Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6

issues on Win64.

Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk

7 files changed, 637 insertions, 701 deletions

diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile
index 43119fcc6f..924b1789b0 100644
--- a/libavcodec/x86/Makefile
+++ b/libavcodec/x86/Makefile
@@ -26,15 +26,12 @@ MMX-OBJS-$(CONFIG_GPL)                 += x86/idct_mmx.o
 MMX-OBJS-$(CONFIG_LPC)                 += x86/lpc_mmx.o
 MMX-OBJS-$(CONFIG_DWT)                 += x86/snowdsp_mmx.o
 MMX-OBJS-$(CONFIG_VC1_DECODER)         += x86/vc1dsp_mmx.o
-MMX-OBJS-$(CONFIG_VP3_DECODER)         += x86/vp3dsp_mmx.o              \
-                                          x86/vp3dsp_sse2.o
-MMX-OBJS-$(CONFIG_VP5_DECODER)         += x86/vp3dsp_mmx.o              \
-                                          x86/vp3dsp_sse2.o             \
-                                          x86/vp56dsp_init.o
-YASM-OBJS-$(CONFIG_VP6_DECODER)        += x86/vp56dsp.o
-MMX-OBJS-$(CONFIG_VP6_DECODER)         += x86/vp3dsp_mmx.o              \
-                                          x86/vp3dsp_sse2.o             \
-                                          x86/vp56dsp_init.o
+YASM-OBJS-$(CONFIG_VP3_DECODER)        += x86/vp3dsp.o
+YASM-OBJS-$(CONFIG_VP5_DECODER)        += x86/vp3dsp.o
+MMX-OBJS-$(CONFIG_VP5_DECODER)         += x86/vp56dsp_init.o
+YASM-OBJS-$(CONFIG_VP6_DECODER)        += x86/vp3dsp.o                  \
+                                          x86/vp56dsp.o
+MMX-OBJS-$(CONFIG_VP6_DECODER)         += x86/vp56dsp_init.o
 YASM-OBJS-$(CONFIG_VP8_DECODER)        += x86/vp8dsp.o
 MMX-OBJS-$(CONFIG_VP8_DECODER)         += x86/vp8dsp-init.o
 MMX-OBJS-$(HAVE_YASM)                  += x86/dsputil_yasm.o            \
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index 3bdcc9b092..9e064caf89 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -28,8 +28,6 @@
 #include "libavcodec/mpegvideo.h"
 #include "libavcodec/simple_idct.h"
 #include "dsputil_mmx.h"
-#include "vp3dsp_mmx.h"
-#include "vp3dsp_sse2.h"
 #include "idct_xvid.h"
 
 //#undef NDEBUG
@@ -2376,6 +2374,19 @@ static void float_to_int16_sse2(int16_t *dst, const float *src, long len){
     );
 }
 
+void ff_vp3_idct_mmx(int16_t *input_data);
+void ff_vp3_idct_put_mmx(uint8_t *dest, int line_size, DCTELEM *block);
+void ff_vp3_idct_add_mmx(uint8_t *dest, int line_size, DCTELEM *block);
+
+void ff_vp3_idct_dc_add_mmx2(uint8_t *dest, int line_size, const DCTELEM *block);
+
+void ff_vp3_v_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values);
+void ff_vp3_h_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values);
+
+void ff_vp3_idct_sse2(int16_t *input_data);
+void ff_vp3_idct_put_sse2(uint8_t *dest, int line_size, DCTELEM *block);
+void ff_vp3_idct_add_sse2(uint8_t *dest, int line_size, DCTELEM *block);
+
 void ff_float_to_int16_interleave6_sse(int16_t *dst, const float **src, int len);
 void ff_float_to_int16_interleave6_3dnow(int16_t *dst, const float **src, int len);
 void ff_float_to_int16_interleave6_3dn2(int16_t *dst, const float **src, int len);
diff --git a/libavcodec/x86/vp3dsp.asm b/libavcodec/x86/vp3dsp.asm
new file mode 100644
index 0000000000..95ea561ed0
--- /dev/null
+++ b/libavcodec/x86/vp3dsp.asm
@@ -0,0 +1,618 @@
+;******************************************************************************
+;* MMX/SSE2-optimized functions for the VP3 decoder
+;* Copyright (c) 2007 Aurelien Jacobs <aurel@gnuage.org>
+;*
+;* This file is part of FFmpeg.
+;*
+;* FFmpeg is free software; you can redistribute it and/or
+;* modify it under the terms of the GNU Lesser General Public
+;* License as published by the Free Software Foundation; either
+;* version 2.1 of the License, or (at your option) any later version.
+;*
+;* FFmpeg is distributed in the hope that it will be useful,
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+;* Lesser General Public License for more details.
+;*
+;* You should have received a copy of the GNU Lesser General Public
+;* License along with FFmpeg; if not, write to the Free Software
+;* 51, Inc., Foundation Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+;******************************************************************************
+
+%include "x86inc.asm"
+%include "x86util.asm"
+
+; MMX-optimized functions cribbed from the original VP3 source code.
+
+SECTION_RODATA
+
+vp3_idct_data: times 8 dw 64277
+               times 8 dw 60547
+               times 8 dw 54491
+               times 8 dw 46341
+               times 8 dw 36410
+               times 8 dw 25080
+               times 8 dw 12785
+
+cextern pb_1
+cextern pb_3
+cextern pb_7
+cextern pb_1F
+cextern pb_81
+
+cextern pw_8
+
+cextern put_signed_pixels_clamped_mmx
+cextern add_pixels_clamped_mmx
+
+SECTION .text
+
+; this is off by one or two for some cases when filter_limit is greater than 63
+; in:  p0 in mm6, p1 in mm4, p2 in mm2, p3 in mm1
+; out: p1 in mm4, p2 in mm3
+%macro VP3_LOOP_FILTER 0
+    movq          m7, m6
+    pand          m6, [pb_7]    ; p0&7
+    psrlw         m7, 3
+    pand          m7, [pb_1F]   ; p0>>3
+    movq          m3, m2        ; p2
+    pxor          m2, m4
+    pand          m2, [pb_1]    ; (p2^p1)&1
+    movq          m5, m2
+    paddb         m2, m2
+    paddb         m2, m5        ; 3*(p2^p1)&1
+    paddb         m2, m6        ; extra bits lost in shifts
+    pcmpeqb       m0, m0
+    pxor          m1, m0        ; 255 - p3
+    pavgb         m1, m2        ; (256 - p3 + extrabits) >> 1
+    pxor          m0, m4        ; 255 - p1
+    pavgb         m0, m3        ; (256 + p2-p1) >> 1
+    paddb         m1, [pb_3]
+    pavgb         m1, m0        ; 128+2+(   p2-p1  - p3) >> 2
+    pavgb         m1, m0        ; 128+1+(3*(p2-p1) - p3) >> 3
+    paddusb       m7, m1        ; d+128+1
+    movq          m6, [pb_81]
+    psubusb       m6, m7
+    psubusb       m7, [pb_81]
+
+    movq          m5, [r2+516]  ; flim
+    pminub        m6, m5
+    pminub        m7, m5
+    movq          m0, m6
+    movq          m1, m7
+    paddb         m6, m6
+    paddb         m7, m7
+    pminub        m6, m5
+    pminub        m7, m5
+    psubb         m6, m0
+    psubb         m7, m1
+    paddusb       m4, m7
+    psubusb       m4, m6
+    psubusb       m3, m7
+    paddusb       m3, m6
+%endmacro
+
+%macro STORE_4_WORDS 1
+    movd          r2, %1
+    mov  [r0     -1], r2w
+    psrlq         %1, 32
+    shr           r2, 16
+    mov  [r0+r1  -1], r2w
+    movd          r2, %1
+    mov  [r0+r1*2-1], r2w
+    shr           r2, 16
+    mov  [r0+r3  -1], r2w
+%endmacro
+
+INIT_MMX
+cglobal vp3_v_loop_filter_mmx2, 3, 4
+%ifdef ARCH_X86_64
+    movsxd        r1, r1d
+%endif
+    mov           r3, r1
+    neg           r1
+    movq          m6, [r0+r1*2]
+    movq          m4, [r0+r1  ]
+    movq          m2, [r0     ]
+    movq          m1, [r0+r3  ]
+
+    VP3_LOOP_FILTER
+
+    movq     [r0+r1], m4
+    movq     [r0   ], m3
+    RET
+
+cglobal vp3_h_loop_filter_mmx2, 3, 4
+%ifdef ARCH_X86_64
+    movsxd        r1, r1d
+%endif
+    lea           r3, [r1*3]
+
+    movd          m6, [r0     -2]
+    movd          m4, [r0+r1  -2]
+    movd          m2, [r0+r1*2-2]
+    movd          m1, [r0+r3  -2]
+    lea           r0, [r0+r1*4  ]
+    punpcklbw     m6, [r0     -2]
+    punpcklbw     m4, [r0+r1  -2]
+    punpcklbw     m2, [r0+r1*2-2]
+    punpcklbw     m1, [r0+r3  -2]
+    sub           r0, r3
+    sub           r0, r1
+
+    TRANSPOSE4x4B  6, 4, 2, 1, 0
+    VP3_LOOP_FILTER
+    SBUTTERFLY    bw, 4, 3, 5
+
+    STORE_4_WORDS m4
+    lea           r0, [r0+r1*4  ]
+    STORE_4_WORDS m3
+    RET
+
+; from original comments: The Macro does IDct on 4 1-D Dcts
+%macro BeginIDCT 0
+    movq          m2, I(3)
+    movq          m6, C(3)
+    movq          m4, m2
+    movq          m7, J(5)
+    pmulhw        m4, m6        ; r4 = c3*i3 - i3
+    movq          m1, C(5)
+    pmulhw        m6, m7        ; r6 = c3*i5 - i5
+    movq          m5, m1
+    pmulhw        m1, m2        ; r1 = c5*i3 - i3
+    movq          m3, I(1)
+    pmulhw        m5, m7        ; r5 = c5*i5 - i5
+    movq          m0, C(1)
+    paddw         m4, m2        ; r4 = c3*i3
+    paddw         m6, m7        ; r6 = c3*i5
+    paddw         m2, m1        ; r2 = c5*i3
+    movq          m1, J(7)
+    paddw         m7, m5        ; r7 = c5*i5
+    movq          m5, m0        ; r5 = c1
+    pmulhw        m0, m3        ; r0 = c1*i1 - i1
+    paddsw        m4, m7        ; r4 = C = c3*i3 + c5*i5
+    pmulhw        m5, m1        ; r5 = c1*i7 - i7
+    movq          m7, C(7)
+    psubsw        m6, m2        ; r6 = D = c3*i5 - c5*i3
+    paddw         m0, m3        ; r0 = c1*i1
+    pmulhw        m3, m7        ; r3 = c7*i1
+    movq          m2, I(2)
+    pmulhw        m7, m1        ; r7 = c7*i7
+    paddw         m5, m1        ; r5 = c1*i7
+    movq          m1, m2        ; r1 = i2
+    pmulhw        m2, C(2)      ; r2 = c2*i2 - i2
+    psubsw        m3, m5        ; r3 = B = c7*i1 - c1*i7
+    movq          m5, J(6)
+    paddsw        m0, m7        ; r0 = A = c1*i1 + c7*i7
+    movq          m7, m5        ; r7 = i6
+    psubsw        m0, m4        ; r0 = A - C
+    pmulhw        m5, C(2)      ; r5 = c2*i6 - i6
+    paddw         m2, m1        ; r2 = c2*i2
+    pmulhw        m1, C(6)      ; r1 = c6*i2
+    paddsw        m4, m4        ; r4 = C + C
+    paddsw        m4, m0        ; r4 = C. = A + C
+    psubsw        m3, m6        ; r3 = B - D
+    paddw         m5, m7        ; r5 = c2*i6
+    paddsw        m6, m6        ; r6 = D + D
+    pmulhw        m7, C(6)      ; r7 = c6*i6
+    paddsw        m6, m3        ; r6 = D. = B + D
+    movq        I(1), m4        ; save C. at I(1)
+    psubsw        m1, m5        ; r1 = H = c6*i2 - c2*i6
+    movq          m4, C(4)
+    movq          m5, m3        ; r5 = B - D
+    pmulhw        m3, m4        ; r3 = (c4 - 1) * (B - D)
+    paddsw        m7, m2        ; r3 = (c4 - 1) * (B - D)
+    movq        I(2), m6        ; save D. at I(2)
+    movq          m2, m0        ; r2 = A - C
+    movq          m6, I(0)
+    pmulhw        m0, m4        ; r0 = (c4 - 1) * (A - C)
+    paddw         m5, m3        ; r5 = B. = c4 * (B - D)
+    movq          m3, J(4)
+    psubsw        m5, m1        ; r5 = B.. = B. - H
+    paddw         m2, m0        ; r0 = A. = c4 * (A - C)
+    psubsw        m6, m3        ; r6 = i0 - i4
+    movq          m0, m6
+    pmulhw        m6, m4        ; r6 = (c4 - 1) * (i0 - i4)
+    paddsw        m3, m3        ; r3 = i4 + i4
+    paddsw        m1, m1        ; r1 = H + H
+    paddsw        m3, m0        ; r3 = i0 + i4
+    paddsw        m1, m5        ; r1 = H. = B + H
+    pmulhw        m4, m3        ; r4 = (c4 - 1) * (i0 + i4)
+    paddsw        m6, m0        ; r6 = F = c4 * (i0 - i4)
+    psubsw        m6, m2        ; r6 = F. = F - A.
+    paddsw        m2, m2        ; r2 = A. + A.
+    movq          m0, I(1)      ; r0 = C.
+    paddsw        m2, m6        ; r2 = A.. = F + A.
+    paddw         m4, m3        ; r4 = E = c4 * (i0 + i4)
+    psubsw        m2, m1        ; r2 = R2 = A.. - H.
+%endmacro
+
+; RowIDCT gets ready to transpose
+%macro RowIDCT 0
+    BeginIDCT
+    movq          m3, I(2)      ; r3 = D.
+    psubsw        m4, m7        ; r4 = E. = E - G
+    paddsw        m1, m1        ; r1 = H. + H.
+    paddsw        m7, m7        ; r7 = G + G
+    paddsw        m1, m2        ; r1 = R1 = A.. + H.
+    paddsw        m7, m4        ; r1 = R1 = A.. + H.
+    psubsw        m4, m3        ; r4 = R4 = E. - D.
+    paddsw        m3, m3
+    psubsw        m6, m5        ; r6 = R6 = F. - B..
+    paddsw        m5, m5
+    paddsw        m3, m4        ; r3 = R3 = E. + D.
+    paddsw        m5, m6        ; r5 = R5 = F. + B..
+    psubsw        m7, m0        ; r7 = R7 = G. - C.
+    paddsw        m0, m0
+    movq        I(1), m1        ; save R1
+    paddsw        m0, m7        ; r0 = R0 = G. + C.
+%endmacro
+
+; Column IDCT normalizes and stores final results
+%macro ColumnIDCT 0
+    BeginIDCT
+    paddsw        m2, OC_8      ; adjust R2 (and R1) for shift
+    paddsw        m1, m1        ; r1 = H. + H.
+    paddsw        m1, m2        ; r1 = R1 = A.. + H.
+    psraw         m2, 4         ; r2 = NR2
+    psubsw        m4, m7        ; r4 = E. = E - G
+    psraw         m1, 4         ; r1 = NR2
+    movq          m3, I(2)      ; r3 = D.
+    paddsw        m7, m7        ; r7 = G + G
+    movq        I(2), m2        ; store NR2 at I2
+    paddsw        m7, m4        ; r7 = G. = E + G
+    movq        I(1), m1        ; store NR1 at I1
+    psubsw        m4, m3        ; r4 = R4 = E. - D.
+    paddsw        m4, OC_8      ; adjust R4 (and R3) for shift
+    paddsw        m3, m3        ; r3 = D. + D.
+    paddsw        m3, m4        ; r3 = R3 = E. + D.
+    psraw         m4, 4         ; r4 = NR4
+    psubsw        m6, m5        ; r6 = R6 = F. - B..
+    psraw         m3, 4         ; r3 = NR3
+    paddsw        m6, OC_8      ; adjust R6 (and R5) for shift
+    paddsw        m5, m5        ; r5 = B.. + B..
+    paddsw        m5, m6        ; r5 = R5 = F. + B..
+    psraw         m6, 4         ; r6 = NR6
+    movq        J(4), m4        ; store NR4 at J4
+    psraw         m5, 4         ; r5 = NR5
+    movq        I(3), m3        ; store NR3 at I3
+    psubsw        m7, m0        ; r7 = R7 = G. - C.
+    paddsw        m7, OC_8      ; adjust R7 (and R0) for shift
+    paddsw        m0, m0        ; r0 = C. + C.
+    paddsw        m0, m7        ; r0 = R0 = G. + C.
+    psraw         m7, 4         ; r7 = NR7
+    movq        J(6), m6        ; store NR6 at J6
+    psraw         m0, 4         ; r0 = NR0
+    movq        J(5), m5        ; store NR5 at J5
+    movq        J(7), m7        ; store NR7 at J7
+    movq        I(0), m0        ; store NR0 at I0
+%endmacro
+
+; Following macro does two 4x4 transposes in place.
+;
+; At entry (we assume):
+;
+;   r0 = a3 a2 a1 a0
+;   I(1) = b3 b2 b1 b0
+;   r2 = c3 c2 c1 c0
+;   r3 = d3 d2 d1 d0
+;
+;   r4 = e3 e2 e1 e0
+;   r5 = f3 f2 f1 f0
+;   r6 = g3 g2 g1 g0
+;   r7 = h3 h2 h1 h0
+;
+; At exit, we have:
+;
+;   I(0) = d0 c0 b0 a0
+;   I(1) = d1 c1 b1 a1
+;   I(2) = d2 c2 b2 a2
+;   I(3) = d3 c3 b3 a3
+;
+;   J(4) = h0 g0 f0 e0
+;   J(5) = h1 g1 f1 e1
+;   J(6) = h2 g2 f2 e2
+;   J(7) = h3 g3 f3 e3
+;
+;  I(0) I(1) I(2) I(3)  is the transpose of r0 I(1) r2 r3.
+;  J(4) J(5) J(6) J(7)  is the transpose of r4 r5 r6 r7.
+;
+;  Since r1 is free at entry, we calculate the Js first.
+%macro Transpose 0
+    movq          m1, m4        ; r1 = e3 e2 e1 e0
+    punpcklwd     m4, m5        ; r4 = f1 e1 f0 e0
+    movq        I(0), m0        ; save a3 a2 a1 a0
+    punpckhwd     m1, m5        ; r1 = f3 e3 f2 e2
+    movq          m0, m6        ; r0 = g3 g2 g1 g0
+    punpcklwd     m6, m7        ; r6 = h1 g1 h0 g0
+    movq          m5, m4        ; r5 = f1 e1 f0 e0
+    punpckldq     m4, m6        ; r4 = h0 g0 f0 e0 = R4
+    punpckhdq     m5, m6        ; r5 = h1 g1 f1 e1 = R5
+    movq          m6, m1        ; r6 = f3 e3 f2 e2
+    movq        J(4), m4
+    punpckhwd     m0, m7        ; r0 = h3 g3 h2 g2
+    movq        J(5), m5
+    punpckhdq     m6, m0        ; r6 = h3 g3 f3 e3 = R7
+    movq          m4, I(0)      ; r4 = a3 a2 a1 a0
+    punpckldq     m1, m0        ; r1 = h2 g2 f2 e2 = R6
+    movq          m5, I(1)      ; r5 = b3 b2 b1 b0
+    movq          m0, m4        ; r0 = a3 a2 a1 a0
+    movq        J(7), m6
+    punpcklwd     m0, m5        ; r0 = b1 a1 b0 a0
+    movq        J(6), m1
+    punpckhwd     m4, m5        ; r4 = b3 a3 b2 a2
+    movq          m5, m2        ; r5 = c3 c2 c1 c0
+    punpcklwd     m2, m3        ; r2 = d1 c1 d0 c0
+    movq          m1, m0        ; r1 = b1 a1 b0 a0
+    punpckldq     m0, m2        ; r0 = d0 c0 b0 a0 = R0
+    punpckhdq     m1, m2        ; r1 = d1 c1 b1 a1 = R1
+    movq          m2, m4        ; r2 = b3 a3 b2 a2
+    movq        I(0), m0
+    punpckhwd     m5, m3        ; r5 = d3 c3 d2 c2
+    movq        I(1), m1
+    punpckhdq     m4, m5        ; r4 = d3 c3 b3 a3 = R3
+    punpckldq     m2, m5        ; r2 = d2 c2 b2 a2 = R2
+    movq        I(3), m4
+    movq        I(2), m2
+%endmacro
+
+%macro VP3_IDCT_mmx 1
+    ; eax = quantized input
+    ; ebx = dequantizer matrix
+    ; ecx = IDCT constants
+    ;  M(I) = ecx + MaskOffset(0) + I * 8
+    ;  C(I) = ecx + CosineOffset(32) + (I-1) * 8
+    ; edx = output
+    ; r0..r7 = mm0..mm7
+%define OC_8 [pw_8]
+%define C(x) [vp3_idct_data+16*(x-1)]
+
+    ; at this point, function has completed dequantization + dezigzag +
+    ; partial transposition; now do the idct itself
+%define I(x) [%1+16* x     ]
+%define J(x) [%1+16*(x-4)+8]
+    RowIDCT
+    Transpose
+
+%define I(x) [%1+16* x   +64]
+%define J(x) [%1+16*(x-4)+72]
+    RowIDCT
+    Transpose
+
+%define I(x) [%1+16*x]
+%define J(x) [%1+16*x]
+    ColumnIDCT
+
+%define I(x) [%1+16*x+8]
+%define J(x) [%1+16*x+8]
+    ColumnIDCT
+%endmacro
+
+%macro VP3_1D_IDCT_SSE2 0
+    movdqa        m2, I(3)      ; xmm2 = i3
+    movdqa        m6, C(3)      ; xmm6 = c3
+    movdqa        m4, m2        ; xmm4 = i3
+    movdqa        m7, I(5)      ; xmm7 = i5
+    pmulhw        m4, m6        ; xmm4 = c3 * i3 - i3
+    movdqa        m1, C(5)      ; xmm1 = c5
+    pmulhw        m6, m7        ; xmm6 = c3 * i5 - i5
+    movdqa        m5, m1        ; xmm5 = c5
+    pmulhw        m1, m2        ; xmm1 = c5 * i3 - i3
+    movdqa        m3, I(1)      ; xmm3 = i1
+    pmulhw        m5, m7        ; xmm5 = c5 * i5 - i5
+    movdqa        m0, C(1)      ; xmm0 = c1
+    paddw         m4, m2        ; xmm4 = c3 * i3
+    paddw         m6, m7        ; xmm6 = c3 * i5
+    paddw         m2, m1        ; xmm2 = c5 * i3
+    movdqa        m1, I(7)      ; xmm1 = i7
+    paddw         m7, m5        ; xmm7 = c5 * i5
+    movdqa        m5, m0        ; xmm5 = c1
+    pmulhw        m0, m3        ; xmm0 = c1 * i1 - i1
+    paddsw        m4, m7        ; xmm4 = c3 * i3 + c5 * i5 = C
+    pmulhw        m5, m1        ; xmm5 = c1 * i7 - i7
+    movdqa        m7, C(7)      ; xmm7 = c7
+    psubsw        m6, m2        ; xmm6 = c3 * i5 - c5 * i3 = D
+    paddw         m0, m3        ; xmm0 = c1 * i1
+    pmulhw        m3, m7        ; xmm3 = c7 * i1
+    movdqa        m2, I(2)      ; xmm2 = i2
+    pmulhw        m7, m1        ; xmm7 = c7 * i7
+    paddw         m5, m1        ; xmm5 = c1 * i7
+    movdqa        m1, m2        ; xmm1 = i2
+    pmulhw        m2, C(2)      ; xmm2 = i2 * c2 -i2
+    psubsw        m3, m5        ; xmm3 = c7 * i1 - c1 * i7 = B
+    movdqa        m5, I(6)      ; xmm5 = i6
+    paddsw        m0, m7        ; xmm0 = c1 * i1 + c7 * i7 = A
+    movdqa        m7, m5        ; xmm7 = i6
+    psubsw        m0, m4        ; xmm0 = A - C
+    pmulhw        m5, C(2)      ; xmm5 = c2 * i6 - i6
+    paddw         m2, m1        ; xmm2 = i2 * c2
+    pmulhw        m1, C(6)      ; xmm1 = c6 * i2
+    paddsw        m4, m4        ; xmm4 = C + C
+    paddsw        m4, m0        ; xmm4 = A + C = C.
+    psubsw        m3, m6        ; xmm3 = B - D
+    paddw         m5, m7        ; xmm5 = c2 * i6
+    paddsw        m6, m6        ; xmm6 = D + D
+    pmulhw        m7, C(6)      ; xmm7 = c6 * i6
+    paddsw        m6, m3        ; xmm6 = B + D = D.
+    movdqa      I(1), m4        ; Save C. at I(1)
+    psubsw        m1, m5        ; xmm1 = c6 * i2 - c2 * i6 = H
+    movdqa        m4, C(4)      ; xmm4 = C4
+    movdqa        m5, m3        ; xmm5 = B - D
+    pmulhw        m3, m4        ; xmm3 = ( c4 -1 ) * ( B - D )
+    paddsw        m7, m2        ; xmm7 = c2 * i2 + c6 * i6 = G
+    movdqa      I(2), m6        ; save D. at I(2)
+    movdqa        m2, m0        ; xmm2 = A - C
+    movdqa        m6, I(0)      ; xmm6 = i0
+    pmulhw        m0, m4        ; xmm0 = ( c4 - 1 ) * ( A - C ) = A.
+    paddw         m5, m3        ; xmm5 = c4 * ( B - D ) = B.
+    movdqa        m3, I(4)      ; xmm3 = i4
+    psubsw        m5, m1        ; xmm5 = B. - H = B..
+    paddw         m2, m0        ; xmm2 = c4 * ( A - C) = A.
+    psubsw        m6, m3        ; xmm6 = i0 - i4
+    movdqa        m0, m6        ; xmm0 = i0 - i4
+    pmulhw        m6, m4        ; xmm6 = (c4 - 1) * (i0 - i4) = F
+    paddsw        m3, m3        ; xmm3 = i4 + i4
+    paddsw        m1, m1        ; xmm1 = H + H
+    paddsw        m3, m0        ; xmm3 = i0 + i4
+    paddsw        m1, m5        ; xmm1 = B. + H = H.
+    pmulhw        m4, m3        ; xmm4 = ( c4 - 1 ) * ( i0 + i4 )
+    paddw         m6, m0        ; xmm6 = c4 * ( i0 - i4 )
+    psubsw        m6, m2        ; xmm6 = F - A. = F.
+    paddsw        m2, m2        ; xmm2 = A. + A.
+    movdqa        m0, I(1)      ; Load        C. from I(1)
+    paddsw        m2, m6        ; xmm2 = F + A. = A..
+    paddw         m4, m3        ; xmm4 = c4 * ( i0 + i4 ) = 3
+    psubsw        m2, m1        ; xmm2 = A.. - H. = R2
+    ADD(m2)                     ; Adjust R2 and R1 before shifting
+    paddsw        m1, m1        ; xmm1 = H. + H.
+    paddsw        m1, m2        ; xmm1 = A.. + H. = R1
+    SHIFT(m2)                   ; xmm2 = op2
+    psubsw        m4, m7        ; xmm4 = E - G = E.
+    SHIFT(m1)                   ; xmm1 = op1
+    movdqa        m3, I(2)      ; Load D. from I(2)
+    paddsw        m7, m7        ; xmm7 = G + G
+    paddsw        m7, m4        ; xmm7 = E + G = G.
+    psubsw        m4, m3        ; xmm4 = E. - D. = R4
+    ADD(m4)                     ; Adjust R4 and R3 before shifting
+    paddsw        m3, m3        ; xmm3 = D. + D.
+    paddsw        m3, m4        ; xmm3 = E. + D. = R3
+    SHIFT(m4)                   ; xmm4 = op4
+    psubsw        m6, m5        ; xmm6 = F. - B..= R6
+    SHIFT(m3)                   ; xmm3 = op3
+    ADD(m6)                     ; Adjust R6 and R5 before shifting
+    paddsw        m5, m5        ; xmm5 = B.. + B..
+    paddsw        m5, m6        ; xmm5 = F. + B.. = R5
+    SHIFT(m6)                   ; xmm6 = op6
+    SHIFT(m5)                   ; xmm5 = op5
+    psubsw        m7, m0        ; xmm7 = G. - C. = R7
+    ADD(m7)                     ; Adjust R7 and R0 before shifting
+    paddsw        m0, m0        ; xmm0 = C. + C.
+    paddsw        m0, m7        ; xmm0 = G. + C.
+    SHIFT(m7)                   ; xmm7 = op7
+    SHIFT(m0)                   ; xmm0 = op0
+%endmacro
+
+%macro PUT_BLOCK 8
+    movdqa      O(0), m%1
+    movdqa      O(1), m%2
+    movdqa      O(2), m%3
+    movdqa      O(3), m%4
+    movdqa      O(4), m%5
+    movdqa      O(5), m%6
+    movdqa      O(6), m%7
+    movdqa      O(7), m%8
+%endmacro
+
+%macro VP3_IDCT_sse2 1
+%define I(x) [%1+16*x]
+%define O(x) [%1+16*x]
+%define C(x) [vp3_idct_data+16*(x-1)]
+%define SHIFT(x)
+%define ADD(x)
+        VP3_1D_IDCT_SSE2
+%ifdef ARCH_X86_64
+        TRANSPOSE8x8W 0, 1, 2, 3, 4, 5, 6, 7, 8
+%else
+        TRANSPOSE8x8W 0, 1, 2, 3, 4, 5, 6, 7, [%1], [%1+16]
+%endif
+        PUT_BLOCK 0, 1, 2, 3, 4, 5, 6, 7
+
+%define SHIFT(x) psraw  x, 4
+%define ADD(x)   paddsw x, [pw_8]
+        VP3_1D_IDCT_SSE2
+        PUT_BLOCK 0, 1, 2, 3, 4, 5, 6, 7
+%endmacro
+
+%macro vp3_idct_funcs 3
+cglobal vp3_idct_%1, 1, 1, %2
+    VP3_IDCT_%1   r0
+    RET
+
+cglobal vp3_idct_put_%1, 3, %3, %2
+    VP3_IDCT_%1   r2
+%ifdef ARCH_X86_64
+    mov           r3, r2
+    mov           r2, r1
+    mov           r1, r0
+    mov           r0, r3
+%else
+    mov          r0m, r2
+    mov          r1m, r0
+    mov          r2m, r1
+%endif
+%ifdef WIN64
+    call put_signed_pixels_clamped_mmx
+    RET
+%else
+    jmp put_signed_pixels_clamped_mmx
+%endif
+
+cglobal vp3_idct_add_%1, 3, %3, %2
+    VP3_IDCT_%1   r2
+%ifdef ARCH_X86_64
+    mov           r3, r2
+    mov           r2, r1
+    mov           r1, r0
+    mov           r0, r3
+%else
+    mov          r0m, r2
+    mov          r1m, r0
+    mov          r2m, r1
+%endif
+%ifdef WIN64
+    call add_pixels_clamped_mmx
+    RET
+%else
+    jmp add_pixels_clamped_mmx
+%endif
+%endmacro
+
+%ifdef ARCH_X86_64
+%define REGS 4
+%else
+%define REGS 3
+%endif
+INIT_MMX
+vp3_idct_funcs mmx,  0, REGS
+INIT_XMM
+vp3_idct_funcs sse2, 9, REGS
+%undef REGS
+
+%macro DC_ADD 0
+    movq          m2, [r0     ]
+    movq          m3, [r0+r1  ]
+    paddusb       m2, m0
+    movq          m4, [r0+r1*2]
+    paddusb       m3, m0
+    movq          m5, [r0+r3  ]
+    paddusb       m4, m0
+    paddusb       m5, m0
+    psubusb       m2, m1
+    psubusb       m3, m1
+    movq   [r0     ], m2
+    psubusb       m4, m1
+    movq   [r0+r1  ], m3
+    psubusb       m5, m1
+    movq   [r0+r1*2], m4
+    movq   [r0+r3  ], m5
+%endmacro
+
+INIT_MMX
+cglobal vp3_idct_dc_add_mmx2, 3, 4
+%ifdef ARCH_X86_64
+    movsxd        r1, r1d
+%endif
+    lea           r3, [r1*3]
+    movsx         r2, word [r2]
+    add           r2, 15
+    sar           r2, 5
+    movd          m0, r2
+    pshufw        m0, m0, 0x0
+    pxor          m1, m1
+    psubw         m1, m0
+    packuswb      m0, m0
+    packuswb      m1, m1
+    DC_ADD
+    lea           r0, [r0+r1*4]
+    DC_ADD
+    RET
diff --git a/libavcodec/x86/vp3dsp_mmx.c b/libavcodec/x86/vp3dsp_mmx.c
deleted file mode 100644
index 33119fb53f..0000000000
--- a/libavcodec/x86/vp3dsp_mmx.c
+++ /dev/null
@@ -1,436 +0,0 @@
-/*
- * Copyright (C) 2004 the ffmpeg project
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * FFmpeg is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-/**
- * @file
- * MMX-optimized functions cribbed from the original VP3 source code.
- */
-
-#include "libavutil/x86_cpu.h"
-#include "libavcodec/dsputil.h"
-#include "dsputil_mmx.h"
-#include "vp3dsp_mmx.h"
-
-extern const uint16_t ff_vp3_idct_data[];
-
-// this is off by one or two for some cases when filter_limit is greater than 63
-// in:  p0 in mm6, p1 in mm4, p2 in mm2, p3 in mm1
-// out: p1 in mm4, p2 in mm3
-#define VP3_LOOP_FILTER(flim) \
-    "movq       %%mm6, %%mm7 \n\t" \
-    "pand    "MANGLE(ff_pb_7 )", %%mm6 \n\t" /* p0&7 */ \
-    "psrlw         $3, %%mm7 \n\t" \
-    "pand    "MANGLE(ff_pb_1F)", %%mm7 \n\t" /* p0>>3 */ \
-    "movq       %%mm2, %%mm3 \n\t" /* mm3 = p2 */ \
-    "pxor       %%mm4, %%mm2 \n\t" \
-    "pand    "MANGLE(ff_pb_1 )", %%mm2 \n\t" /* (p2^p1)&1 */ \
-    "movq       %%mm2, %%mm5 \n\t" \
-    "paddb      %%mm2, %%mm2 \n\t" \
-    "paddb      %%mm5, %%mm2 \n\t" /* 3*(p2^p1)&1 */ \
-    "paddb      %%mm6, %%mm2 \n\t" /* extra bits lost in shifts */ \
-    "pcmpeqb    %%mm0, %%mm0 \n\t" \
-    "pxor       %%mm0, %%mm1 \n\t" /* 255 - p3 */ \
-    "pavgb      %%mm2, %%mm1 \n\t" /* (256 - p3 + extrabits) >> 1 */ \
-    "pxor       %%mm4, %%mm0 \n\t" /* 255 - p1 */ \
-    "pavgb      %%mm3, %%mm0 \n\t" /* (256 + p2-p1) >> 1 */ \
-    "paddb   "MANGLE(ff_pb_3 )", %%mm1 \n\t" \
-    "pavgb      %%mm0, %%mm1 \n\t" /* 128+2+(   p2-p1  - p3) >> 2 */ \
-    "pavgb      %%mm0, %%mm1 \n\t" /* 128+1+(3*(p2-p1) - p3) >> 3 */ \
-    "paddusb    %%mm1, %%mm7 \n\t" /* d+128+1 */ \
-    "movq    "MANGLE(ff_pb_81)", %%mm6 \n\t" \
-    "psubusb    %%mm7, %%mm6 \n\t" \
-    "psubusb "MANGLE(ff_pb_81)", %%mm7 \n\t" \
-\
-    "movq     "#flim", %%mm5 \n\t" \
-    "pminub     %%mm5, %%mm6 \n\t" \
-    "pminub     %%mm5, %%mm7 \n\t" \
-    "movq       %%mm6, %%mm0 \n\t" \
-    "movq       %%mm7, %%mm1 \n\t" \
-    "paddb      %%mm6, %%mm6 \n\t" \
-    "paddb      %%mm7, %%mm7 \n\t" \
-    "pminub     %%mm5, %%mm6 \n\t" \
-    "pminub     %%mm5, %%mm7 \n\t" \
-    "psubb      %%mm0, %%mm6 \n\t" \
-    "psubb      %%mm1, %%mm7 \n\t" \
-    "paddusb    %%mm7, %%mm4 \n\t" \
-    "psubusb    %%mm6, %%mm4 \n\t" \
-    "psubusb    %%mm7, %%mm3 \n\t" \
-    "paddusb    %%mm6, %%mm3 \n\t"
-
-#define STORE_4_WORDS(dst0, dst1, dst2, dst3, mm) \
-    "movd "#mm", %0        \n\t" \
-    "movw   %w0, -1"#dst0" \n\t" \
-    "psrlq  $32, "#mm"     \n\t" \
-    "shr    $16, %0        \n\t" \
-    "movw   %w0, -1"#dst1" \n\t" \
-    "movd "#mm", %0        \n\t" \
-    "movw   %w0, -1"#dst2" \n\t" \
-    "shr    $16, %0        \n\t" \
-    "movw   %w0, -1"#dst3" \n\t"
-
-void ff_vp3_v_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values)
-{
-    __asm__ volatile(
-        "movq          %0, %%mm6 \n\t"
-        "movq          %1, %%mm4 \n\t"
-        "movq          %2, %%mm2 \n\t"
-        "movq          %3, %%mm1 \n\t"
-
-        VP3_LOOP_FILTER(%4)
-
-        "movq       %%mm4, %1    \n\t"
-        "movq       %%mm3, %2    \n\t"
-
-        : "+m" (*(uint64_t*)(src - 2*stride)),
-          "+m" (*(uint64_t*)(src - 1*stride)),
-          "+m" (*(uint64_t*)(src + 0*stride)),
-          "+m" (*(uint64_t*)(src + 1*stride))
-        : "m"(*(uint64_t*)(bounding_values+129))
-    );
-}
-
-void ff_vp3_h_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values)
-{
-    x86_reg tmp;
-
-    __asm__ volatile(
-        "movd -2(%1),      %%mm6 \n\t"
-        "movd -2(%1,%3),   %%mm0 \n\t"
-        "movd -2(%1,%3,2), %%mm1 \n\t"
-        "movd -2(%1,%4),   %%mm4 \n\t"
-
-        TRANSPOSE8x4(%%mm6, %%mm0, %%mm1, %%mm4, -2(%2), -2(%2,%3), -2(%2,%3,2), -2(%2,%4), %%mm2)
-        VP3_LOOP_FILTER(%5)
-        SBUTTERFLY(%%mm4, %%mm3, %%mm5, bw, q)
-
-        STORE_4_WORDS((%1), (%1,%3), (%1,%3,2), (%1,%4), %%mm4)
-        STORE_4_WORDS((%2), (%2,%3), (%2,%3,2), (%2,%4), %%mm5)
-
-        : "=&r"(tmp)
-        : "r"(src), "r"(src+4*stride), "r"((x86_reg)stride), "r"((x86_reg)3*stride),
-          "m"(*(uint64_t*)(bounding_values+129))
-        : "memory"
-    );
-}
-
-/* from original comments: The Macro does IDct on 4 1-D Dcts */
-#define BeginIDCT() \
-    "movq   "I(3)", %%mm2 \n\t" \
-    "movq   "C(3)", %%mm6 \n\t" \
-    "movq    %%mm2, %%mm4 \n\t" \
-    "movq   "J(5)", %%mm7 \n\t" \
-    "pmulhw  %%mm6, %%mm4 \n\t"    /* r4 = c3*i3 - i3 */ \
-    "movq   "C(5)", %%mm1 \n\t" \
-    "pmulhw  %%mm7, %%mm6 \n\t"    /* r6 = c3*i5 - i5 */ \
-    "movq    %%mm1, %%mm5 \n\t" \
-    "pmulhw  %%mm2, %%mm1 \n\t"    /* r1 = c5*i3 - i3 */ \
-    "movq   "I(1)", %%mm3 \n\t" \
-    "pmulhw  %%mm7, %%mm5 \n\t"    /* r5 = c5*i5 - i5 */ \
-    "movq   "C(1)", %%mm0 \n\t" \
-    "paddw   %%mm2, %%mm4 \n\t"    /* r4 = c3*i3 */ \
-    "paddw   %%mm7, %%mm6 \n\t"    /* r6 = c3*i5 */ \
-    "paddw   %%mm1, %%mm2 \n\t"    /* r2 = c5*i3 */ \
-    "movq   "J(7)", %%mm1 \n\t" \
-    "paddw   %%mm5, %%mm7 \n\t"    /* r7 = c5*i5 */ \
-    "movq    %%mm0, %%mm5 \n\t"    /* r5 = c1 */ \
-    "pmulhw  %%mm3, %%mm0 \n\t"    /* r0 = c1*i1 - i1 */ \
-    "paddsw  %%mm7, %%mm4 \n\t"    /* r4 = C = c3*i3 + c5*i5 */ \
-    "pmulhw  %%mm1, %%mm5 \n\t"    /* r5 = c1*i7 - i7 */ \
-    "movq   "C(7)", %%mm7 \n\t" \
-    "psubsw  %%mm2, %%mm6 \n\t"    /* r6 = D = c3*i5 - c5*i3 */ \
-    "paddw   %%mm3, %%mm0 \n\t"    /* r0 = c1*i1 */ \
-    "pmulhw  %%mm7, %%mm3 \n\t"    /* r3 = c7*i1 */ \
-    "movq   "I(2)", %%mm2 \n\t" \
-    "pmulhw  %%mm1, %%mm7 \n\t"    /* r7 = c7*i7 */ \
-    "paddw   %%mm1, %%mm5 \n\t"    /* r5 = c1*i7 */ \
-    "movq    %%mm2, %%mm1 \n\t"    /* r1 = i2 */ \
-    "pmulhw "C(2)", %%mm2 \n\t"    /* r2 = c2*i2 - i2 */ \
-    "psubsw  %%mm5, %%mm3 \n\t"    /* r3 = B = c7*i1 - c1*i7 */ \
-    "movq   "J(6)", %%mm5 \n\t" \
-    "paddsw  %%mm7, %%mm0 \n\t"    /* r0 = A = c1*i1 + c7*i7 */ \
-    "movq    %%mm5, %%mm7 \n\t"    /* r7 = i6 */ \
-    "psubsw  %%mm4, %%mm0 \n\t"    /* r0 = A - C */ \
-    "pmulhw "C(2)", %%mm5 \n\t"    /* r5 = c2*i6 - i6 */ \
-    "paddw   %%mm1, %%mm2 \n\t"    /* r2 = c2*i2 */ \
-    "pmulhw "C(6)", %%mm1 \n\t"    /* r1 = c6*i2 */ \
-    "paddsw  %%mm4, %%mm4 \n\t"    /* r4 = C + C */ \
-    "paddsw  %%mm0, %%mm4 \n\t"    /* r4 = C. = A + C */ \
-    "psubsw  %%mm6, %%mm3 \n\t"    /* r3 = B - D */ \
-    "paddw   %%mm7, %%mm5 \n\t"    /* r5 = c2*i6 */ \
-    "paddsw  %%mm6, %%mm6 \n\t"    /* r6 = D + D */ \
-    "pmulhw "C(6)", %%mm7 \n\t"    /* r7 = c6*i6 */ \
-    "paddsw  %%mm3, %%mm6 \n\t"    /* r6 = D. = B + D */ \
-    "movq    %%mm4, "I(1)"\n\t"    /* save C. at I(1) */ \
-    "psubsw  %%mm5, %%mm1 \n\t"    /* r1 = H = c6*i2 - c2*i6 */ \
-    "movq   "C(4)", %%mm4 \n\t" \
-    "movq    %%mm3, %%mm5 \n\t"    /* r5 = B - D */ \
-    "pmulhw  %%mm4, %%mm3 \n\t"    /* r3 = (c4 - 1) * (B - D) */ \
-    "paddsw  %%mm2, %%mm7 \n\t"    /* r3 = (c4 - 1) * (B - D) */ \
-    "movq    %%mm6, "I(2)"\n\t"    /* save D. at I(2) */ \
-    "movq    %%mm0, %%mm2 \n\t"    /* r2 = A - C */ \
-    "movq   "I(0)", %%mm6 \n\t" \
-    "pmulhw  %%mm4, %%mm0 \n\t"    /* r0 = (c4 - 1) * (A - C) */ \
-    "paddw   %%mm3, %%mm5 \n\t"    /* r5 = B. = c4 * (B - D) */ \
-    "movq   "J(4)", %%mm3 \n\t" \
-    "psubsw  %%mm1, %%mm5 \n\t"    /* r5 = B.. = B. - H */ \
-    "paddw   %%mm0, %%mm2 \n\t"    /* r0 = A. = c4 * (A - C) */ \
-    "psubsw  %%mm3, %%mm6 \n\t"    /* r6 = i0 - i4 */ \
-    "movq    %%mm6, %%mm0 \n\t" \
-    "pmulhw  %%mm4, %%mm6 \n\t"    /* r6 = (c4 - 1) * (i0 - i4) */ \
-    "paddsw  %%mm3, %%mm3 \n\t"    /* r3 = i4 + i4 */ \
-    "paddsw  %%mm1, %%mm1 \n\t"    /* r1 = H + H */ \
-    "paddsw  %%mm0, %%mm3 \n\t"    /* r3 = i0 + i4 */ \
-    "paddsw  %%mm5, %%mm1 \n\t"    /* r1 = H. = B + H */ \
-    "pmulhw  %%mm3, %%mm4 \n\t"    /* r4 = (c4 - 1) * (i0 + i4) */ \
-    "paddsw  %%mm0, %%mm6 \n\t"    /* r6 = F = c4 * (i0 - i4) */ \
-    "psubsw  %%mm2, %%mm6 \n\t"    /* r6 = F. = F - A. */ \
-    "paddsw  %%mm2, %%mm2 \n\t"    /* r2 = A. + A. */ \
-    "movq   "I(1)", %%mm0 \n\t"    /* r0 = C. */ \
-    "paddsw  %%mm6, %%mm2 \n\t"    /* r2 = A.. = F + A. */ \
-    "paddw   %%mm3, %%mm4 \n\t"    /* r4 = E = c4 * (i0 + i4) */ \
-    "psubsw  %%mm1, %%mm2 \n\t"    /* r2 = R2 = A.. - H. */
-
-/* RowIDCT gets ready to transpose */
-#define RowIDCT() \
-    BeginIDCT() \
-    "movq   "I(2)", %%mm3 \n\t"    /* r3 = D. */ \
-    "psubsw  %%mm7, %%mm4 \n\t"    /* r4 = E. = E - G */ \
-    "paddsw  %%mm1, %%mm1 \n\t"    /* r1 = H. + H. */ \
-    "paddsw  %%mm7, %%mm7 \n\t"    /* r7 = G + G */ \
-    "paddsw  %%mm2, %%mm1 \n\t"    /* r1 = R1 = A.. + H. */ \
-    "paddsw  %%mm4, %%mm7 \n\t"    /* r1 = R1 = A.. + H. */ \
-    "psubsw  %%mm3, %%mm4 \n\t"    /* r4 = R4 = E. - D. */ \
-    "paddsw  %%mm3, %%mm3 \n\t" \
-    "psubsw  %%mm5, %%mm6 \n\t"    /* r6 = R6 = F. - B.. */ \
-    "paddsw  %%mm5, %%mm5 \n\t" \
-    "paddsw  %%mm4, %%mm3 \n\t"    /* r3 = R3 = E. + D. */ \
-    "paddsw  %%mm6, %%mm5 \n\t"    /* r5 = R5 = F. + B.. */ \
-    "psubsw  %%mm0, %%mm7 \n\t"    /* r7 = R7 = G. - C. */ \
-    "paddsw  %%mm0, %%mm0 \n\t" \
-    "movq    %%mm1, "I(1)"\n\t"    /* save R1 */ \
-    "paddsw  %%mm7, %%mm0 \n\t"    /* r0 = R0 = G. + C. */
-
-/* Column IDCT normalizes and stores final results */
-#define ColumnIDCT() \
-    BeginIDCT() \
-    "paddsw "OC_8", %%mm2 \n\t"    /* adjust R2 (and R1) for shift */ \
-    "paddsw  %%mm1, %%mm1 \n\t"    /* r1 = H. + H. */ \
-    "paddsw  %%mm2, %%mm1 \n\t"    /* r1 = R1 = A.. + H. */ \
-    "psraw      $4, %%mm2 \n\t"    /* r2 = NR2 */ \
-    "psubsw  %%mm7, %%mm4 \n\t"    /* r4 = E. = E - G */ \
-    "psraw      $4, %%mm1 \n\t"    /* r1 = NR1 */ \
-    "movq   "I(2)", %%mm3 \n\t"    /* r3 = D. */ \
-    "paddsw  %%mm7, %%mm7 \n\t"    /* r7 = G + G */ \
-    "movq    %%mm2, "I(2)"\n\t"    /* store NR2 at I2 */ \
-    "paddsw  %%mm4, %%mm7 \n\t"    /* r7 = G. = E + G */ \
-    "movq    %%mm1, "I(1)"\n\t"    /* store NR1 at I1 */ \
-    "psubsw  %%mm3, %%mm4 \n\t"    /* r4 = R4 = E. - D. */ \
-    "paddsw "OC_8", %%mm4 \n\t"    /* adjust R4 (and R3) for shift */ \
-    "paddsw  %%mm3, %%mm3 \n\t"    /* r3 = D. + D. */ \
-    "paddsw  %%mm4, %%mm3 \n\t"    /* r3 = R3 = E. + D. */ \
-    "psraw      $4, %%mm4 \n\t"    /* r4 = NR4 */ \
-    "psubsw  %%mm5, %%mm6 \n\t"    /* r6 = R6 = F. - B.. */ \
-    "psraw      $4, %%mm3 \n\t"    /* r3 = NR3 */ \
-    "paddsw "OC_8", %%mm6 \n\t"    /* adjust R6 (and R5) for shift */ \
-    "paddsw  %%mm5, %%mm5 \n\t"    /* r5 = B.. + B.. */ \
-    "paddsw  %%mm6, %%mm5 \n\t"    /* r5 = R5 = F. + B.. */ \
-    "psraw      $4, %%mm6 \n\t"    /* r6 = NR6 */ \
-    "movq    %%mm4, "J(4)"\n\t"    /* store NR4 at J4 */ \
-    "psraw      $4, %%mm5 \n\t"    /* r5 = NR5 */ \
-    "movq    %%mm3, "I(3)"\n\t"    /* store NR3 at I3 */ \
-    "psubsw  %%mm0, %%mm7 \n\t"    /* r7 = R7 = G. - C. */ \
-    "paddsw "OC_8", %%mm7 \n\t"    /* adjust R7 (and R0) for shift */ \
-    "paddsw  %%mm0, %%mm0 \n\t"    /* r0 = C. + C. */ \
-    "paddsw  %%mm7, %%mm0 \n\t"    /* r0 = R0 = G. + C. */ \
-    "psraw      $4, %%mm7 \n\t"    /* r7 = NR7 */ \
-    "movq    %%mm6, "J(6)"\n\t"    /* store NR6 at J6 */ \
-    "psraw      $4, %%mm0 \n\t"    /* r0 = NR0 */ \
-    "movq    %%mm5, "J(5)"\n\t"    /* store NR5 at J5 */ \
-    "movq    %%mm7, "J(7)"\n\t"    /* store NR7 at J7 */ \
-    "movq    %%mm0, "I(0)"\n\t"    /* store NR0 at I0 */
-
-/* Following macro does two 4x4 transposes in place.
-
-  At entry (we assume):
-
-    r0 = a3 a2 a1 a0
-    I(1) = b3 b2 b1 b0
-    r2 = c3 c2 c1 c0
-    r3 = d3 d2 d1 d0
-
-    r4 = e3 e2 e1 e0
-    r5 = f3 f2 f1 f0
-    r6 = g3 g2 g1 g0
-    r7 = h3 h2 h1 h0
-
-  At exit, we have:
-
-    I(0) = d0 c0 b0 a0
-    I(1) = d1 c1 b1 a1
-    I(2) = d2 c2 b2 a2
-    I(3) = d3 c3 b3 a3
-
-    J(4) = h0 g0 f0 e0
-    J(5) = h1 g1 f1 e1
-    J(6) = h2 g2 f2 e2
-    J(7) = h3 g3 f3 e3
-
-   I(0) I(1) I(2) I(3)  is the transpose of r0 I(1) r2 r3.
-   J(4) J(5) J(6) J(7)  is the transpose of r4 r5 r6 r7.
-
-   Since r1 is free at entry, we calculate the Js first. */
-#define Transpose() \
-    "movq       %%mm4, %%mm1 \n\t"    /* r1 = e3 e2 e1 e0 */ \
-    "punpcklwd  %%mm5, %%mm4 \n\t"    /* r4 = f1 e1 f0 e0 */ \
-    "movq       %%mm0, "I(0)"\n\t"    /* save a3 a2 a1 a0 */ \
-    "punpckhwd  %%mm5, %%mm1 \n\t"    /* r1 = f3 e3 f2 e2 */ \
-    "movq       %%mm6, %%mm0 \n\t"    /* r0 = g3 g2 g1 g0 */ \
-    "punpcklwd  %%mm7, %%mm6 \n\t"    /* r6 = h1 g1 h0 g0 */ \
-    "movq       %%mm4, %%mm5 \n\t"    /* r5 = f1 e1 f0 e0 */ \
-    "punpckldq  %%mm6, %%mm4 \n\t"    /* r4 = h0 g0 f0 e0 = R4 */ \
-    "punpckhdq  %%mm6, %%mm5 \n\t"    /* r5 = h1 g1 f1 e1 = R5 */ \
-    "movq       %%mm1, %%mm6 \n\t"    /* r6 = f3 e3 f2 e2 */ \
-    "movq       %%mm4, "J(4)"\n\t" \
-    "punpckhwd  %%mm7, %%mm0 \n\t"    /* r0 = h3 g3 h2 g2 */ \
-    "movq       %%mm5, "J(5)"\n\t" \
-    "punpckhdq  %%mm0, %%mm6 \n\t"    /* r6 = h3 g3 f3 e3 = R7 */ \
-    "movq      "I(0)", %%mm4 \n\t"    /* r4 = a3 a2 a1 a0 */ \
-    "punpckldq  %%mm0, %%mm1 \n\t"    /* r1 = h2 g2 f2 e2 = R6 */ \
-    "movq      "I(1)", %%mm5 \n\t"    /* r5 = b3 b2 b1 b0 */ \
-    "movq       %%mm4, %%mm0 \n\t"    /* r0 = a3 a2 a1 a0 */ \
-    "movq       %%mm6, "J(7)"\n\t" \
-    "punpcklwd  %%mm5, %%mm0 \n\t"    /* r0 = b1 a1 b0 a0 */ \
-    "movq       %%mm1, "J(6)"\n\t" \
-    "punpckhwd  %%mm5, %%mm4 \n\t"    /* r4 = b3 a3 b2 a2 */ \
-    "movq       %%mm2, %%mm5 \n\t"    /* r5 = c3 c2 c1 c0 */ \
-    "punpcklwd  %%mm3, %%mm2 \n\t"    /* r2 = d1 c1 d0 c0 */ \
-    "movq       %%mm0, %%mm1 \n\t"    /* r1 = b1 a1 b0 a0 */ \
-    "punpckldq  %%mm2, %%mm0 \n\t"    /* r0 = d0 c0 b0 a0 = R0 */ \
-    "punpckhdq  %%mm2, %%mm1 \n\t"    /* r1 = d1 c1 b1 a1 = R1 */ \
-    "movq       %%mm4, %%mm2 \n\t"    /* r2 = b3 a3 b2 a2 */ \
-    "movq       %%mm0, "I(0)"\n\t" \
-    "punpckhwd  %%mm3, %%mm5 \n\t"    /* r5 = d3 c3 d2 c2 */ \
-    "movq       %%mm1, "I(1)"\n\t" \
-    "punpckhdq  %%mm5, %%mm4 \n\t"    /* r4 = d3 c3 b3 a3 = R3 */ \
-    "punpckldq  %%mm5, %%mm2 \n\t"    /* r2 = d2 c2 b2 a2 = R2 */ \
-    "movq       %%mm4, "I(3)"\n\t" \
-    "movq       %%mm2, "I(2)"\n\t"
-
-void ff_vp3_idct_mmx(int16_t *output_data)
-{
-    /* eax = quantized input
-     * ebx = dequantizer matrix
-     * ecx = IDCT constants
-     *  M(I) = ecx + MaskOffset(0) + I * 8
-     *  C(I) = ecx + CosineOffset(32) + (I-1) * 8
-     * edx = output
-     * r0..r7 = mm0..mm7
-     */
-
-#define C(x) AV_STRINGIFY(16*(x-1))"(%1)"
-#define OC_8 "%2"
-
-    /* at this point, function has completed dequantization + dezigzag +
-     * partial transposition; now do the idct itself */
-#define I(x) AV_STRINGIFY(16* x       )"(%0)"
-#define J(x) AV_STRINGIFY(16*(x-4) + 8)"(%0)"
-
-    __asm__ volatile (
-        RowIDCT()
-        Transpose()
-
-#undef I
-#undef J
-#define I(x) AV_STRINGIFY(16* x    + 64)"(%0)"
-#define J(x) AV_STRINGIFY(16*(x-4) + 72)"(%0)"
-
-        RowIDCT()
-        Transpose()
-
-#undef I
-#undef J
-#define I(x) AV_STRINGIFY(16*x)"(%0)"
-#define J(x) AV_STRINGIFY(16*x)"(%0)"
-
-        ColumnIDCT()
-
-#undef I
-#undef J
-#define I(x) AV_STRINGIFY(16*x + 8)"(%0)"
-#define J(x) AV_STRINGIFY(16*x + 8)"(%0)"
-
-        ColumnIDCT()
-        :: "r"(output_data), "r"(ff_vp3_idct_data), "m"(ff_pw_8)
-    );
-#undef I
-#undef J
-
-}
-
-void ff_vp3_idct_put_mmx(uint8_t *dest, int line_size, DCTELEM *block)
-{
-    ff_vp3_idct_mmx(block);
-    ff_put_signed_pixels_clamped_mmx(block, dest, line_size);
-}
-
-void ff_vp3_idct_add_mmx(uint8_t *dest, int line_size, DCTELEM *block)
-{
-    ff_vp3_idct_mmx(block);
-    ff_add_pixels_clamped_mmx(block, dest, line_size);
-}
-
-void ff_vp3_idct_dc_add_mmx2(uint8_t *dest, int linesize, const DCTELEM *block)
-{
-    int dc = (block[0] + 15) >> 5;
-
-    __asm__ volatile(
-        "movd          %3, %%mm0 \n\t"
-        "pshufw $0, %%mm0, %%mm0 \n\t"
-        "pxor       %%mm1, %%mm1 \n\t"
-        "psubw      %%mm0, %%mm1 \n\t"
-        "packuswb   %%mm0, %%mm0 \n\t"
-        "packuswb   %%mm1, %%mm1 \n\t"
-
-#define DC_ADD \
-        "movq        (%0), %%mm2 \n\t" \
-        "movq     (%0,%1), %%mm3 \n\t" \
-        "paddusb    %%mm0, %%mm2 \n\t" \
-        "movq   (%0,%1,2), %%mm4 \n\t" \
-        "paddusb    %%mm0, %%mm3 \n\t" \
-        "movq     (%0,%2), %%mm5 \n\t" \
-        "paddusb    %%mm0, %%mm4 \n\t" \
-        "paddusb    %%mm0, %%mm5 \n\t" \
-        "psubusb    %%mm1, %%mm2 \n\t" \
-        "psubusb    %%mm1, %%mm3 \n\t" \
-        "movq       %%mm2, (%0)  \n\t" \
-        "psubusb    %%mm1, %%mm4 \n\t" \
-        "movq       %%mm3, (%0,%1) \n\t" \
-        "psubusb    %%mm1, %%mm5 \n\t" \
-        "movq       %%mm4, (%0,%1,2) \n\t" \
-        "movq       %%mm5, (%0,%2) \n\t"
-
-        DC_ADD
-        "lea    (%0,%1,4), %0 \n\t"
-        DC_ADD
-
-        : "+r"(dest)
-        : "r"((x86_reg)linesize), "r"((x86_reg)3*linesize), "r"(dc)
-    );
-}
diff --git a/libavcodec/x86/vp3dsp_mmx.h b/libavcodec/x86/vp3dsp_mmx.h
deleted file mode 100644
index e0ebf0b0f4..0000000000
--- a/libavcodec/x86/vp3dsp_mmx.h
+++ /dev/null
@@ -1,36 +0,0 @@
-/*
- * vp3dsp MMX function declarations
- * Copyright (c) 2007 Aurelien Jacobs <aurel@gnuage.org>
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * FFmpeg is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#ifndef AVCODEC_X86_VP3DSP_MMX_H
-#define AVCODEC_X86_VP3DSP_MMX_H
-
-#include <stdint.h>
-#include "libavcodec/dsputil.h"
-
-void ff_vp3_idct_mmx(int16_t *data);
-void ff_vp3_idct_put_mmx(uint8_t *dest, int line_size, DCTELEM *block);
-void ff_vp3_idct_add_mmx(uint8_t *dest, int line_size, DCTELEM *block);
-void ff_vp3_idct_dc_add_mmx2(uint8_t *dest, int line_size, const DCTELEM *block);
-
-void ff_vp3_v_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values);
-void ff_vp3_h_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values);
-
-#endif /* AVCODEC_X86_VP3DSP_MMX_H */
diff --git a/libavcodec/x86/vp3dsp_sse2.c b/libavcodec/x86/vp3dsp_sse2.c
deleted file mode 100644
index 144f763845..0000000000
--- a/libavcodec/x86/vp3dsp_sse2.c
+++ /dev/null
@@ -1,187 +0,0 @@
-/*
- * Copyright (C) 2004 the ffmpeg project
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * FFmpeg is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-/**
- * @file
- * SSE2-optimized functions cribbed from the original VP3 source code.
- */
-
-#include "libavcodec/dsputil.h"
-#include "dsputil_mmx.h"
-#include "vp3dsp_sse2.h"
-
-DECLARE_ALIGNED(16, const uint16_t, ff_vp3_idct_data)[7 * 8] =
-{
-    64277,64277,64277,64277,64277,64277,64277,64277,
-    60547,60547,60547,60547,60547,60547,60547,60547,
-    54491,54491,54491,54491,54491,54491,54491,54491,
-    46341,46341,46341,46341,46341,46341,46341,46341,
-    36410,36410,36410,36410,36410,36410,36410,36410,
-    25080,25080,25080,25080,25080,25080,25080,25080,
-    12785,12785,12785,12785,12785,12785,12785,12785
-};
-
-
-#define VP3_1D_IDCT_SSE2(ADD, SHIFT) \
-    "movdqa "I(3)", %%xmm2 \n\t"     /* xmm2 = i3 */ \
-    "movdqa "C(3)", %%xmm6 \n\t"     /* xmm6 = c3 */ \
-    "movdqa %%xmm2, %%xmm4 \n\t"     /* xmm4 = i3 */ \
-    "movdqa "I(5)", %%xmm7 \n\t"     /* xmm7 = i5 */ \
-    "pmulhw %%xmm6, %%xmm4 \n\t"     /* xmm4 = c3 * i3 - i3 */ \
-    "movdqa "C(5)", %%xmm1 \n\t"     /* xmm1 = c5 */ \
-    "pmulhw %%xmm7, %%xmm6 \n\t"     /* xmm6 = c3 * i5 - i5 */ \
-    "movdqa %%xmm1, %%xmm5 \n\t"     /* xmm5 = c5 */ \
-    "pmulhw %%xmm2, %%xmm1 \n\t"     /* xmm1 = c5 * i3 - i3 */ \
-    "movdqa "I(1)", %%xmm3 \n\t"     /* xmm3 = i1 */ \
-    "pmulhw %%xmm7, %%xmm5 \n\t"     /* xmm5 = c5 * i5 - i5 */ \
-    "movdqa "C(1)", %%xmm0 \n\t"     /* xmm0 = c1 */ \
-    "paddw  %%xmm2, %%xmm4 \n\t"     /* xmm4 = c3 * i3 */ \
-    "paddw  %%xmm7, %%xmm6 \n\t"     /* xmm6 = c3 * i5 */ \
-    "paddw  %%xmm1, %%xmm2 \n\t"     /* xmm2 = c5 * i3 */ \
-    "movdqa "I(7)", %%xmm1 \n\t"     /* xmm1 = i7 */ \
-    "paddw  %%xmm5, %%xmm7 \n\t"     /* xmm7 = c5 * i5 */ \
-    "movdqa %%xmm0, %%xmm5 \n\t"     /* xmm5 = c1 */ \
-    "pmulhw %%xmm3, %%xmm0 \n\t"     /* xmm0 = c1 * i1 - i1 */ \
-    "paddsw %%xmm7, %%xmm4 \n\t"     /* xmm4 = c3 * i3 + c5 * i5 = C */ \
-    "pmulhw %%xmm1, %%xmm5 \n\t"     /* xmm5 = c1 * i7 - i7 */ \
-    "movdqa "C(7)", %%xmm7 \n\t"     /* xmm7 = c7 */ \
-    "psubsw %%xmm2, %%xmm6 \n\t"     /* xmm6 = c3 * i5 - c5 * i3 = D */ \
-    "paddw  %%xmm3, %%xmm0 \n\t"     /* xmm0 = c1 * i1 */ \
-    "pmulhw %%xmm7, %%xmm3 \n\t"     /* xmm3 = c7 * i1 */ \
-    "movdqa "I(2)", %%xmm2 \n\t"     /* xmm2 = i2 */ \
-    "pmulhw %%xmm1, %%xmm7 \n\t"     /* xmm7 = c7 * i7 */ \
-    "paddw  %%xmm1, %%xmm5 \n\t"     /* xmm5 = c1 * i7 */ \
-    "movdqa %%xmm2, %%xmm1 \n\t"     /* xmm1 = i2 */ \
-    "pmulhw "C(2)", %%xmm2 \n\t"     /* xmm2 = i2 * c2 -i2 */ \
-    "psubsw %%xmm5, %%xmm3 \n\t"     /* xmm3 = c7 * i1 - c1 * i7 = B */ \
-    "movdqa "I(6)", %%xmm5 \n\t"     /* xmm5 = i6 */ \
-    "paddsw %%xmm7, %%xmm0 \n\t"     /* xmm0 = c1 * i1 + c7 * i7 = A */ \
-    "movdqa %%xmm5, %%xmm7 \n\t"     /* xmm7 = i6 */ \
-    "psubsw %%xmm4, %%xmm0 \n\t"     /* xmm0 = A - C */ \
-    "pmulhw "C(2)", %%xmm5 \n\t"     /* xmm5 = c2 * i6 - i6 */ \
-    "paddw  %%xmm1, %%xmm2 \n\t"     /* xmm2 = i2 * c2 */ \
-    "pmulhw "C(6)", %%xmm1 \n\t"     /* xmm1 = c6 * i2 */ \
-    "paddsw %%xmm4, %%xmm4 \n\t"     /* xmm4 = C + C */ \
-    "paddsw %%xmm0, %%xmm4 \n\t"     /* xmm4 = A + C = C. */ \
-    "psubsw %%xmm6, %%xmm3 \n\t"     /* xmm3 = B - D */ \
-    "paddw  %%xmm7, %%xmm5 \n\t"     /* xmm5 = c2 * i6 */ \
-    "paddsw %%xmm6, %%xmm6 \n\t"     /* xmm6 = D + D */ \
-    "pmulhw "C(6)", %%xmm7 \n\t"     /* xmm7 = c6 * i6 */ \
-    "paddsw %%xmm3, %%xmm6 \n\t"     /* xmm6 = B + D = D. */ \
-    "movdqa %%xmm4, "I(1)" \n\t"     /* Save C. at I(1) */ \
-    "psubsw %%xmm5, %%xmm1 \n\t"     /* xmm1 = c6 * i2 - c2 * i6 = H */ \
-    "movdqa "C(4)", %%xmm4 \n\t"     /* xmm4 = c4 */ \
-    "movdqa %%xmm3, %%xmm5 \n\t"     /* xmm5 = B - D */ \
-    "pmulhw %%xmm4, %%xmm3 \n\t"     /* xmm3 = ( c4 -1 ) * ( B - D ) */ \
-    "paddsw %%xmm2, %%xmm7 \n\t"     /* xmm7 = c2 * i2 + c6 * i6 = G */ \
-    "movdqa %%xmm6, "I(2)" \n\t"     /* Save D. at I(2) */ \
-    "movdqa %%xmm0, %%xmm2 \n\t"     /* xmm2 = A - C */ \
-    "movdqa "I(0)", %%xmm6 \n\t"     /* xmm6 = i0 */ \
-    "pmulhw %%xmm4, %%xmm0 \n\t"     /* xmm0 = ( c4 - 1 ) * ( A - C ) = A. */ \
-    "paddw  %%xmm3, %%xmm5 \n\t"     /* xmm5 = c4 * ( B - D ) = B. */ \
-    "movdqa "I(4)", %%xmm3 \n\t"     /* xmm3 = i4 */ \
-    "psubsw %%xmm1, %%xmm5 \n\t"     /* xmm5 = B. - H = B.. */ \
-    "paddw  %%xmm0, %%xmm2 \n\t"     /* xmm2 = c4 * ( A - C) = A. */ \
-    "psubsw %%xmm3, %%xmm6 \n\t"     /* xmm6 = i0 - i4 */ \
-    "movdqa %%xmm6, %%xmm0 \n\t"     /* xmm0 = i0 - i4 */ \
-    "pmulhw %%xmm4, %%xmm6 \n\t"     /* xmm6 = (c4 - 1) * (i0 - i4) = F */ \
-    "paddsw %%xmm3, %%xmm3 \n\t"     /* xmm3 = i4 + i4 */ \
-    "paddsw %%xmm1, %%xmm1 \n\t"     /* xmm1 = H + H */ \
-    "paddsw %%xmm0, %%xmm3 \n\t"     /* xmm3 = i0 + i4 */ \
-    "paddsw %%xmm5, %%xmm1 \n\t"     /* xmm1 = B. + H = H. */ \
-    "pmulhw %%xmm3, %%xmm4 \n\t"     /* xmm4 = ( c4 - 1 ) * ( i0 + i4 )  */ \
-    "paddw  %%xmm0, %%xmm6 \n\t"     /* xmm6 = c4 * ( i0 - i4 ) */ \
-    "psubsw %%xmm2, %%xmm6 \n\t"     /* xmm6 = F - A. = F. */ \
-    "paddsw %%xmm2, %%xmm2 \n\t"     /* xmm2 = A. + A. */ \
-    "movdqa "I(1)", %%xmm0 \n\t"     /* Load        C. from I(1) */ \
-    "paddsw %%xmm6, %%xmm2 \n\t"     /* xmm2 = F + A. = A.. */ \
-    "paddw  %%xmm3, %%xmm4 \n\t"     /* xmm4 = c4 * ( i0 + i4 ) = 3 */ \
-    "psubsw %%xmm1, %%xmm2 \n\t"     /* xmm2 = A.. - H. = R2 */ \
-    ADD(%%xmm2)                      /* Adjust R2 and R1 before shifting */ \
-    "paddsw %%xmm1, %%xmm1 \n\t"     /* xmm1 = H. + H. */ \
-    "paddsw %%xmm2, %%xmm1 \n\t"     /* xmm1 = A.. + H. = R1 */ \
-    SHIFT(%%xmm2)                    /* xmm2 = op2 */ \
-    "psubsw %%xmm7, %%xmm4 \n\t"     /* xmm4 = E - G = E. */ \
-    SHIFT(%%xmm1)                    /* xmm1 = op1 */ \
-    "movdqa "I(2)", %%xmm3 \n\t"     /* Load D. from I(2) */ \
-    "paddsw %%xmm7, %%xmm7 \n\t"     /* xmm7 = G + G */ \
-    "paddsw %%xmm4, %%xmm7 \n\t"     /* xmm7 = E + G = G. */ \
-    "psubsw %%xmm3, %%xmm4 \n\t"     /* xmm4 = E. - D. = R4 */ \
-    ADD(%%xmm4)                      /* Adjust R4 and R3 before shifting */ \
-    "paddsw %%xmm3, %%xmm3 \n\t"     /* xmm3 = D. + D. */ \
-    "paddsw %%xmm4, %%xmm3 \n\t"     /* xmm3 = E. + D. = R3 */ \
-    SHIFT(%%xmm4)                    /* xmm4 = op4 */ \
-    "psubsw %%xmm5, %%xmm6 \n\t"     /* xmm6 = F. - B..= R6 */ \
-    SHIFT(%%xmm3)                    /* xmm3 = op3 */ \
-    ADD(%%xmm6)                      /* Adjust R6 and R5 before shifting */ \
-    "paddsw %%xmm5, %%xmm5 \n\t"     /* xmm5 = B.. + B.. */ \
-    "paddsw %%xmm6, %%xmm5 \n\t"     /* xmm5 = F. + B.. = R5 */ \
-    SHIFT(%%xmm6)                    /* xmm6 = op6 */ \
-    SHIFT(%%xmm5)                    /* xmm5 = op5 */ \
-    "psubsw %%xmm0, %%xmm7 \n\t"     /* xmm7 = G. - C. = R7 */ \
-    ADD(%%xmm7)                      /* Adjust R7 and R0 before shifting */ \
-    "paddsw %%xmm0, %%xmm0 \n\t"     /* xmm0 = C. + C. */ \
-    "paddsw %%xmm7, %%xmm0 \n\t"     /* xmm0 = G. + C. */ \
-    SHIFT(%%xmm7)                    /* xmm7 = op7 */ \
-    SHIFT(%%xmm0)                    /* xmm0 = op0 */
-
-#define PUT_BLOCK(r0, r1, r2, r3, r4, r5, r6, r7) \
-    "movdqa " #r0 ", " O(0) "\n\t" \
-    "movdqa " #r1 ", " O(1) "\n\t" \
-    "movdqa " #r2 ", " O(2) "\n\t" \
-    "movdqa " #r3 ", " O(3) "\n\t" \
-    "movdqa " #r4 ", " O(4) "\n\t" \
-    "movdqa " #r5 ", " O(5) "\n\t" \
-    "movdqa " #r6 ", " O(6) "\n\t" \
-    "movdqa " #r7 ", " O(7) "\n\t"
-
-#define NOP(xmm)
-#define SHIFT4(xmm) "psraw  $4, "#xmm"\n\t"
-#define ADD8(xmm)   "paddsw %2, "#xmm"\n\t"
-
-void ff_vp3_idct_sse2(int16_t *input_data)
-{
-#define I(x) AV_STRINGIFY(16*x)"(%0)"
-#define O(x) I(x)
-#define C(x) AV_STRINGIFY(16*(x-1))"(%1)"
-
-    __asm__ volatile (
-        VP3_1D_IDCT_SSE2(NOP, NOP)
-
-        TRANSPOSE8(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm6, %%xmm7, (%0))
-        PUT_BLOCK(%%xmm0, %%xmm5, %%xmm7, %%xmm3, %%xmm6, %%xmm4, %%xmm2, %%xmm1)
-
-        VP3_1D_IDCT_SSE2(ADD8, SHIFT4)
-        PUT_BLOCK(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm6, %%xmm7)
-        :: "r"(input_data), "r"(ff_vp3_idct_data), "m"(ff_pw_8)
-    );
-}
-
-void ff_vp3_idct_put_sse2(uint8_t *dest, int line_size, DCTELEM *block)
-{
-    ff_vp3_idct_sse2(block);
-    ff_put_signed_pixels_clamped_mmx(block, dest, line_size);
-}
-
-void ff_vp3_idct_add_sse2(uint8_t *dest, int line_size, DCTELEM *block)
-{
-    ff_vp3_idct_sse2(block);
-    ff_add_pixels_clamped_mmx(block, dest, line_size);
-}
diff --git a/libavcodec/x86/vp3dsp_sse2.h b/libavcodec/x86/vp3dsp_sse2.h
deleted file mode 100644
index 9094620eb2..0000000000
--- a/libavcodec/x86/vp3dsp_sse2.h
+++ /dev/null
@@ -1,31 +0,0 @@
-/*
- * vp3dsp SSE2 function declarations
- * Copyright (c) 2007 Aurelien Jacobs <aurel@gnuage.org>
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * FFmpeg is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#ifndef AVCODEC_X86_VP3DSP_SSE2_H
-#define AVCODEC_X86_VP3DSP_SSE2_H
-
-#include "libavcodec/dsputil.h"
-
-void ff_vp3_idct_sse2(int16_t *input_data);
-void ff_vp3_idct_put_sse2(uint8_t *dest, int line_size, DCTELEM *block);
-void ff_vp3_idct_add_sse2(uint8_t *dest, int line_size, DCTELEM *block);
-
-#endif /* AVCODEC_X86_VP3DSP_SSE2_H */