huffyuvencdsp: move functions only used by huffyuv from lossless_videodsp

Signed-off-by: James Almer <jamrial@gmail.com>

14 files changed, 196 insertions, 126 deletions

diff --git a/configure b/configure
index d393983687..d4232f5d42 100755
--- a/configure
+++ b/configure
@@ -2430,7 +2430,7 @@ hap_encoder_deps="libsnappy"
 hap_encoder_select="texturedspenc"
 hevc_decoder_select="bswapdsp cabac golomb videodsp"
 huffyuv_decoder_select="bswapdsp huffyuvdsp llviddsp"
-huffyuv_encoder_select="bswapdsp huffman huffyuvencdsp llviddsp"
+huffyuv_encoder_select="bswapdsp huffman huffyuvencdsp"
 iac_decoder_select="imc_decoder"
 imc_decoder_select="bswapdsp fft mdct sinewin"
 indeo3_decoder_select="hpeldsp"
diff --git a/libavcodec/huffyuv.c b/libavcodec/huffyuv.c
index 492155550f..e582060cc3 100644
--- a/libavcodec/huffyuv.c
+++ b/libavcodec/huffyuv.c
@@ -76,7 +76,6 @@ av_cold void ff_huffyuv_common_init(AVCodecContext *avctx)
     s->flags = avctx->flags;
 
     ff_bswapdsp_init(&s->bdsp);
-    ff_llviddsp_init(&s->llviddsp, avctx);
 
     s->width = avctx->width;
     s->height = avctx->height;
diff --git a/libavcodec/huffyuvdec.c b/libavcodec/huffyuvdec.c
index 1b4112ce2f..d0682040b3 100644
--- a/libavcodec/huffyuvdec.c
+++ b/libavcodec/huffyuvdec.c
@@ -298,6 +298,7 @@ static av_cold int decode_init(AVCodecContext *avctx)
         return ret;
 
     ff_huffyuvdsp_init(&s->hdsp);
+    ff_llviddsp_init(&s->llviddsp, avctx);
     memset(s->vlc, 0, 4 * sizeof(VLC));
 
     s->interlaced = avctx->height > 288;
diff --git a/libavcodec/huffyuvenc.c b/libavcodec/huffyuvenc.c
index 572de16a1f..f5bc99f196 100644
--- a/libavcodec/huffyuvenc.c
+++ b/libavcodec/huffyuvenc.c
@@ -43,7 +43,7 @@ static inline void diff_bytes(HYuvContext *s, uint8_t *dst,
     if (s->bps <= 8) {
         s->hencdsp.diff_bytes(dst, src0, src1, w);
     } else {
-        s->llviddsp.diff_int16((uint16_t *)dst, (const uint16_t *)src0, (const uint16_t *)src1, s->n - 1, w);
+        s->hencdsp.diff_int16((uint16_t *)dst, (const uint16_t *)src0, (const uint16_t *)src1, s->n - 1, w);
     }
 }
 
@@ -84,7 +84,7 @@ static inline int sub_left_prediction(HYuvContext *s, uint8_t *dst,
                 dst16[i] = temp - left;
                 left   = temp;
             }
-            s->llviddsp.diff_int16(dst16 + 16, src16 + 16, src16 + 15, s->n - 1, w - 16);
+            s->hencdsp.diff_int16(dst16 + 16, src16 + 16, src16 + 15, s->n - 1, w - 16);
             return src16[w-1];
         }
     }
@@ -158,7 +158,7 @@ static void sub_median_prediction(HYuvContext *s, uint8_t *dst, const uint8_t *s
     if (s->bps <= 8) {
         s->hencdsp.sub_hfyu_median_pred(dst, src1, src2, w , left, left_top);
     } else {
-        s->llviddsp.sub_hfyu_median_pred_int16((uint16_t *)dst, (const uint16_t *)src1, (const uint16_t *)src2, s->n - 1, w , left, left_top);
+        s->hencdsp.sub_hfyu_median_pred_int16((uint16_t *)dst, (const uint16_t *)src1, (const uint16_t *)src2, s->n - 1, w , left, left_top);
     }
 }
 
@@ -217,7 +217,7 @@ static av_cold int encode_init(AVCodecContext *avctx)
     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(avctx->pix_fmt);
 
     ff_huffyuv_common_init(avctx);
-    ff_huffyuvencdsp_init(&s->hencdsp);
+    ff_huffyuvencdsp_init(&s->hencdsp, avctx);
 
     avctx->extradata = av_mallocz(3*MAX_N + 4);
     if (s->flags&AV_CODEC_FLAG_PASS1) {
diff --git a/libavcodec/huffyuvencdsp.c b/libavcodec/huffyuvencdsp.c
index fdcd0b06aa..f051021094 100644
--- a/libavcodec/huffyuvencdsp.c
+++ b/libavcodec/huffyuvencdsp.c
@@ -53,6 +53,32 @@ static void diff_bytes_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
         dst[i + 0] = src1[i + 0] - src2[i + 0];
 }
 
+static void diff_int16_c(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w){
+    long i;
+#if !HAVE_FAST_UNALIGNED
+    if((long)src2 & (sizeof(long)-1)){
+        for(i=0; i+3<w; i+=4){
+            dst[i+0] = (src1[i+0]-src2[i+0]) & mask;
+            dst[i+1] = (src1[i+1]-src2[i+1]) & mask;
+            dst[i+2] = (src1[i+2]-src2[i+2]) & mask;
+            dst[i+3] = (src1[i+3]-src2[i+3]) & mask;
+        }
+    }else
+#endif
+    {
+        unsigned long pw_lsb = (mask >> 1) * 0x0001000100010001ULL;
+        unsigned long pw_msb = pw_lsb +  0x0001000100010001ULL;
+
+        for (i = 0; i <= w - (int)sizeof(long)/2; i += sizeof(long)/2) {
+            long a = *(long*)(src1+i);
+            long b = *(long*)(src2+i);
+            *(long*)(dst+i) = ((a|pw_msb) - (b&pw_lsb)) ^ ((a^b^pw_msb)&pw_msb);
+        }
+    }
+    for (; i<w; i++)
+        dst[i] = (src1[i] - src2[i]) & mask;
+}
+
 static void sub_hfyu_median_pred_c(uint8_t *dst, const uint8_t *src1,
                                    const uint8_t *src2, intptr_t w,
                                    int *left, int *left_top)
@@ -74,11 +100,31 @@ static void sub_hfyu_median_pred_c(uint8_t *dst, const uint8_t *src1,
     *left_top = lt;
 }
 
-av_cold void ff_huffyuvencdsp_init(HuffYUVEncDSPContext *c)
+static void sub_hfyu_median_pred_int16_c(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w, int *left, int *left_top){
+    int i;
+    uint16_t l, lt;
+
+    l  = *left;
+    lt = *left_top;
+
+    for(i=0; i<w; i++){
+        const int pred = mid_pred(l, src1[i], (l + src1[i] - lt) & mask);
+        lt = src1[i];
+        l  = src2[i];
+        dst[i] = (l - pred) & mask;
+    }
+
+    *left     = l;
+    *left_top = lt;
+}
+
+av_cold void ff_huffyuvencdsp_init(HuffYUVEncDSPContext *c, AVCodecContext *avctx)
 {
     c->diff_bytes           = diff_bytes_c;
+    c->diff_int16           = diff_int16_c;
     c->sub_hfyu_median_pred = sub_hfyu_median_pred_c;
+    c->sub_hfyu_median_pred_int16 = sub_hfyu_median_pred_int16_c;
 
     if (ARCH_X86)
-        ff_huffyuvencdsp_init_x86(c);
+        ff_huffyuvencdsp_init_x86(c, avctx);
 }
diff --git a/libavcodec/huffyuvencdsp.h b/libavcodec/huffyuvencdsp.h
index 9d09095374..141dad8692 100644
--- a/libavcodec/huffyuvencdsp.h
+++ b/libavcodec/huffyuvencdsp.h
@@ -21,11 +21,18 @@
 
 #include <stdint.h>
 
+#include "avcodec.h"
+
 typedef struct HuffYUVEncDSPContext {
     void (*diff_bytes)(uint8_t *dst /* align 16 */,
                        const uint8_t *src1 /* align 16 */,
                        const uint8_t *src2 /* align 1 */,
                        intptr_t w);
+    void (*diff_int16)(uint16_t *dst /* align 16 */,
+                       const uint16_t *src1 /* align 16 */,
+                       const uint16_t *src2 /* align 1 */,
+                       unsigned mask, int w);
+
     /**
      * Subtract HuffYUV's variant of median prediction.
      * Note, this might read from src1[-1], src2[-1].
@@ -33,9 +40,12 @@ typedef struct HuffYUVEncDSPContext {
     void (*sub_hfyu_median_pred)(uint8_t *dst, const uint8_t *src1,
                                  const uint8_t *src2, intptr_t w,
                                  int *left, int *left_top);
+    void (*sub_hfyu_median_pred_int16)(uint16_t *dst, const uint16_t *src1,
+                                       const uint16_t *src2, unsigned mask,
+                                       int w, int *left, int *left_top);
 } HuffYUVEncDSPContext;
 
-void ff_huffyuvencdsp_init(HuffYUVEncDSPContext *c);
-void ff_huffyuvencdsp_init_x86(HuffYUVEncDSPContext *c);
+void ff_huffyuvencdsp_init(HuffYUVEncDSPContext *c, AVCodecContext *avctx);
+void ff_huffyuvencdsp_init_x86(HuffYUVEncDSPContext *c, AVCodecContext *avctx);
 
 #endif /* AVCODEC_HUFFYUVENCDSP_H */
diff --git a/libavcodec/lossless_videodsp.c b/libavcodec/lossless_videodsp.c
index 5440ce2e31..b93d4e7214 100644
--- a/libavcodec/lossless_videodsp.c
+++ b/libavcodec/lossless_videodsp.c
@@ -92,32 +92,6 @@ static void add_int16_c(uint16_t *dst, const uint16_t *src, unsigned mask, int w
         dst[i] = (dst[i] + src[i]) & mask;
 }
 
-static void diff_int16_c(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w){
-    long i;
-#if !HAVE_FAST_UNALIGNED
-    if((long)src2 & (sizeof(long)-1)){
-        for(i=0; i+3<w; i+=4){
-            dst[i+0] = (src1[i+0]-src2[i+0]) & mask;
-            dst[i+1] = (src1[i+1]-src2[i+1]) & mask;
-            dst[i+2] = (src1[i+2]-src2[i+2]) & mask;
-            dst[i+3] = (src1[i+3]-src2[i+3]) & mask;
-        }
-    }else
-#endif
-    {
-        unsigned long pw_lsb = (mask >> 1) * 0x0001000100010001ULL;
-        unsigned long pw_msb = pw_lsb +  0x0001000100010001ULL;
-
-        for (i = 0; i <= w - (int)sizeof(long)/2; i += sizeof(long)/2) {
-            long a = *(long*)(src1+i);
-            long b = *(long*)(src2+i);
-            *(long*)(dst+i) = ((a|pw_msb) - (b&pw_lsb)) ^ ((a^b^pw_msb)&pw_msb);
-        }
-    }
-    for (; i<w; i++)
-        dst[i] = (src1[i] - src2[i]) & mask;
-}
-
 static void add_hfyu_median_pred_int16_c(uint16_t *dst, const uint16_t *src, const uint16_t *diff, unsigned mask, int w, int *left, int *left_top){
     int i;
     uint16_t l, lt;
@@ -135,24 +109,6 @@ static void add_hfyu_median_pred_int16_c(uint16_t *dst, const uint16_t *src, con
     *left_top = lt;
 }
 
-static void sub_hfyu_median_pred_int16_c(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w, int *left, int *left_top){
-    int i;
-    uint16_t l, lt;
-
-    l  = *left;
-    lt = *left_top;
-
-    for(i=0; i<w; i++){
-        const int pred = mid_pred(l, src1[i], (l + src1[i] - lt) & mask);
-        lt = src1[i];
-        l  = src2[i];
-        dst[i] = (l - pred) & mask;
-    }
-
-    *left     = l;
-    *left_top = lt;
-}
-
 static int add_hfyu_left_pred_int16_c(uint16_t *dst, const uint16_t *src, unsigned mask, int w, unsigned acc){
     int i;
 
@@ -180,10 +136,8 @@ void ff_llviddsp_init(LLVidDSPContext *c, AVCodecContext *avctx)
     c->add_left_pred              = add_left_pred_c;
 
     c->add_int16 = add_int16_c;
-    c->diff_int16= diff_int16_c;
     c->add_hfyu_left_pred_int16   = add_hfyu_left_pred_int16_c;
     c->add_hfyu_median_pred_int16 = add_hfyu_median_pred_int16_c;
-    c->sub_hfyu_median_pred_int16 = sub_hfyu_median_pred_int16_c;
 
     if (ARCH_X86)
         ff_llviddsp_init_x86(c, avctx);
diff --git a/libavcodec/lossless_videodsp.h b/libavcodec/lossless_videodsp.h
index e8ba175a5d..7f3168339f 100644
--- a/libavcodec/lossless_videodsp.h
+++ b/libavcodec/lossless_videodsp.h
@@ -35,9 +35,7 @@ typedef struct LLVidDSPContext {
                          intptr_t w, int left);
 
     void (*add_int16)(uint16_t *dst/*align 16*/, const uint16_t *src/*align 16*/, unsigned mask, int w);
-    void (*diff_int16)(uint16_t *dst/*align 16*/, const uint16_t *src1/*align 16*/, const uint16_t *src2/*align 1*/, unsigned mask, int w);
 
-    void (*sub_hfyu_median_pred_int16)(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w, int *left, int *left_top);
     void (*add_hfyu_median_pred_int16)(uint16_t *dst, const uint16_t *top, const uint16_t *diff, unsigned mask, int w, int *left, int *left_top);
     int  (*add_hfyu_left_pred_int16)(uint16_t *dst, const uint16_t *src, unsigned mask, int w, unsigned left);
 } LLVidDSPContext;
diff --git a/libavcodec/pngenc.c b/libavcodec/pngenc.c
index 51ae094770..3aeff83f42 100644
--- a/libavcodec/pngenc.c
+++ b/libavcodec/pngenc.c
@@ -1015,7 +1015,7 @@ FF_DISABLE_DEPRECATION_WARNINGS
 FF_ENABLE_DEPRECATION_WARNINGS
 #endif
 
-    ff_huffyuvencdsp_init(&s->hdsp);
+    ff_huffyuvencdsp_init(&s->hdsp, avctx);
 
 #if FF_API_PRIVATE_OPT
 FF_DISABLE_DEPRECATION_WARNINGS
diff --git a/libavcodec/utvideoenc.c b/libavcodec/utvideoenc.c
index fd27a15fd8..d82f6a3872 100644
--- a/libavcodec/utvideoenc.c
+++ b/libavcodec/utvideoenc.c
@@ -120,7 +120,7 @@ static av_cold int utvideo_encode_init(AVCodecContext *avctx)
     }
 
     ff_bswapdsp_init(&c->bdsp);
-    ff_huffyuvencdsp_init(&c->hdsp);
+    ff_huffyuvencdsp_init(&c->hdsp, avctx);
 
 #if FF_API_PRIVATE_OPT
 FF_DISABLE_DEPRECATION_WARNINGS
diff --git a/libavcodec/x86/huffyuvencdsp.asm b/libavcodec/x86/huffyuvencdsp.asm
index a55a1de65d..78ad202249 100644
--- a/libavcodec/x86/huffyuvencdsp.asm
+++ b/libavcodec/x86/huffyuvencdsp.asm
@@ -148,3 +148,116 @@ DIFF_BYTES_PROLOGUE
     DIFF_BYTES_BODY    u, u
 %undef i
 %endif
+
+%macro INT16_LOOP 2 ; %1 = a/u (aligned/unaligned), %2 = add/sub
+    movd    m4, maskd
+    SPLATW  m4, m4
+    add     wd, wd
+    test    wq, 2*mmsize - 1
+    jz %%.tomainloop
+    push  tmpq
+%%.wordloop:
+    sub     wq, 2
+%ifidn %2, add
+    mov   tmpw, [srcq+wq]
+    add   tmpw, [dstq+wq]
+%else
+    mov   tmpw, [src1q+wq]
+    sub   tmpw, [src2q+wq]
+%endif
+    and   tmpw, maskw
+    mov     [dstq+wq], tmpw
+    test    wq, 2*mmsize - 1
+    jnz %%.wordloop
+    pop   tmpq
+%%.tomainloop:
+%ifidn %2, add
+    add     srcq, wq
+%else
+    add     src1q, wq
+    add     src2q, wq
+%endif
+    add     dstq, wq
+    neg     wq
+    jz      %%.end
+%%.loop:
+%ifidn %2, add
+    mov%1   m0, [srcq+wq]
+    mov%1   m1, [dstq+wq]
+    mov%1   m2, [srcq+wq+mmsize]
+    mov%1   m3, [dstq+wq+mmsize]
+%else
+    mov%1   m0, [src1q+wq]
+    mov%1   m1, [src2q+wq]
+    mov%1   m2, [src1q+wq+mmsize]
+    mov%1   m3, [src2q+wq+mmsize]
+%endif
+    p%2w    m0, m1
+    p%2w    m2, m3
+    pand    m0, m4
+    pand    m2, m4
+    mov%1   [dstq+wq]       , m0
+    mov%1   [dstq+wq+mmsize], m2
+    add     wq, 2*mmsize
+    jl %%.loop
+%%.end:
+    RET
+%endmacro
+
+%if ARCH_X86_32
+INIT_MMX mmx
+cglobal diff_int16, 5,5,5, dst, src1, src2, mask, w, tmp
+    INT16_LOOP a, sub
+%endif
+
+INIT_XMM sse2
+cglobal diff_int16, 5,5,5, dst, src1, src2, mask, w, tmp
+    test src1q, mmsize-1
+    jnz .unaligned
+    test src2q, mmsize-1
+    jnz .unaligned
+    test dstq, mmsize-1
+    jnz .unaligned
+    INT16_LOOP a, sub
+.unaligned:
+    INT16_LOOP u, sub
+
+INIT_MMX mmxext
+cglobal sub_hfyu_median_pred_int16, 7,7,0, dst, src1, src2, mask, w, left, left_top
+    add      wd, wd
+    movd    mm7, maskd
+    SPLATW  mm7, mm7
+    movq    mm0, [src1q]
+    movq    mm2, [src2q]
+    psllq   mm0, 16
+    psllq   mm2, 16
+    movd    mm6, [left_topq]
+    por     mm0, mm6
+    movd    mm6, [leftq]
+    por     mm2, mm6
+    xor     maskq, maskq
+.loop:
+    movq    mm1, [src1q + maskq]
+    movq    mm3, [src2q + maskq]
+    movq    mm4, mm2
+    psubw   mm2, mm0
+    paddw   mm2, mm1
+    pand    mm2, mm7
+    movq    mm5, mm4
+    pmaxsw  mm4, mm1
+    pminsw  mm1, mm5
+    pminsw  mm4, mm2
+    pmaxsw  mm4, mm1
+    psubw   mm3, mm4
+    pand    mm3, mm7
+    movq    [dstq + maskq], mm3
+    add     maskq, 8
+    movq    mm0, [src1q + maskq - 2]
+    movq    mm2, [src2q + maskq - 2]
+    cmp     maskq, wq
+        jb .loop
+    movzx maskd, word [src1q + wq - 2]
+    mov [left_topq], maskd
+    movzx maskd, word [src2q + wq - 2]
+    mov [leftq], maskd
+    RET
diff --git a/libavcodec/x86/huffyuvencdsp_mmx.c b/libavcodec/x86/huffyuvencdsp_mmx.c
index 9767b212da..2402021823 100644
--- a/libavcodec/x86/huffyuvencdsp_mmx.c
+++ b/libavcodec/x86/huffyuvencdsp_mmx.c
@@ -24,6 +24,7 @@
 
 #include "libavutil/attributes.h"
 #include "libavutil/cpu.h"
+#include "libavutil/pixdesc.h"
 #include "libavutil/x86/asm.h"
 #include "libavutil/x86/cpu.h"
 #include "libavcodec/huffyuvencdsp.h"
@@ -35,6 +36,12 @@ void ff_diff_bytes_sse2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
                         intptr_t w);
 void ff_diff_bytes_avx2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
                         intptr_t w);
+void ff_diff_int16_mmx (uint16_t *dst, const uint16_t *src1, const uint16_t *src2,
+                        unsigned mask, int w);
+void ff_diff_int16_sse2(uint16_t *dst, const uint16_t *src1, const uint16_t *src2,
+                        unsigned mask, int w);
+void ff_sub_hfyu_median_pred_int16_mmxext(uint16_t *dst, const uint16_t *src1, const uint16_t *src2,
+                                          unsigned mask, int w, int *left, int *left_top);
 
 #if HAVE_INLINE_ASM
 
@@ -80,12 +87,14 @@ static void sub_hfyu_median_pred_mmxext(uint8_t *dst, const uint8_t *src1,
 
 #endif /* HAVE_INLINE_ASM */
 
-av_cold void ff_huffyuvencdsp_init_x86(HuffYUVEncDSPContext *c)
+av_cold void ff_huffyuvencdsp_init_x86(HuffYUVEncDSPContext *c, AVCodecContext *avctx)
 {
     av_unused int cpu_flags = av_get_cpu_flags();
+    const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(avctx->pix_fmt);
 
     if (ARCH_X86_32 && EXTERNAL_MMX(cpu_flags)) {
         c->diff_bytes = ff_diff_bytes_mmx;
+        c->diff_int16 = ff_diff_int16_mmx;
     }
 
 #if HAVE_INLINE_ASM
@@ -94,8 +103,13 @@ av_cold void ff_huffyuvencdsp_init_x86(HuffYUVEncDSPContext *c)
     }
 #endif /* HAVE_INLINE_ASM */
 
+    if (EXTERNAL_MMXEXT(cpu_flags) && pix_desc && pix_desc->comp[0].depth<16) {
+        c->sub_hfyu_median_pred_int16 = ff_sub_hfyu_median_pred_int16_mmxext;
+    }
+
     if (EXTERNAL_SSE2(cpu_flags)) {
         c->diff_bytes = ff_diff_bytes_sse2;
+        c->diff_int16 = ff_diff_int16_sse2;
     }
 
     if (EXTERNAL_AVX2_FAST(cpu_flags)) {
diff --git a/libavcodec/x86/lossless_videodsp.asm b/libavcodec/x86/lossless_videodsp.asm
index a6ce5fe62b..bcc40ec061 100644
--- a/libavcodec/x86/lossless_videodsp.asm
+++ b/libavcodec/x86/lossless_videodsp.asm
@@ -288,25 +288,6 @@ cglobal add_int16, 4,4,5, dst, src, mask, w, tmp
 .unaligned:
     INT16_LOOP u, add
 
-%if ARCH_X86_32
-INIT_MMX mmx
-cglobal diff_int16, 5,5,5, dst, src1, src2, mask, w, tmp
-    INT16_LOOP a, sub
-%endif
-
-INIT_XMM sse2
-cglobal diff_int16, 5,5,5, dst, src1, src2, mask, w, tmp
-    test src1q, mmsize-1
-    jnz .unaligned
-    test src2q, mmsize-1
-    jnz .unaligned
-    test dstq, mmsize-1
-    jnz .unaligned
-    INT16_LOOP a, sub
-.unaligned:
-    INT16_LOOP u, sub
-
-
 %macro ADD_HFYU_LEFT_LOOP_INT16 2 ; %1 = dst alignment (a/u), %2 = src alignment (a/u)
     add     wd, wd
     add     srcq, wq
@@ -443,42 +424,3 @@ cglobal add_hfyu_median_pred_int16, 7,7,0, dst, top, diff, mask, w, left, left_t
     movzx   r2d, word [topq-2]
     mov [left_topq], r2d
     RET
-
-cglobal sub_hfyu_median_pred_int16, 7,7,0, dst, src1, src2, mask, w, left, left_top
-    add      wd, wd
-    movd    mm7, maskd
-    SPLATW  mm7, mm7
-    movq    mm0, [src1q]
-    movq    mm2, [src2q]
-    psllq   mm0, 16
-    psllq   mm2, 16
-    movd    mm6, [left_topq]
-    por     mm0, mm6
-    movd    mm6, [leftq]
-    por     mm2, mm6
-    xor     maskq, maskq
-.loop:
-    movq    mm1, [src1q + maskq]
-    movq    mm3, [src2q + maskq]
-    movq    mm4, mm2
-    psubw   mm2, mm0
-    paddw   mm2, mm1
-    pand    mm2, mm7
-    movq    mm5, mm4
-    pmaxsw  mm4, mm1
-    pminsw  mm1, mm5
-    pminsw  mm4, mm2
-    pmaxsw  mm4, mm1
-    psubw   mm3, mm4
-    pand    mm3, mm7
-    movq    [dstq + maskq], mm3
-    add     maskq, 8
-    movq    mm0, [src1q + maskq - 2]
-    movq    mm2, [src2q + maskq - 2]
-    cmp     maskq, wq
-        jb .loop
-    movzx maskd, word [src1q + wq - 2]
-    mov [left_topq], maskd
-    movzx maskd, word [src2q + wq - 2]
-    mov [leftq], maskd
-    RET
diff --git a/libavcodec/x86/lossless_videodsp_init.c b/libavcodec/x86/lossless_videodsp_init.c
index 465feef81b..2dc662d8b1 100644
--- a/libavcodec/x86/lossless_videodsp_init.c
+++ b/libavcodec/x86/lossless_videodsp_init.c
@@ -41,12 +41,9 @@ int  ff_add_left_pred_sse4(uint8_t *dst, const uint8_t *src,
 
 void ff_add_int16_mmx(uint16_t *dst, const uint16_t *src, unsigned mask, int w);
 void ff_add_int16_sse2(uint16_t *dst, const uint16_t *src, unsigned mask, int w);
-void ff_diff_int16_mmx (uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w);
-void ff_diff_int16_sse2(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w);
 int ff_add_hfyu_left_pred_int16_ssse3(uint16_t *dst, const uint16_t *src, unsigned mask, int w, unsigned acc);
 int ff_add_hfyu_left_pred_int16_sse4(uint16_t *dst, const uint16_t *src, unsigned mask, int w, unsigned acc);
 void ff_add_hfyu_median_pred_int16_mmxext(uint16_t *dst, const uint16_t *top, const uint16_t *diff, unsigned mask, int w, int *left, int *left_top);
-void ff_sub_hfyu_median_pred_int16_mmxext(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w, int *left, int *left_top);
 
 #if HAVE_INLINE_ASM && HAVE_7REGS && ARCH_X86_32
 static void add_median_pred_cmov(uint8_t *dst, const uint8_t *top,
@@ -98,9 +95,7 @@ void ff_llviddsp_init_x86(LLVidDSPContext *c, AVCodecContext *avctx)
 
     if (ARCH_X86_32 && EXTERNAL_MMX(cpu_flags)) {
         c->add_bytes = ff_add_bytes_mmx;
-
         c->add_int16 = ff_add_int16_mmx;
-        c->diff_int16 = ff_diff_int16_mmx;
     }
 
     if (ARCH_X86_32 && EXTERNAL_MMXEXT(cpu_flags)) {
@@ -111,7 +106,6 @@ void ff_llviddsp_init_x86(LLVidDSPContext *c, AVCodecContext *avctx)
 
     if (EXTERNAL_MMXEXT(cpu_flags) && pix_desc && pix_desc->comp[0].depth<16) {
         c->add_hfyu_median_pred_int16 = ff_add_hfyu_median_pred_int16_mmxext;
-        c->sub_hfyu_median_pred_int16 = ff_sub_hfyu_median_pred_int16_mmxext;
     }
 
     if (EXTERNAL_SSE2(cpu_flags)) {
@@ -119,7 +113,6 @@ void ff_llviddsp_init_x86(LLVidDSPContext *c, AVCodecContext *avctx)
         c->add_median_pred = ff_add_median_pred_sse2;
 
         c->add_int16 = ff_add_int16_sse2;
-        c->diff_int16 = ff_diff_int16_sse2;
     }
 
     if (EXTERNAL_SSSE3(cpu_flags)) {