huffyuvencdsp: move functions only used by huffyuv from lossless_videodsp

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

1 files changed, 113 insertions, 0 deletions

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