gistfile1.asm

pb_perm_unpackw_0: db  0,  1,  2,  3, 16, 17, 18, 19,  8,  9, 10, 11, 24, 25, 26, 27
pb_perm_unpackw_1: db  8,  9, 10, 11, 24, 25, 26, 27, 12, 13, 14, 15, 28, 29, 30, 31

[...]

    punpcklwd  m0, m1            ; m0 =  0,  1,  6,  7, 12, 13, 18, 19
    punpcklwd  m2, m3            ; m2 =  4,  5, 10, 11, 16, 17, 22, 23
    punpcklwd  m4, m5            ; m4 =  2,  3,  8,  9, 14, 15, 20, 21
%if cpuflag(xop)
    vpperm     m1, m0, m4, [pb_perm_unpackw_0]  ; m1 =  0,  1,  2,  3, 12, 13, 14, 15
    vpperm     m4, m4, m2, [pb_perm_unpackw_1]  ; m4 =  8,  9, 10, 11, 20, 21, 22, 23
    blendpd    m2, m0, 1010b                    ; m2 =  4,  5,  6,  7, 16, 17, 18, 19
    SWAP 0,4,2
%else
                                 ; blend dwords
    shufps     m1, m0, m2, q2020 ; m1 =  0,  1, 12, 13,  2,  3, 14, 15
    shufps     m0, m4, q2031     ; m0 =  6,  7, 18, 19,  4,  5, 16, 17
    shufps     m2, m4, q3131     ; m2 =  8,  9, 20, 21, 10, 11, 22, 23
                                 ; shuffle dwords
    pshufd     m0, m0, q1302     ; m0 =  4,  5,  6,  7, 16, 17, 18, 19
    pshufd     m1, m1, q3120     ; m1 =  0,  1,  2,  3, 12, 13, 14, 15
    pshufd     m2, m2, q3120     ; m2 =  8,  9, 10, 11, 20, 21, 22, 23
%endif

conv_s16p_to_s16_6ch