si128((__m128i*)(p_buffer+8), xmm6); \
+    xmm0 = _mm_unpackhi_epi16(xmm0, xmm2);  \
+    _mm_stream_si128((__m128i*)(p_buffer+12), xmm0);
+
+#define SSE2_UNPACK_32_BGRA_UNALIGNED       \
+    xmm3 = _mm_setzero_si128();             \
+    xmm4 = xmm2;                            \
+    xmm4 = _mm_unpacklo_epi8(xmm4, xmm0);   \
+    xmm3 = _mm_unpacklo_epi8(xmm3, xmm1);   \
+    xmm5 = xmm3;                            \
+    xmm3 = _mm_unpacklo_epi16(xmm3, xmm4);  \
     _mm_storeu_si128((__m128i*)(p_buffer), xmm3); \
-    xmm5 = _mm_unpackhi_epi16(xmm5, xmm4); \
+    xmm5 = _mm_unpackhi_epi16(xmm5, xmm4);  \
     _mm_storeu_si128((__m128i*)(p_buffer+4), xmm5); \
-    xmm3 = _mm_setzero_si128(); \
-    xmm4 = xmm2; \
-    xmm0 = _mm_unpackhi_epi8(xmm0, xmm4); \
-    xmm1 = _mm_unpackhi_epi8(xmm1, xmm3); \
-    xmm5 = xmm3; \
-    xmm3 = _mm_unpacklo_epi16(xmm3, xmm1); \
-    _mm_storeu_si128((__m128i*)(p_buffer+8), xmm3); \
-    xmm5 = _mm_unpackhi_epi16(xmm5, xmm4); \
-    _mm_storeu_si128((__m128i*)(p_buffer+12), xmm5); \
+    xmm6 = _mm_setzero_si128();             \
+    xmm2 = _mm_unpackhi_epi8(xmm2, xmm0);   \
+    xmm6 = _mm_unpackhi_epi8(xmm6, xmm1);   \
+    xmm0 = xmm6;                            \
+    xmm6 = _mm_unpacklo_epi16(xmm6, xmm2);  \
+    _mm_storeu_si128((__m128i*)(p_buffer+8), xmm6); \
+    xmm0 = _mm_unpackhi_epi16(xmm0, xmm2);  \
+    _mm_storeu_si128((__m128i*)(p_buffer+12), xmm0);
+
+#define SSE2_UNPACK_32_ABGR_ALIGNED         \
+    ;
+
+#define SSE2_UNPACK_32_ABGR_UNALIGNED       \
+    ;
 
+#endif
+
+#endif