mov    r3, r1m    lea    r4, [r3*3]    sub    r0, 4    lea    r1, [r0+r4]    %assign pad 0x78-(stack_offset&15)    SUB    esp, pad%define pix_tmp esp+12    ; transpose 6x16 -> tmp space    TRANSPOSE6x8_MEM  PASS8ROWS(r0, r1, r3, r4), pix_tmp    lea    r0, [r0+r3*8]    lea    r1, [r1+r3*8]    TRANSPOSE6x8_MEM  PASS8ROWS(r0, r1, r3, r4), pix_tmp+8    ; vertical filter    lea    r0, [pix_tmp+0x30]    PUSH   dword r4m    PUSH   dword r3m    PUSH   dword r2m    PUSH   dword 16    PUSH   dword r0    call   x264_deblock_%2_luma_%1%ifidn %2, v8    add    dword [esp   ], 8 ; pix_tmp+0x38    add    dword [esp+16], 2 ; tc0+2    call   x264_deblock_%2_luma_%1%endif    ADD    esp, 20    ; transpose 16x4 -> original space  (only the middle 4 rows were changed by the filter)    mov    r0, r0mp    sub    r0, 2    lea    r1, [r0+r4]    movq   m0, [pix_tmp+0x10]    movq   m1, [pix_tmp+0x20]    movq   m2, [pix_tmp+0x30]    movq   m3, [pix_tmp+0x40]    TRANSPOSE8x4_STORE  PASS8ROWS(r0, r1, r3, r4)    lea    r0, [r0+r3*8]    lea    r1, [r1+r3*8]    movq   m0, [pix_tmp+0x18]    movq   m1, [pix_tmp+0x28]    movq   m2, [pix_tmp+0x38]    movq   m3, [pix_tmp+0x48]    TRANSPOSE8x4_STORE  PASS8ROWS(r0, r1, r3, r4)    ADD    esp, pad    RET%endmacro ; DEBLOCK_LUMAINIT_MMXDEBLOCK_LUMA mmxext, v8, 8INIT_XMMDEBLOCK_LUMA sse2, v, 16%endif ; ARCH%macro LUMA_INTRA_P012 4 ; p0..p3 in memory    mova  t0, p2    mova  t1, p0    pavgb t0, p1    pavgb t1, q0    pavgb t0, t1 ; ((p2+p1+1)/2 + (p0+q0+1)/2 + 1)/2    mova  t5, t1    mova  t2, p2    mova  t3, p0    paddb t2, p1    paddb t3, q0    paddb t2, t3    mova  t3, t2    mova  t4, t2    psrlw t2, 1    pavgb t2, mpb_00    pxor  t2, t0    pand  t2, mpb_01    psubb t0, t2 ; p1' = (p2+p1+p0+q0+2)/4;    mova  t1, p2    mova  t2, p2    pavgb t1, q1    psubb t2, q1    paddb t3, t3    psubb t3, t2 ; p2+2*p1+2*p0+2*q0+q1    pand  t2, mpb_01    psubb t1, t2    pavgb t1, p1    pavgb t1, t5 ; (((p2+q1)/2 + p1+1)/2 + (p0+q0+1)/2 + 1)/2    psrlw t3, 2    pavgb t3, mpb_00    pxor  t3, t1    pand  t3, mpb_01    psubb t1, t3 ; p0'a = (p2+2*p1+2*p0+2*q0+q1+4)/8    mova  t3, p0    mova  t2, p0    pxor  t3, q1    pavgb t2, q1    pand  t3, mpb_01    psubb t2, t3    pavgb t2, p1 ; p0'b = (2*p1+p0+q0+2)/4    pxor  t1, t2    pxor  t2, p0    pand  t1, mask1p    pand  t2, mask0    pxor  t1, t2    pxor  t1, p0    mova  %1, t1 ; store p0    mova  t1, %4 ; p3    mova  t2, t1    pavgb t1, p2    paddb t2, p2    pavgb t1, t0 ; (p3+p2+1)/2 + (p2+p1+p0+q0+2)/4    paddb t2, t2    paddb t2, t4 ; 2*p3+3*p2+p1+p0+q0    psrlw t2, 2    pavgb t2, mpb_00    pxor  t2, t1    pand  t2, mpb_01    psubb t1, t2 ; p2' = (2*p3+3*p2+p1+p0+q0+4)/8    pxor  t0, p1    pxor  t1, p2    pand  t0, mask1p    pand  t1, mask1p    pxor  t0, p1    pxor  t1, p2    mova  %2, t0 ; store p1    mova  %3, t1 ; store p2%endmacro%macro LUMA_INTRA_SWAP_PQ 0    %define q1 m0    %define q0 m1    %define p0 m2    %define p1 m3    %define p2 q2    %define mask1p mask1q%endmacro%macro DEBLOCK_LUMA_INTRA 2    %define p1 m0    %define p0 m1    %define q0 m2    %define q1 m3    %define t0 m4    %define t1 m5    %define t2 m6    %define t3 m7%ifdef ARCH_X86_64    %define p2 m8    %define q2 m9    %define t4 m10    %define t5 m11    %define mask0 m12    %define mask1p m13    %define mask1q [rsp-24]    %define mpb_00 m14    %define mpb_01 m15%else    %define spill(x) [esp+16*x+((stack_offset+4)&15)]    %define p2 [r4+r1]    %define q2 [r0+2*r1]    %define t4 spill(0)    %define t5 spill(1)    %define mask0 spill(2)    %define mask1p spill(3)    %define mask1q spill(4)    %define mpb_00 [pb_00 GLOBAL]    %define mpb_01 [pb_01 GLOBAL]%endif;-----------------------------------------------------------------------------; void x264_deblock_v_luma_intra_sse2( uint8_t *pix, int stride, int alpha, int beta );-----------------------------------------------------------------------------cglobal x264_deblock_%2_luma_intra_%1, 4,6,16%ifndef ARCH_X86_64    sub     esp, 0x60%endif    lea     r4, [r1*4]    lea     r5, [r1*3] ; 3*stride    dec     r2d        ; alpha-1    jl .end    neg     r4    dec     r3d        ; beta-1    jl .end    add     r4, r0     ; pix-4*stride    mova    p1, [r4+2*r1]    mova    p0, [r4+r5]    mova    q0, [r0]    mova    q1, [r0+r1]%ifdef ARCH_X86_64    pxor    mpb_00, mpb_00    mova    mpb_01, [pb_01 GLOBAL]    LOAD_MASK r2d, r3d, t5 ; m5=beta-1, t5=alpha-1, m7=mask0    SWAP    7, 12 ; m12=mask0    pavgb   t5, mpb_00    pavgb   t5, mpb_01 ; alpha/4+1    movdqa  p2, [r4+r1]    movdqa  q2, [r0+2*r1]    DIFF_GT2 p0, q0, t5, t0, t3 ; t0 = |p0-q0| > alpha/4+1    DIFF_GT2 p0, p2, m5, t2, t5 ; mask1 = |p2-p0| > beta-1    DIFF_GT2 q0, q2, m5, t4, t5 ; t4 = |q2-q0| > beta-1    pand    t0, mask0    pand    t4, t0    pand    t2, t0    mova    mask1q, t4    mova    mask1p, t2%else    LOAD_MASK r2d, r3d, t5 ; m5=beta-1, t5=alpha-1, m7=mask0    mova    m4, t5    mova    mask0, m7    pavgb   m4, [pb_00 GLOBAL]    pavgb   m4, [pb_01 GLOBAL] ; alpha/4+1    DIFF_GT2 p0, q0, m4, m6, m7 ; m6 = |p0-q0| > alpha/4+1    pand    m6, mask0    DIFF_GT2 p0, p2, m5, m4, m7 ; m4 = |p2-p0| > beta-1    pand    m4, m6    mova    mask1p, m4    DIFF_GT2 q0, q2, m5, m4, m7 ; m4 = |q2-q0| > beta-1    pand    m4, m6    mova    mask1q, m4%endif    LUMA_INTRA_P012 [r4+r5], [r4+2*r1], [r4+r1], [r4]    LUMA_INTRA_SWAP_PQ    LUMA_INTRA_P012 [r0], [r0+r1], [r0+2*r1], [r0+r5].end:%ifndef ARCH_X86_64    add     esp, 0x60%endif    RETINIT_MMX%ifdef ARCH_X86_64;-----------------------------------------------------------------------------; void x264_deblock_h_luma_intra_sse2( uint8_t *pix, int stride, int alpha, int beta );-----------------------------------------------------------------------------cglobal x264_deblock_h_luma_intra_%1, 4,7    movsxd r10, r1d    lea    r11, [r10*3]    lea    r6,  [r0-4]    lea    r5,  [r0-4+r11]    sub    rsp, 0x88    %define pix_tmp rsp    ; transpose 8x16 -> tmp space    TRANSPOSE8x8_MEM  PASS8ROWS(r6, r5, r10, r11), PASS8ROWS(pix_tmp, pix_tmp+0x30, 0x10, 0x30)    lea    r6, [r6+r10*8]    lea    r5, [r5+r10*8]    TRANSPOSE8x8_MEM  PASS8ROWS(r6, r5, r10, r11), PASS8ROWS(pix_tmp+8, pix_tmp+0x38, 0x10, 0x30)    lea    r0,  [pix_tmp+0x40]    mov    r1,  0x10    call   x264_deblock_v_luma_intra_%1    ; transpose 16x6 -> original space (but we can't write only 6 pixels, so really 16x8)    lea    r5, [r6+r11]    TRANSPOSE8x8_MEM  PASS8ROWS(pix_tmp+8, pix_tmp+0x38, 0x10, 0x30), PASS8ROWS(r6, r5, r10, r11)    shl    r10, 3    sub    r6,  r10    sub    r5,  r10    shr    r10, 3    TRANSPOSE8x8_MEM  PASS8ROWS(pix_tmp, pix_tmp+0x30, 0x10, 0x30), PASS8ROWS(r6, r5, r10, r11)    add    rsp, 0x88    RET%elsecglobal x264_deblock_h_luma_intra_%1, 2,4    lea    r3,  [r1*3]    sub    r0,  4    lea    r2,  [r0+r3]%assign pad 0x8c-(stack_offset&15)    SUB    rsp, pad    %define pix_tmp rsp    ; transpose 8x16 -> tmp space    TRANSPOSE8x8_MEM  PASS8ROWS(r0, r2, r1, r3), PASS8ROWS(pix_tmp, pix_tmp+0x30, 0x10, 0x30)    lea    r0,  [r0+r1*8]    lea    r2,  [r2+r1*8]    TRANSPOSE8x8_MEM  PASS8ROWS(r0, r2, r1, r3), PASS8ROWS(pix_tmp+8, pix_tmp+0x38, 0x10, 0x30)    lea    r0,  [pix_tmp+0x40]    PUSH   dword r3m    PUSH   dword r2m    PUSH   dword 16    PUSH   r0    call   x264_deblock_%2_luma_intra_%1%ifidn %2, v8    add    dword [rsp], 8 ; pix_tmp+8    call   x264_deblock_%2_luma_intra_%1%endif    ADD    esp, 16    mov    r1,  r1m    mov    r0,  r0mp    lea    r3,  [r1*3]    sub    r0,  4    lea    r2,  [r0+r3]    ; transpose 16x6 -> original space (but we can't write only 6 pixels, so really 16x8)    TRANSPOSE8x8_MEM  PASS8ROWS(pix_tmp, pix_tmp+0x30, 0x10, 0x30), PASS8ROWS(r0, r2, r1, r3)    lea    r0,  [r0+r1*8]    lea    r2,  [r2+r1*8]    TRANSPOSE8x8_MEM  PASS8ROWS(pix_tmp+8, pix_tmp+0x38, 0x10, 0x30), PASS8ROWS(r0, r2, r1, r3)    ADD    rsp, pad    RET%endif ; ARCH_X86_64%endmacro ; DEBLOCK_LUMA_INTRAINIT_XMMDEBLOCK_LUMA_INTRA sse2, v%ifndef ARCH_X86_64INIT_MMXDEBLOCK_LUMA_INTRA mmxext, v8%endifINIT_MMX%macro CHROMA_V_START 0    dec    r2d      ; alpha-1    dec    r3d      ; beta-1    mov    t5, r0    sub    t5, r1    sub    t5, r1%endmacro%macro CHROMA_H_START 0    dec    r2d    dec    r3d    sub    r0, 2    lea    t6, [r1*3]    mov    t5, r0    add    r0, t6%endmacro%define t5 r5%define t6 r6;-----------------------------------------------------------------------------; void x264_deblock_v_chroma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );-----------------------------------------------------------------------------cglobal x264_deblock_v_chroma_mmxext, 5,6    CHROMA_V_START    movq  m0, [t5]    movq  m1, [t5+r1]    movq  m2, [r0]    movq  m3, [r0+r1]    call chroma_inter_body_mmxext    movq  [t5+r1], m1    movq  [r0], m2    RET;-----------------------------------------------------------------------------; void x264_deblock_h_chroma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );-----------------------------------------------------------------------------cglobal x264_deblock_h_chroma_mmxext, 5,7%ifdef ARCH_X86_64    %define buf0 [rsp-24]    %define buf1 [rsp-16]%else    %define buf0 r0m    %define buf1 r2m%endif    CHROMA_H_START    TRANSPOSE4x8_LOAD  PASS8ROWS(t5, r0, r1, t6)    movq  buf0, m0    movq  buf1, m3    call chroma_inter_body_mmxext    movq  m0, buf0    movq  m3, buf1    TRANSPOSE8x4_STORE PASS8ROWS(t5, r0, r1, t6)    RETALIGN 16chroma_inter_body_mmxext:    LOAD_MASK  r2d, r3d    movd       m6, [r4] ; tc0    punpcklbw  m6, m6    pand       m7, m6    DEBLOCK_P0_Q0    ret; in: %1=p0 %2=p1 %3=q1; out: p0 = (p0 + q1 + 2*p1 + 2) >> 2%macro CHROMA_INTRA_P0 3    movq    m4, %1    pxor    m4, %3    pand    m4, [pb_01 GLOBAL] ; m4 = (p0^q1)&1    pavgb   %1, %3    psubusb %1, m4    pavgb   %1, %2             ; dst = avg(p1, avg(p0,q1) - ((p0^q1)&1))%endmacro%define t5 r4%define t6 r5;-----------------------------------------------------------------------------; void x264_deblock_v_chroma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );-----------------------------------------------------------------------------cglobal x264_deblock_v_chroma_intra_mmxext, 4,5    CHROMA_V_START    movq  m0, [t5]    movq  m1, [t5+r1]    movq  m2, [r0]    movq  m3, [r0+r1]    call chroma_intra_body_mmxext    movq  [t5+r1], m1    movq  [r0], m2    RET;-----------------------------------------------------------------------------; void x264_deblock_h_chroma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );-----------------------------------------------------------------------------cglobal x264_deblock_h_chroma_intra_mmxext, 4,6    CHROMA_H_START    TRANSPOSE4x8_LOAD  PASS8ROWS(t5, r0, r1, t6)    call chroma_intra_body_mmxext    TRANSPOSE8x4_STORE PASS8ROWS(t5, r0, r1, t6)    RETALIGN 16chroma_intra_body_mmxext:    LOAD_MASK r2d, r3d    movq   m5, m1    movq   m6, m2    CHROMA_INTRA_P0  m1, m0, m3    CHROMA_INTRA_P0  m2, m3, m0    psubb  m1, m5    psubb  m2, m6    pand   m1, m7    pand   m2, m7    paddb  m1, m5    paddb  m2, m6    ret