Merge remote-tracking branch 'qatar/master'

* qatar/master:
  lavr: fix handling of custom mix matrices
  fate: force pix_fmt in lagarith-rgb32 test
  fate: add tests for lagarith lossless video codec.
  ARMv6: vp8: fix stack allocation with Apple's assembler
  ARM: vp56: allow inline asm to build with clang
  fft: 3dnow: fix register name typo in DECL_IMDCT macro
  x86: dct32: port to cpuflags
  x86: build: replace mmx2 by mmxext
  Revert "wmapro: prevent division by zero when sample rate is unspecified"
  wmapro: prevent division by zero when sample rate is unspecified
  lagarith: fix color plane inversion for YUY2 output.
  lagarith: pad RGB buffer by 1 byte.
  dsputil: make add_hfyu_left_prediction_sse4() support unaligned src.

Conflicts:
	doc/APIchanges
	libavcodec/lagarith.c
	libavfilter/x86/gradfun.c
	libavutil/cpu.h
	libavutil/version.h
	libswscale/utils.c
	libswscale/version.h
	libswscale/x86/yuv2rgb.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

20 files changed, 67 insertions, 86 deletions

diff --git a/libavcodec/x86/ac3dsp.asm b/libavcodec/x86/ac3dsp.asm
index 4e0fce35d1..acbab35cae 100644
--- a/libavcodec/x86/ac3dsp.asm
+++ b/libavcodec/x86/ac3dsp.asm
@@ -68,7 +68,7 @@ cglobal ac3_exponent_min_%1, 3,4,2, exp, reuse_blks, expn, offset
 %define LOOP_ALIGN
 INIT_MMX
 AC3_EXPONENT_MIN mmx
-%if HAVE_MMX2
+%if HAVE_MMXEXT
 %define PMINUB PMINUB_MMXEXT
 %define LOOP_ALIGN ALIGN 16
 AC3_EXPONENT_MIN mmxext
diff --git a/libavcodec/x86/ac3dsp_mmx.c b/libavcodec/x86/ac3dsp_mmx.c
index 54fa380130..5549f3e550 100644
--- a/libavcodec/x86/ac3dsp_mmx.c
+++ b/libavcodec/x86/ac3dsp_mmx.c
@@ -65,7 +65,7 @@ av_cold void ff_ac3dsp_init_x86(AC3DSPContext *c, int bit_exact)
             c->float_to_fixed24 = ff_float_to_fixed24_3dnow;
         }
     }
-    if (mm_flags & AV_CPU_FLAG_MMX2 && HAVE_MMX2) {
+    if (mm_flags & AV_CPU_FLAG_MMXEXT && HAVE_MMXEXT) {
         c->ac3_exponent_min = ff_ac3_exponent_min_mmxext;
         c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmx2;
     }
diff --git a/libavcodec/x86/cavsdsp_mmx.c b/libavcodec/x86/cavsdsp_mmx.c
index eae6fea536..05f192fc71 100644
--- a/libavcodec/x86/cavsdsp_mmx.c
+++ b/libavcodec/x86/cavsdsp_mmx.c
@@ -486,7 +486,7 @@ void ff_cavsdsp_init_mmx(CAVSDSPContext *c, AVCodecContext *avctx)
     int mm_flags = av_get_cpu_flags();
 
 #if HAVE_INLINE_ASM
-    if (mm_flags & AV_CPU_FLAG_MMX2)  ff_cavsdsp_init_mmx2 (c, avctx);
+    if (mm_flags & AV_CPU_FLAG_MMXEXT) ff_cavsdsp_init_mmx2(c, avctx);
     if (mm_flags & AV_CPU_FLAG_3DNOW) ff_cavsdsp_init_3dnow(c, avctx);
 #endif /* HAVE_INLINE_ASM */
 }
diff --git a/libavcodec/x86/dct32_sse.asm b/libavcodec/x86/dct32_sse.asm
index ca44106433..02b5f3fc89 100644
--- a/libavcodec/x86/dct32_sse.asm
+++ b/libavcodec/x86/dct32_sse.asm
@@ -42,39 +42,24 @@ ps_cos_vec: dd   0.500603,  0.505471,  0.515447,  0.531043
 align 32
 ps_p1p1m1m1: dd 0, 0, 0x80000000, 0x80000000, 0, 0, 0x80000000, 0x80000000
 
-%macro BUTTERFLY_SSE 4
-    movaps %4, %1
-    subps  %1, %2
-    addps  %2, %4
-    mulps  %1, %3
-%endmacro
-
-%macro BUTTERFLY_AVX 4
-    vsubps  %4, %1, %2
-    vaddps  %2, %2, %1
-    vmulps  %1, %4, %3
-%endmacro
-
-%macro BUTTERFLY0_SSE 5
-    movaps %4, %1
-    shufps %1, %1, %5
-    xorps  %4, %2
-    addps  %1, %4
-    mulps  %1, %3
+%macro BUTTERFLY 4
+    subps  %4, %1, %2
+    addps  %2, %2, %1
+    mulps  %1, %4, %3
 %endmacro
 
-%macro BUTTERFLY0_SSE2 5
+%macro BUTTERFLY0 5
+%if cpuflag(sse2) && notcpuflag(avx)
     pshufd %4, %1, %5
     xorps  %1, %2
     addps  %1, %4
     mulps  %1, %3
-%endmacro
-
-%macro BUTTERFLY0_AVX 5
-    vshufps %4, %1, %1, %5
-    vxorps  %1, %1, %2
-    vaddps  %4, %4, %1
-    vmulps  %1, %4, %3
+%else
+    shufps %4, %1, %1, %5
+    xorps  %1, %1, %2
+    addps  %4, %4, %1
+    mulps  %1, %4, %3
+%endif
 %endmacro
 
 %macro BUTTERFLY2 4
@@ -206,14 +191,11 @@ ps_p1p1m1m1: dd 0, 0, 0x80000000, 0x80000000, 0, 0, 0x80000000, 0x80000000
     movss [outq+116], m6
 %endmacro
 
-%define BUTTERFLY  BUTTERFLY_AVX
-%define BUTTERFLY0 BUTTERFLY0_AVX
-
-INIT_YMM
+INIT_YMM avx
 SECTION_TEXT
 %if HAVE_AVX
 ; void ff_dct32_float_avx(FFTSample *out, const FFTSample *in)
-cglobal dct32_float_avx, 2,3,8, out, in, tmp
+cglobal dct32_float, 2,3,8, out, in, tmp
     ; pass 1
     vmovaps     m4, [inq+0]
     vinsertf128 m5, m5, [inq+96], 1
@@ -286,9 +268,6 @@ INIT_XMM
     RET
 %endif
 
-%define BUTTERFLY  BUTTERFLY_SSE
-%define BUTTERFLY0 BUTTERFLY0_SSE
-
 %if ARCH_X86_64
 %define SPILL SWAP
 %define UNSPILL SWAP
@@ -411,10 +390,9 @@ INIT_XMM
 %endif
 
 
-INIT_XMM
-%macro DCT32_FUNC 1
 ; void ff_dct32_float_sse(FFTSample *out, const FFTSample *in)
-cglobal dct32_float_%1, 2,3,16, out, in, tmp
+%macro DCT32_FUNC 0
+cglobal dct32_float, 2, 3, 16, out, in, tmp
     ; pass 1
 
     movaps      m0, [inq+0]
@@ -498,18 +476,16 @@ cglobal dct32_float_%1, 2,3,16, out, in, tmp
     RET
 %endmacro
 
-%macro LOAD_INV_SSE 2
+%macro LOAD_INV 2
+%if cpuflag(sse2)
+    pshufd      %1, %2, 0x1b
+%elif cpuflag(sse)
     movaps      %1, %2
     shufps      %1, %1, 0x1b
+%endif
 %endmacro
 
-%define LOAD_INV LOAD_INV_SSE
-DCT32_FUNC sse
-
-%macro LOAD_INV_SSE2 2
-    pshufd      %1, %2, 0x1b
-%endmacro
-
-%define LOAD_INV LOAD_INV_SSE2
-%define BUTTERFLY0 BUTTERFLY0_SSE2
-DCT32_FUNC sse2
+INIT_XMM sse
+DCT32_FUNC
+INIT_XMM sse2
+DCT32_FUNC
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index 47b8ef1fc3..f4ed7565e7 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -3171,7 +3171,7 @@ void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx)
                     c->idct_add              = ff_idct_xvid_sse2_add;
                     c->idct                  = ff_idct_xvid_sse2;
                     c->idct_permutation_type = FF_SSE2_IDCT_PERM;
-                } else if (mm_flags & AV_CPU_FLAG_MMX2) {
+                } else if (mm_flags & AV_CPU_FLAG_MMXEXT) {
                     c->idct_put              = ff_idct_xvid_mmx2_put;
                     c->idct_add              = ff_idct_xvid_mmx2_add;
                     c->idct                  = ff_idct_xvid_mmx2;
@@ -3187,7 +3187,7 @@ void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx)
         dsputil_init_mmx(c, avctx, mm_flags);
     }
 
-    if (mm_flags & AV_CPU_FLAG_MMX2)
+    if (mm_flags & AV_CPU_FLAG_MMXEXT)
         dsputil_init_mmx2(c, avctx, mm_flags);
 
     if (mm_flags & AV_CPU_FLAG_3DNOW && HAVE_AMD3DNOW)
diff --git a/libavcodec/x86/dsputil_yasm.asm b/libavcodec/x86/dsputil_yasm.asm
index 5a6c3d1eae..06d2027c69 100644
--- a/libavcodec/x86/dsputil_yasm.asm
+++ b/libavcodec/x86/dsputil_yasm.asm
@@ -388,12 +388,16 @@ cglobal add_hfyu_median_prediction_mmx2, 6,6,0, dst, top, diff, w, left, left_to
     RET
 
 
-%macro ADD_HFYU_LEFT_LOOP 1 ; %1 = is_aligned
+%macro ADD_HFYU_LEFT_LOOP 2 ; %1 = dst_is_aligned, %2 = src_is_aligned
     add     srcq, wq
     add     dstq, wq
     neg     wq
 %%.loop:
+%if %2
     mova    m1, [srcq+wq]
+%else
+    movu    m1, [srcq+wq]
+%endif
     mova    m2, m1
     psllw   m1, 8
     paddb   m1, m2
@@ -435,7 +439,7 @@ cglobal add_hfyu_left_prediction_ssse3, 3,3,7, dst, src, w, left
     mova    m3, [pb_zz11zz55zz99zzdd]
     movd    m0, leftm
     psllq   m0, 56
-    ADD_HFYU_LEFT_LOOP 1
+    ADD_HFYU_LEFT_LOOP 1, 1
 
 INIT_XMM
 cglobal add_hfyu_left_prediction_sse4, 3,3,7, dst, src, w, left
@@ -446,12 +450,14 @@ cglobal add_hfyu_left_prediction_sse4, 3,3,7, dst, src, w, left
     movd    m0, leftm
     pslldq  m0, 15
     test    srcq, 15
-    jnz add_hfyu_left_prediction_ssse3.skip_prologue
+    jnz .src_unaligned
     test    dstq, 15
-    jnz .unaligned
-    ADD_HFYU_LEFT_LOOP 1
-.unaligned:
-    ADD_HFYU_LEFT_LOOP 0
+    jnz .dst_unaligned
+    ADD_HFYU_LEFT_LOOP 1, 1
+.dst_unaligned:
+    ADD_HFYU_LEFT_LOOP 0, 1
+.src_unaligned:
+    ADD_HFYU_LEFT_LOOP 0, 0
 
 
 ; float scalarproduct_float_sse(const float *v1, const float *v2, int len)
diff --git a/libavcodec/x86/dsputilenc_mmx.c b/libavcodec/x86/dsputilenc_mmx.c
index cdc384ca9d..77b7b7620e 100644
--- a/libavcodec/x86/dsputilenc_mmx.c
+++ b/libavcodec/x86/dsputilenc_mmx.c
@@ -1112,7 +1112,7 @@ void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
             (dct_algo==FF_DCT_AUTO || dct_algo==FF_DCT_MMX)) {
             if(mm_flags & AV_CPU_FLAG_SSE2){
                 c->fdct = ff_fdct_sse2;
-            }else if(mm_flags & AV_CPU_FLAG_MMX2){
+            } else if (mm_flags & AV_CPU_FLAG_MMXEXT) {
                 c->fdct = ff_fdct_mmx2;
             }else{
                 c->fdct = ff_fdct_mmx;
@@ -1145,8 +1145,7 @@ void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
 
         c->ssd_int8_vs_int16 = ssd_int8_vs_int16_mmx;
 
-
-        if (mm_flags & AV_CPU_FLAG_MMX2) {
+        if (mm_flags & AV_CPU_FLAG_MMXEXT) {
             c->sum_abs_dctelem= sum_abs_dctelem_mmx2;
             c->vsad[4]= vsad_intra16_mmx2;
 
@@ -1187,7 +1186,7 @@ void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
         c->hadamard8_diff[0] = ff_hadamard8_diff16_mmx;
         c->hadamard8_diff[1] = ff_hadamard8_diff_mmx;
 
-        if (mm_flags & AV_CPU_FLAG_MMX2) {
+        if (mm_flags & AV_CPU_FLAG_MMXEXT) {
             c->hadamard8_diff[0] = ff_hadamard8_diff16_mmx2;
             c->hadamard8_diff[1] = ff_hadamard8_diff_mmx2;
         }
diff --git a/libavcodec/x86/fft_mmx.asm b/libavcodec/x86/fft_mmx.asm
index f41381760d..faa27a01c6 100644
--- a/libavcodec/x86/fft_mmx.asm
+++ b/libavcodec/x86/fft_mmx.asm
@@ -1041,7 +1041,7 @@ cglobal imdct_half, 3,12,8; FFTContext *s, FFTSample *output, const FFTSample *i
     mova [r1+r5*8], m0
     mova [r1+r6*8], m2
     add    r4, 2
-    sub    r4, 2
+    sub    r3, 2
 %else
 %if ARCH_X86_64
     movzx  r5,  word [rrevtab+r4-4]
diff --git a/libavcodec/x86/h264_intrapred_init.c b/libavcodec/x86/h264_intrapred_init.c
index 5975d07706..59ab3ea27e 100644
--- a/libavcodec/x86/h264_intrapred_init.c
+++ b/libavcodec/x86/h264_intrapred_init.c
@@ -198,7 +198,7 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
             }
         }
 
-        if (mm_flags & AV_CPU_FLAG_MMX2) {
+        if (mm_flags & AV_CPU_FLAG_MMXEXT) {
             h->pred16x16[HOR_PRED8x8            ] = ff_pred16x16_horizontal_mmx2;
             h->pred16x16[DC_PRED8x8             ] = ff_pred16x16_dc_mmx2;
             if (chroma_format_idc == 1)
@@ -308,7 +308,7 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
             }
         }
     } else if (bit_depth == 10) {
-        if (mm_flags & AV_CPU_FLAG_MMX2) {
+        if (mm_flags & AV_CPU_FLAG_MMXEXT) {
             h->pred4x4[DC_PRED             ] = ff_pred4x4_dc_10_mmxext;
             h->pred4x4[HOR_UP_PRED         ] = ff_pred4x4_horizontal_up_10_mmxext;
 
diff --git a/libavcodec/x86/h264dsp_mmx.c b/libavcodec/x86/h264dsp_mmx.c
index bb77a96999..87b9452501 100644
--- a/libavcodec/x86/h264dsp_mmx.c
+++ b/libavcodec/x86/h264dsp_mmx.c
@@ -218,7 +218,7 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
 #if HAVE_YASM
     int mm_flags = av_get_cpu_flags();
 
-    if (chroma_format_idc == 1 && mm_flags & AV_CPU_FLAG_MMX2)
+    if (chroma_format_idc == 1 && mm_flags & AV_CPU_FLAG_MMXEXT)
         c->h264_loop_filter_strength = ff_h264_loop_filter_strength_mmx2;
 
     if (bit_depth == 8) {
@@ -236,7 +236,7 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
             if (mm_flags & AV_CPU_FLAG_CMOV)
                 c->h264_luma_dc_dequant_idct = ff_h264_luma_dc_dequant_idct_mmx;
 
-            if (mm_flags & AV_CPU_FLAG_MMX2) {
+            if (mm_flags & AV_CPU_FLAG_MMXEXT) {
                 c->h264_idct_dc_add  = ff_h264_idct_dc_add_8_mmx2;
                 c->h264_idct8_dc_add = ff_h264_idct8_dc_add_8_mmx2;
                 c->h264_idct_add16   = ff_h264_idct_add16_8_mmx2;
@@ -304,7 +304,7 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
         }
     } else if (bit_depth == 10) {
         if (mm_flags & AV_CPU_FLAG_MMX) {
-            if (mm_flags & AV_CPU_FLAG_MMX2) {
+            if (mm_flags & AV_CPU_FLAG_MMXEXT) {
 #if ARCH_X86_32
                 c->h264_v_loop_filter_chroma       = ff_deblock_v_chroma_10_mmx2;
                 c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_mmx2;
diff --git a/libavcodec/x86/motion_est_mmx.c b/libavcodec/x86/motion_est_mmx.c
index a903c96950..1d3545a5e8 100644
--- a/libavcodec/x86/motion_est_mmx.c
+++ b/libavcodec/x86/motion_est_mmx.c
@@ -444,7 +444,7 @@ void ff_dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx)
         c->sad[0]= sad16_mmx;
         c->sad[1]= sad8_mmx;
     }
-    if (mm_flags & AV_CPU_FLAG_MMX2) {
+    if (mm_flags & AV_CPU_FLAG_MMXEXT) {
         c->pix_abs[0][0] = sad16_mmx2;
         c->pix_abs[1][0] = sad8_mmx2;
 
diff --git a/libavcodec/x86/mpegvideo_mmx.c b/libavcodec/x86/mpegvideo_mmx.c
index 8e72852b0a..44d4cd3a8a 100644
--- a/libavcodec/x86/mpegvideo_mmx.c
+++ b/libavcodec/x86/mpegvideo_mmx.c
@@ -595,15 +595,15 @@ static void  denoise_dct_sse2(MpegEncContext *s, DCTELEM *block){
 #define HAVE_SSSE3 0
 
 #undef HAVE_SSE2
-#undef HAVE_MMX2
+#undef HAVE_MMXEXT
 #define HAVE_SSE2 0
-#define HAVE_MMX2 0
+#define HAVE_MMXEXT 0
 #define RENAME(a) a ## _MMX
 #define RENAMEl(a) a ## _mmx
 #include "mpegvideo_mmx_template.c"
 
-#undef HAVE_MMX2
-#define HAVE_MMX2 1
+#undef HAVE_MMXEXT
+#define HAVE_MMXEXT 1
 #undef RENAME
 #undef RENAMEl
 #define RENAME(a) a ## _MMX2
@@ -660,7 +660,7 @@ void ff_MPV_common_init_mmx(MpegEncContext *s)
 #endif
             if(mm_flags & AV_CPU_FLAG_SSE2){
                 s->dct_quantize= dct_quantize_SSE2;
-            } else if(mm_flags & AV_CPU_FLAG_MMX2){
+            } else if (mm_flags & AV_CPU_FLAG_MMXEXT) {
                 s->dct_quantize= dct_quantize_MMX2;
             } else {
                 s->dct_quantize= dct_quantize_MMX;
diff --git a/libavcodec/x86/mpegvideo_mmx_template.c b/libavcodec/x86/mpegvideo_mmx_template.c
index 9119476d36..82e4ffa558 100644
--- a/libavcodec/x86/mpegvideo_mmx_template.c
+++ b/libavcodec/x86/mpegvideo_mmx_template.c
@@ -48,7 +48,7 @@
 #define MMREG_WIDTH "8"
 #define MM "%%mm"
 #define MOVQ "movq"
-#if HAVE_MMX2
+#if HAVE_MMXEXT
 #define SPREADW(a) "pshufw $0, "a", "a" \n\t"
 #define PMAXW(a,b) "pmaxsw "a", "b"     \n\t"
 #define PMAX(a,b) \
diff --git a/libavcodec/x86/pngdsp-init.c b/libavcodec/x86/pngdsp-init.c
index f122b242fb..7a12730620 100644
--- a/libavcodec/x86/pngdsp-init.c
+++ b/libavcodec/x86/pngdsp-init.c
@@ -41,7 +41,7 @@ void ff_pngdsp_init_x86(PNGDSPContext *dsp)
     if (flags & AV_CPU_FLAG_MMX)
         dsp->add_bytes_l2         = ff_add_bytes_l2_mmx;
 #endif
-    if (flags & AV_CPU_FLAG_MMX2)
+    if (flags & AV_CPU_FLAG_MMXEXT)
         dsp->add_paeth_prediction = ff_add_png_paeth_prediction_mmx2;
     if (flags & AV_CPU_FLAG_SSE2)
         dsp->add_bytes_l2         = ff_add_bytes_l2_sse2;
diff --git a/libavcodec/x86/rv34dsp_init.c b/libavcodec/x86/rv34dsp_init.c
index d91818c375..7284a9beaf 100644
--- a/libavcodec/x86/rv34dsp_init.c
+++ b/libavcodec/x86/rv34dsp_init.c
@@ -37,7 +37,7 @@ av_cold void ff_rv34dsp_init_x86(RV34DSPContext* c, DSPContext *dsp)
 
     if (mm_flags & AV_CPU_FLAG_MMX)
         c->rv34_idct_dc_add = ff_rv34_idct_dc_add_mmx;
-    if (mm_flags & AV_CPU_FLAG_MMX2) {
+    if (mm_flags & AV_CPU_FLAG_MMXEXT) {
         c->rv34_inv_transform_dc = ff_rv34_idct_dc_noround_mmx2;
         c->rv34_idct_add         = ff_rv34_idct_add_mmx2;
     }
diff --git a/libavcodec/x86/rv40dsp_init.c b/libavcodec/x86/rv40dsp_init.c
index bbb1c8eb57..c508ac9328 100644
--- a/libavcodec/x86/rv40dsp_init.c
+++ b/libavcodec/x86/rv40dsp_init.c
@@ -204,7 +204,7 @@ void ff_rv40dsp_init_x86(RV34DSPContext *c, DSPContext *dsp)
         QPEL_MC_SET(put_, _mmx)
 #endif
     }
-    if (mm_flags & AV_CPU_FLAG_MMX2) {
+    if (mm_flags & AV_CPU_FLAG_MMXEXT) {
         c->avg_chroma_pixels_tab[0] = ff_avg_rv40_chroma_mc8_mmx2;
         c->avg_chroma_pixels_tab[1] = ff_avg_rv40_chroma_mc4_mmx2;
         c->rv40_weight_pixels_tab[0][0] = ff_rv40_weight_func_rnd_16_mmx2;
diff --git a/libavcodec/x86/snowdsp_mmx.c b/libavcodec/x86/snowdsp_mmx.c
index eac43364e5..5d47206e81 100644
--- a/libavcodec/x86/snowdsp_mmx.c
+++ b/libavcodec/x86/snowdsp_mmx.c
@@ -889,7 +889,7 @@ void ff_dwt_init_x86(DWTContext *c)
             c->inner_add_yblock = ff_snow_inner_add_yblock_sse2;
         }
         else{
-            if(mm_flags & AV_CPU_FLAG_MMX2){
+            if (mm_flags & AV_CPU_FLAG_MMXEXT) {
             c->horizontal_compose97i = ff_snow_horizontal_compose97i_mmx;
 #if HAVE_7REGS
             c->vertical_compose97i = ff_snow_vertical_compose97i_mmx;
diff --git a/libavcodec/x86/vc1dsp_mmx.c b/libavcodec/x86/vc1dsp_mmx.c
index bddac5ec77..5922a6dbf7 100644
--- a/libavcodec/x86/vc1dsp_mmx.c
+++ b/libavcodec/x86/vc1dsp_mmx.c
@@ -760,7 +760,7 @@ void ff_vc1dsp_init_mmx(VC1DSPContext *dsp)
             dsp->put_no_rnd_vc1_chroma_pixels_tab[0]= ff_put_vc1_chroma_mc8_mmx_nornd;
     }
 
-    if (mm_flags & AV_CPU_FLAG_MMX2){
+    if (mm_flags & AV_CPU_FLAG_MMXEXT) {
         dsp->avg_vc1_mspel_pixels_tab[ 0] = ff_avg_vc1_mspel_mc00_mmx2;
         dsp->avg_vc1_mspel_pixels_tab[ 4] = avg_vc1_mspel_mc01_mmx2;
         dsp->avg_vc1_mspel_pixels_tab[ 8] = avg_vc1_mspel_mc02_mmx2;
@@ -810,7 +810,7 @@ void ff_vc1dsp_init_mmx(VC1DSPContext *dsp)
     if (mm_flags & AV_CPU_FLAG_MMX) {
     }
 
-    if (mm_flags & AV_CPU_FLAG_MMX2) {
+    if (mm_flags & AV_CPU_FLAG_MMXEXT) {
         ASSIGN_LF(mmx2);
     }
 
diff --git a/libavcodec/x86/vp3dsp_init.c b/libavcodec/x86/vp3dsp_init.c
index e0366d0f06..8905452dcf 100644
--- a/libavcodec/x86/vp3dsp_init.c
+++ b/libavcodec/x86/vp3dsp_init.c
@@ -49,7 +49,7 @@ av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags)
     }
 #endif
 
-    if (HAVE_MMX2 && cpuflags & AV_CPU_FLAG_MMX2) {
+    if (HAVE_MMXEXT && cpuflags & AV_CPU_FLAG_MMXEXT) {
         c->idct_dc_add = ff_vp3_idct_dc_add_mmx2;
 
         if (!(flags & CODEC_FLAG_BITEXACT)) {
diff --git a/libavcodec/x86/vp8dsp-init.c b/libavcodec/x86/vp8dsp-init.c
index 33a908ed76..8c17fa0382 100644
--- a/libavcodec/x86/vp8dsp-init.c
+++ b/libavcodec/x86/vp8dsp-init.c
@@ -350,7 +350,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
 
     /* note that 4-tap width=16 functions are missing because w=16
      * is only used for luma, and luma is always a copy or sixtap. */
-    if (mm_flags & AV_CPU_FLAG_MMX2) {
+    if (mm_flags & AV_CPU_FLAG_MMXEXT) {
         VP8_MC_FUNC(2, 4, mmx2);
         VP8_BILINEAR_MC_FUNC(2, 4, mmx2);
 #if ARCH_X86_32