kolibrios-gitea/drivers/audio/infinity/mixer.asm

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;                                                              ;;
;; Copyright (C) KolibriOS team 2006-2014. All rights reserved. ;;
;; Distributed under terms of the GNU General Public License    ;;
;;                                                              ;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;   (C) copyright Serge 2006
;   email: infinity_sound@mail.ru


align 4

mix_list rd 32*3

align 4
proc new_mix stdcall, output:dword
           locals
             main_count   rd 1
             fpu_state    rb 528   ;512+16
           endl

        mov     [main_count], 32
        call    prepare_playlist
        cmp     [play_count], 0
        je      .clear

        lea     eax, [fpu_state+16]
        and     eax, -16            ;must be 16b aligned
        invoke  FpuSave

        call    update_streams
.mix:
        lea     eax, [mix_list]
        call    do_mix_list
        test    eax, eax
        je      .done

if USE_SSE2_MIXER

        cmp     eax, 1
        ja      @F
                                    ;use fast path
        mov     edi, [output]
        lea     edx, [mix_list]
        call    mix_fast
        jmp     .next
@@:
        cmp     eax, 2
        ja      @F

        mov     edi, [output]
        lea     edx, [mix_list]
        call    mix_fast_2_stream
        jmp     .next
@@:

end if

        lea     ebx, [mix_list]
        stdcall mix_all, [output], ebx, eax
.next:
        add     [output], 512
        dec     [main_count]
        jnz     .mix
.exit:
        lea     eax, [fpu_state+16]
        and     eax, -16
        invoke  FpuRestore
        ret
.done:
        mov     ecx, [main_count]
        shl     ecx, 7    ;ecx*= 512/4

        mov     edi, [output]
        xor     eax, eax
        cld
        rep stosd
        jmp     .exit
.clear:
        mov     edi, [output]
        mov     ecx, 4096
        xor     eax, eax
        cld
        rep stosd
        ret
endp

align 4
proc update_streams
           locals
             stream_index  dd ?
             event         rd 6
           endl

        mov     [stream_index], 0
.l1:
        mov     edx, [stream_index]
        mov     esi, [play_list+edx*4]

        add     dword [esi+STREAM.time_stamp], 4096
        adc     dword [esi+STREAM.time_stamp+4], 0
        mov     dword [esi+STREAM.last_ts], 0

        mov     eax, [esi+STREAM.out_rp]
        cmp     eax, [esi+STREAM.out_top]
        jb      @f
        sub     eax, 64*1024
@@:
        mov     [esi+STREAM.out_rp], eax

        cmp     [esi+STREAM.out_count], 16384
        ja      .skip

        test    [esi+STREAM.format], PCM_RING
        jnz     .ring

        stdcall refill, esi
.skip:
        inc     [stream_index]
        dec     [play_count]
        jnz     .l1
        ret
.ring:
        stdcall refill_ring, esi
        jmp     .skip
endp

align 4
proc refill stdcall, str:dword
           locals
             r_size    rd 1
           endl

        mov     ebx, [str]
        mov     edi, [ebx+STREAM.out_wp]
        cmp     edi, [ebx+STREAM.out_top]
        jb      @F
        sub     edi, 0x10000
        mov     [ebx+STREAM.out_wp], edi
@@:
        mov     eax, [ebx+STREAM.in_count]
        test    eax, eax
        jz      .done

        mov     ecx, [ebx+STREAM.r_size]
        cmp     eax, ecx
        jle     @F

        mov     eax, ecx
@@:
        mov     ecx, eax
        cmp     word [ebx+STREAM.format], PCM_1_16_8
        ja      @F

        shr     eax, 1                  ;two channles
@@:
        test    [ebx+STREAM.format], 1  ;even formats mono
        jz      @F

        shr     eax, 1                  ;eax= samples
@@:
        shl     eax, 15   ;eax*=32768 =r_end

        mov     [r_size], ecx

        mov     esi, [ebx+STREAM.in_rp]
        mov     edi, [ebx+STREAM.out_wp]

        stdcall [ebx+STREAM.resample], edi, esi, \
                [ebx+STREAM.r_dt], ecx, eax

        mov     ebx, [str]

        add     [ebx+STREAM.out_count], eax;
        add     [ebx+STREAM.out_wp], eax;

        mov     eax, [ebx+STREAM.in_rp]
        mov     ecx, [r_size]
        add     eax, ecx
        add     [ebx+STREAM.in_free], ecx
        sub     [ebx+STREAM.in_count], ecx

        cmp     eax, [ebx+STREAM.in_top]
        jb      @f

        sub     eax, [ebx+STREAM.in_size]
@@:
        mov     [ebx+STREAM.in_rp], eax

.done:
        mov     eax, [ebx+STREAM.notify_event]
        test    eax, eax
        jz      .exit

        mov     ebx, [ebx+STREAM.notify_id]
        mov     edx, EVENT_WATCHED
        xor     esi, esi
        invoke  RaiseEvent   ;eax, ebx, edx, esi
.exit:
        ret
endp

align 4
proc refill_ring stdcall, str:dword
           locals
             event     rd 6
           endl

        mov     ebx, [str]
        mov     edi, [ebx+STREAM.out_wp]
        cmp     edi, [ebx+STREAM.out_top]
        jb      @F
        sub     edi, 0x10000
        mov     [ebx+STREAM.out_wp], edi
@@:
        mov     ecx, [ebx+STREAM.r_size]
        mov     eax, ecx
        cmp     word [ebx+STREAM.format], PCM_1_16_8
        ja      @F

        shr     eax, 1                  ;two channles
@@:
        test    [ebx+STREAM.format], 1  ;even formats mono
        jz      @F

        shr     eax, 1                  ;eax= samples
@@:
        shl     eax, 15   ;eax*=32768 =r_end

        mov     esi, [ebx+STREAM.in_rp]
        mov     edi, [ebx+STREAM.out_wp]

        stdcall [ebx+STREAM.resample], edi, esi, \
                [ebx+STREAM.r_dt], ecx, eax

        mov     ebx, [str]

        add     [ebx+STREAM.out_count], eax;
        add     [ebx+STREAM.out_wp], eax;

        mov     eax, [ebx+STREAM.in_rp]
        mov     ecx, [ebx+STREAM.r_size]
        add     eax, ecx
        add     [ebx+STREAM.in_free], ecx
        sub     [ebx+STREAM.in_count], ecx

        cmp     eax, [ebx+STREAM.in_top]
        jb      @f

        sub     eax, [ebx+STREAM.in_size]
@@:
        mov     [ebx+STREAM.in_rp], eax

        sub     eax, [ebx+STREAM.in_base]
        sub     eax, 128
        lea     esi, [event]

        mov     dword [esi], RT_INP_EMPTY
        mov     dword [esi+4], 0
        mov     dword [esi+8], ebx
        mov     dword [esi+12], eax

        mov     eax, [ebx+STREAM.notify_event]
        test    eax, eax
        jz      .exit

        mov     ebx, [ebx+STREAM.notify_id]
        xor     edx, edx
        invoke  RaiseEvent   ;eax, ebx, edx, esi
.exit:
        ret
endp

if USE_SSE2_MIXER

align 4
proc mix_all stdcall, dest:dword, list:dword, count:dword

        mov     edi, [dest]
        mov     ebx, 32
.mix:
        mov     edx, [list]
        mov     ecx, [count]

        mov     eax, [edx]

        movdqa  xmm1, [eax]
        movss   xmm2, [edx+4]
        movss   xmm3, [edx+8]

        punpcklwd xmm0, xmm1
        punpckhwd xmm1, xmm1

        shufps  xmm2, xmm3, 0
        shufps  xmm2, xmm2, 0x88

        psrad   xmm0, 16
        psrad   xmm1, 16
        cvtdq2ps xmm0, xmm0
        cvtdq2ps xmm1, xmm1
        mulps   xmm0, xmm2
        mulps   xmm1, xmm2

.mix_loop:
        add     dword [edx], 16
        add     edx, 12
        dec     ecx
        jz      @F

        mov     eax, [edx]

        movdqa  xmm3, [eax]
        movss   xmm4, [edx+4]
        movss   xmm5, [edx+8]

        punpcklwd xmm2, xmm3
        punpckhwd xmm3, xmm3

        shufps  xmm4, xmm5, 0
        shufps  xmm4, xmm4, 0x88

        psrad   xmm2, 16
        psrad   xmm3, 16

        cvtdq2ps xmm2, xmm2
        cvtdq2ps xmm3, xmm3

        mulps   xmm2, xmm4
        mulps   xmm3, xmm4
        addps   xmm0, xmm2
        addps   xmm1, xmm3

        jmp     .mix_loop
@@:
        cvtps2dq xmm0, xmm0
        cvtps2dq xmm1, xmm1
        packssdw xmm0, xmm0
        packssdw xmm1, xmm1
        punpcklqdq xmm0, xmm1
        movntdq [edi], xmm0

        add     edi, 16
        dec     ebx
        jnz     .mix

        ret
endp

; param
;  edi = dest
;  edx = mix_list

align 4
mix_fast:

        mov     ebx, 32
        mov     eax, [edx]

        movss   xmm2, [edx+4]        ; vol Lf
        movss   xmm3, [edx+8]        ; vol Rf
        shufps  xmm2, xmm3, 0        ; Rf Rf Lf Lf
        shufps  xmm2, xmm2, 0x88     ; volume level  Rf Lf Rf Lf
.mix:
        movdqa  xmm1, [eax]          ; R3w L3w  R2w L2w  R1w L1w  R0w L0w
        add     eax, 16
        punpcklwd xmm0, xmm1         ; R1w R1w  L1w L1W  R0w R0w  L0w L0w
        punpckhwd xmm1, xmm1         ; R3w R3w  L3w L3w  R2w R2w  L2w L2w

        psrad   xmm0, 16             ; R1d L1d R0d L0d
        psrad   xmm1, 16             ; R3d L3d R2d L2d

        cvtdq2ps xmm0, xmm0          ; time to use all power
        cvtdq2ps xmm1, xmm1          ; of the dark side

        mulps   xmm0, xmm2           ; R1f' L1f' R0f' L0f'
        mulps   xmm1, xmm2           ; R3f' L3f' R2f' L2f'

        cvtps2dq xmm0, xmm0          ; R1d' L1d' R0d' L0d'
        cvtps2dq xmm1, xmm1          ; R3d' L3d' R2d' L2d'
        packssdw xmm0, xmm0          ; R1w' L1w'  R0w' L0w'  R1w' L1w'  R0w' L0w'
        packssdw xmm1, xmm1          ; R3w' L3w'  R2w' L2w'  R3w' L3w'  R2w' L2w'
        punpcklqdq xmm0, xmm1        ; R3w' L3w'  R2w' L2w'  R1w' L1w'  R0w' L0w'
        movntdq [edi], xmm0

        add     edi, 16
        dec     ebx
        jnz     .mix

        ret

align 4
mix_fast_2_stream:

        mov     ebx, 32
        mov     eax, [edx]

        movss   xmm4, [edx+4]        ; vol Lf
        movss   xmm5, [edx+8]        ; vol Rf
        mov     ecx, [edx+12]

        movss   xmm6, [edx+16]       ; vol Lf
        movss   xmm7, [edx+20]       ; vol Rf

        shufps  xmm4, xmm5, 0        ; Rf Rf Lf Lf
        shufps  xmm4, xmm4, 0x88     ; volume level  Rf Lf Rf Lf

        shufps  xmm6, xmm7, 0        ; Rf Rf Lf Lf
        shufps  xmm6, xmm6, 0x88     ; volume level  Rf Lf Rf Lf

.mix:
        movdqa  xmm1, [eax]          ; R3w L3w  R2w L2w  R1w L1w  R0w L0w
        movdqa  xmm3, [ecx]          ; R3w L3w  R2w L2w  R1w L1w  R0w L0w

        add     eax, 16
        add     ecx, 16

        punpcklwd xmm0, xmm1         ; R1w R1w  L1w L1W  R0w R0w  L0w L0w
        punpckhwd xmm1, xmm1         ; R3w R3w  L3w L3w  R2w R2w  L2w L2w

        psrad   xmm0, 16             ; R1d L1d R0d L0d
        psrad   xmm1, 16             ; R3d L3d R2d L2d

        cvtdq2ps xmm0, xmm0          ; time to use all power
        cvtdq2ps xmm1, xmm1          ; of the dark side

        mulps   xmm0, xmm4           ; R1f' L1f' R0f' L0f'
        mulps   xmm1, xmm4           ; R3f' L3f' R2f' L2f'

        punpcklwd xmm2, xmm3         ; R1w R1w  L1w L1W  R0w R0w  L0w L0w
        punpckhwd xmm3, xmm3         ; R3w R3w  L3w L3w  R2w R2w  L2w L2w

        psrad   xmm2, 16             ; R1d L1d R0d L0d
        psrad   xmm3, 16             ; R3d L3d R2d L2d

        cvtdq2ps xmm2, xmm2          ; time to use all power
        cvtdq2ps xmm3, xmm3          ; of the dark side

        mulps   xmm2, xmm6           ; R1f' L1f' R0f' L0f'
        mulps   xmm3, xmm6           ; R3f' L3f' R2f' L2f'

        addps   xmm0, xmm2
        addps   xmm1, xmm3

        cvtps2dq xmm0, xmm0          ; R1d' L1d' R0d' L0d'
        cvtps2dq xmm1, xmm1          ; R3d' L3d' R2d' L2d'
        packssdw xmm0, xmm0          ; R1w' L1w'  R0w' L0w'  R1w' L1w'  R0w' L0w'
        packssdw xmm1, xmm1          ; R3w' L3w'  R2w' L2w'  R3w' L3w'  R2w' L2w'
        punpcklqdq xmm0, xmm1        ; R3w' L3w'  R2w' L2w'  R1w' L1w'  R0w' L0w'
        movntdq [edi], xmm0

        add     edi, 16
        dec     ebx
        jnz     .mix

        ret

else                                    ; fixed point mmx version

align 4
proc mix_all stdcall, dest:dword, list:dword, count:dword

        mov     edi, [dest]
        mov     ebx, 64
.mix:
        mov     edx, [list]
        mov     ecx, [count]

        mov     eax, [edx]

        movq    mm0, [eax]

        movd    mm1, [edx+4]
        punpckldq mm1, mm1
        pmulhw  mm0, mm1
        psllw   mm0, 1

.mix_loop:
        add     dword [edx], 8
        add     edx, 12
        dec     ecx
        jz      @F

        mov     eax, [edx]
        movq    mm1, [eax]
        movd    mm2, [edx+4]
        punpckldq mm2, mm2
        pmulhw  mm1, mm2
        psllw   mm1, 1
        paddsw  mm0, mm1
        jmp     .mix_loop
@@:
        movq    [edi], mm0
        add     edi, 8
        dec     ebx
        jnz     .mix

        ret
endp

end if


align 4
proc resample_1 stdcall, dest:dword,src:dword,\
                       r_dt:dword, r_size:dword,r_end:dword

; dest equ esp+8
; src  equ esp+12
; r_dt equ esp+16
; r_size equ esp+20
; r_end equ esp+24

        mov     edi, [dest]
        mov     edx, [src]
        sub     edx, 32*2
        mov     eax, 16

align 4
.l1:
        mov     ecx, eax
        mov     esi, eax
        and     ecx, 0x7FFF
        shr     esi, 15
        lea     esi, [edx+esi*2]

        movsx   ebp, word [esi]
        movsx   esi, word [esi+2]
        mov     ebx, 32768
        imul    esi, ecx
        sub     ebx, ecx
        imul    ebx, ebp
        lea     ecx, [ebx+esi+16384]
        sar     ecx, 15
        cmp     ecx, 32767        ; 00007fffH
        jle     @f
        mov     ecx, 32767        ; 00007fffH
        jmp     .write
@@:
        cmp     ecx, -32768       ; ffff8000H
        jge     .write
        mov     ecx, -32768       ; ffff8000H
.write:
        mov     ebx, ecx
        shl     ebx, 16
        mov     bx, cx
        mov     [edi], ebx
        add     edi, 4

        add     eax, [esp+16]
        cmp     eax, [esp+24]
        jb      .l1

        mov     ebp, esp

        sub     edi, [dest]
        mov     eax, edi
        ret
endp

align 4
proc resample_18 stdcall, dest:dword,src:dword,\
                       r_dt:dword, r_size:dword,r_end:dword


        mov     edi, [dest]
        mov     edx, [src]
        sub     edx, 32

        mov     esi, 16

align 4
.l1:
        mov     ecx, esi
        mov     eax, esi
        and     ecx, 0x7FFF
        shr     eax, 15
        lea     eax, [edx+eax]

        mov     bx, word [eax]
        sub     bh, 0x80
        sub     bl, 0x80
        movsx   eax, bh
        shl     eax, 8
        movsx   ebp, bl
        shl     ebp, 8
        mov     ebx, 32768
        imul    eax, ecx
        sub     ebx, ecx
        imul    ebx, ebp
        lea     ecx, [ebx+eax+16384]
        sar     ecx, 15
        cmp     ecx, 32767        ; 00007fffH
        jle     @f
        mov     ecx, 32767        ; 00007fffH
        jmp     .write
@@:
        cmp     ecx, -32768       ; ffff8000H
        jge     .write
        mov     ecx, -32768       ; ffff8000H
.write:
        mov     ebx, ecx
        shl     ebx, 16
        mov     bx, cx
        mov     [edi], ebx
        add     edi, 4

        add     esi, [esp+16]
        cmp     esi, [esp+24]
        jb      .l1

        mov     ebp, esp
        sub     edi, [dest]
        mov     eax, edi
        ret
endp

align 4
proc copy_stream stdcall, dest:dword,src:dword,\
                       r_dt:dword, r_size:dword,r_end:dword

        mov     ecx, [r_size]
        mov     eax, ecx
        shr     ecx, 2
        mov     esi, [src]
        mov     edi, [dest]
        cld
        rep movsd
        ret
endp

align 4
proc resample_2 stdcall, dest:dword,src:dword,\
                       r_dt:dword, r_size:dword,r_end:dword

        mov     edx, [src]
        sub     edx, 32*4
        mov     edi, [dest]
        mov     ebx, [r_dt]
        mov     eax, 16
        emms

align 4
.l1:
        mov     ecx, eax
        mov     esi, eax
        and     ecx, 0x7FFF
        shr     esi, 15
        lea     esi, [edx+esi*4]

        movq    mm0, [esi]
        movq    mm1, mm0

        movd    mm2, ecx
        punpcklwd mm2, mm2
        movq    mm3, qword [m7]    ;0x8000

        psubw   mm3, mm2  ;        ;0x8000 - iconst
        punpckldq mm3, mm2

        pmulhw  mm0, mm3
        pmullw  mm1, mm3

        movq    mm4, mm1
        punpcklwd mm1, mm0
        punpckhwd mm4, mm0
        paddd   mm1, mm4
        psrad   mm1, 15
        packssdw mm1, mm1
        movd    [edi], mm1
        add     edi, 4

        add     eax, ebx
        cmp     eax, [r_end]
        jb      .l1
        emms

        sub     edi, [dest]
        mov     eax, edi
        ret
endp

align 4
proc resample_28 stdcall, dest:dword,src:dword,\
                       r_dt:dword, r_size:dword,r_end:dword

        mov     edx, [src]
        sub     edx, 32*2
        mov     edi, [dest]
        mov     ebx, [r_dt]
        mov     eax, 16
        emms
        movq    mm7, [mm80]
        movq    mm6, [mm_mask]

align 4
.l1:
        mov     ecx, eax
        mov     esi, eax
        and     ecx, 0x7FFF
        shr     esi, 15
        lea     esi, [edx+esi*2]

        movq    mm0, [esi]
        psubb   mm0, mm7
        punpcklbw mm0, mm0
        pand    mm0, mm6

        movq    mm1, mm0

        movd    mm2, ecx
        punpcklwd mm2, mm2
        movq    mm3, qword [m7] ;                  // 0x8000

        psubw   mm3, mm2        ;         // 0x8000 - iconst
        punpckldq mm3, mm2

        pmulhw  mm0, mm3
        pmullw  mm1, mm3

        movq    mm4, mm1
        punpcklwd mm1, mm0
        punpckhwd mm4, mm0
        paddd   mm1, mm4
        psrad   mm1, 15
        packssdw mm1, mm1
        movd    [edi], mm1
        add     edi, 4

        add     eax, ebx
        cmp     eax, [r_end]
        jb      .l1
        emms


        sub     edi, [dest]
        mov     eax, edi
        ret
endp


proc m16_stereo stdcall, dest:dword,src:dword,\
                       r_dt:dword, r_size:dword,r_end:dword

        mov     esi, [src]
        mov     edi, [dest]
        mov     ecx, [r_size]
        shr     ecx, 8
@@:
        call    m16_s_mmx
        add     edi, 128
        add     esi, 64
        call    m16_s_mmx
        add     edi, 128
        add     esi, 64
        call    m16_s_mmx
        add     edi, 128
        add     esi, 64
        call    m16_s_mmx
        add     edi, 128
        add     esi, 64
        dec     ecx
        jnz     @b

        mov     eax, [r_size]
        add     eax, eax
        ret
endp

align 4
proc s8_stereo stdcall, dest:dword,src:dword,\
                       r_dt:dword, r_size:dword,r_end:dword

        mov     esi, [src]
        mov     edi, [dest]
        mov     ecx, [r_size]
        shr     ecx, 7

        movq    mm7, [mm80]
        movq    mm6, [mm_mask]
@@:
        call    s8_s_mmx
        add     edi, 64
        add     esi, 32
        call    s8_s_mmx
        add     edi, 64
        add     esi, 32
        call    s8_s_mmx
        add     edi, 64
        add     esi, 32
        call    s8_s_mmx
        add     edi, 64
        add     esi, 32
        dec     ecx
        jnz     @b

        mov     eax, [r_size]
        add     eax, eax
        ret
endp

proc m8_stereo stdcall, dest:dword,src:dword,\
                       r_dt:dword, r_size:dword,r_end:dword

        mov     esi, [src]
        mov     edi, [dest]
        mov     ecx, [r_size]
        shr     ecx, 6

        movq    mm7, [mm80]
        movq    mm6, [mm_mask]
@@:
        call    m8_s_mmx
        add     edi, 64
        add     esi, 16
        call    m8_s_mmx
        add     edi, 64
        add     esi, 16
        call    m8_s_mmx
        add     edi, 64
        add     esi, 16
        call    m8_s_mmx
        add     edi, 64
        add     esi, 16
        dec     ecx
        jnz     @b

        mov     eax, [r_size]
        add     eax, eax
        add     eax, eax
        ret
endp

align 4
proc alloc_mix_buff

        bsf     eax, [mix_buff_map]
        jnz     .find
        xor     eax, eax
        ret
.find:
        btr     [mix_buff_map], eax
        shl     eax, 9
        add     eax, [mix_buff]
        ret
endp

align 4
proc m16_s_mmx

        movq    mm0, [esi]
        movq    mm1, mm0
        punpcklwd mm0, mm0
        punpckhwd mm1, mm1
        movq    [edi], mm0
        movq    [edi+8], mm1

        movq    mm0, [esi+8]
        movq    mm1, mm0
        punpcklwd mm0, mm0
        punpckhwd mm1, mm1
        movq    [edi+16], mm0
        movq    [edi+24], mm1

        movq    mm0, [esi+16]
        movq    mm1, mm0
        punpcklwd mm0, mm0
        punpckhwd mm1, mm1
        movq    [edi+32], mm0
        movq    [edi+40], mm1

        movq    mm0, [esi+24]
        movq    mm1, mm0
        punpcklwd mm0, mm0
        punpckhwd mm1, mm1
        movq    [edi+48], mm0
        movq    [edi+56], mm1

        movq    mm0, [esi+32]
        movq    mm1, mm0
        punpcklwd mm0, mm0
        punpckhwd mm1, mm1
        movq    [edi+64], mm0
        movq    [edi+72], mm1

        movq    mm0, [esi+40]
        movq    mm1, mm0
        punpcklwd mm0, mm0
        punpckhwd mm1, mm1
        movq    [edi+80], mm0
        movq    [edi+88], mm1


        movq    mm0, [esi+48]
        movq    mm1, mm0
        punpcklwd mm0, mm0
        punpckhwd mm1, mm1
        movq    [edi+96], mm0
        movq    [edi+104], mm1

        movq    mm0, [esi+56]
        movq    mm1, mm0
        punpcklwd mm0, mm0
        punpckhwd mm1, mm1
        movq    [edi+112], mm0
        movq    [edi+120], mm1

        ret
endp

align 4
proc s8_s_mmx

        movq    mm0, [esi]
        psubb   mm0, mm7
        movq    mm1, mm0
        punpcklbw mm0, mm0
        pand    mm0, mm6
        punpckhbw mm1, mm1
        pand    mm1, mm6
        movq    [edi], mm0
        movq    [edi+8], mm1

        movq    mm0, [esi+8]
        psubb   mm0, mm7
        movq    mm1, mm0
        punpcklbw mm0, mm0
        pand    mm0, mm6
        punpckhbw mm1, mm1
        pand    mm1, mm6
        movq    [edi+16], mm0
        movq    [edi+24], mm1

        movq    mm0, [esi+16]
        psubb   mm0, mm7
        movq    mm1, mm0
        punpcklbw mm0, mm0
        pand    mm0, mm6
        punpckhbw mm1, mm1
        pand    mm1, mm6
        movq    [edi+32], mm0
        movq    [edi+40], mm1

        movq    mm0, [esi+24]
        psubb   mm0, mm7
        movq    mm1, mm0
        punpcklbw mm0, mm0
        pand    mm0, mm6
        punpckhbw mm1, mm1
        pand    mm1, mm6
        movq    [edi+48], mm0
        movq    [edi+56], mm1

        ret

endp

align 4
proc m8_s_mmx

        movq    mm0, [esi]
        psubb   mm0, mm7
        movq    mm1, mm0
        punpcklbw mm0, mm0
        pand    mm0, mm6
        punpckhbw mm1, mm1
        pand    mm1, mm6
        movq    mm2, mm0
        punpcklwd mm0, mm0
        punpckhwd mm2, mm2

        movq    mm3, mm1
        punpcklwd mm1, mm1
        punpckhwd mm3, mm3

        movq    [edi], mm0
        movq    [edi+8], mm2
        movq    [edi+16], mm1
        movq    [edi+24], mm3

        movq    mm0, [esi+8]
        psubb   mm0, mm7
        movq    mm1, mm0
        punpcklbw mm0, mm0
        pand    mm0, mm6
        punpckhbw mm1, mm1
        pand    mm1, mm6
        movq    mm2, mm0
        punpcklwd mm0, mm0
        punpckhwd mm2, mm2

        movq    mm3, mm1
        punpcklwd mm1, mm1
        punpckhwd mm3, mm3

        movq    [edi+32], mm0
        movq    [edi+40], mm2
        movq    [edi+48], mm1
        movq    [edi+56], mm3

        ret
endp

align 4
proc mix_2_1 stdcall, output:dword, str0:dword, str1:dword

        mov     edi, [output]
        mov     eax, [str0]
        mov     ebx, [str1]
        mov     esi, 128
        call    [mix_2_core]   ;edi, eax, ebx

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        call    [mix_2_core]   ;edi, eax, ebx

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        call    [mix_2_core]   ;edi, eax, ebx

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        call    [mix_2_core]   ;edi, eax, ebx
        ret
endp

align 4
proc mix_3_1 stdcall, output:dword, str0:dword, str1:dword, str2:dword

        mov     edi, [output]
        mov     eax, [str0]
        mov     ebx, [str1]
        mov     ecx, [str2]
        mov     esi, 128
        call    [mix_3_core]

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        add     ecx, esi
        call    [mix_3_core]

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        add     ecx, esi
        call    [mix_3_core]

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        add     ecx, esi
        call    [mix_3_core]
        ret
endp

align 4
proc mix_4_1 stdcall, str0:dword, str1:dword,\
                      str2:dword, str3:dword

           local output:DWORD

        call    alloc_mix_buff
        and     eax, eax
        jz      .err

        mov     [output], eax

        mov     edi, eax
        mov     eax, [str0]
        mov     ebx, [str1]
        mov     ecx, [str2]
        mov     edx, [str3]
        mov     esi, 128
        call    [mix_4_core]  ;edi, eax, ebx, ecx, edx

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        add     ecx, esi
        add     edx, esi
        call    [mix_4_core]  ;edi, eax, ebx, ecx, edx

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        add     ecx, esi
        add     edx, esi
        call    [mix_4_core]  ;edi, eax, ebx, ecx, edx

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        add     ecx, esi
        add     edx, esi
        call    [mix_4_core]  ;edi, eax, ebx, ecx, edx
        mov     eax, [output]
        ret
.err:
        xor     eax, eax
        ret
endp


align 4
proc final_mix stdcall, output:dword, str0:dword, str1:dword,\
                        str2:dword, str3:dword

        mov     edi, [output]

        mov     eax, [str0]
        mov     ebx, [str1]
        mov     ecx, [str2]
        mov     edx, [str3]
        mov     esi, 128
        call    [mix_4_core]  ;edi, eax, ebx, ecx, edx

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        add     ecx, esi
        add     edx, esi
        call    [mix_4_core]  ;edi, eax, ebx, ecx, edx

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        add     ecx, esi
        add     edx, esi
        call    [mix_4_core]  ;edi, eax, ebx, ecx, edx

        add     edi, esi
        add     eax, esi
        add     ebx, esi
        add     ecx, esi
        add     edx, esi
        call    [mix_4_core]  ;edi, eax, ebx, ecx, edx
        ret
endp

align 4
proc copy_mem stdcall, output:dword, input:dword

        mov     edi, [output]
        mov     esi, [input]
        mov     ecx, 0x80
.l1:
        mov     eax, [esi]
        mov     [edi], eax
        add     esi, 4
        add     edi, 4
        loop    .l1

        ret
endp

proc memcpy
@@:
        mov     eax, [esi]
        mov     [edi], eax
        add     esi, 4
        add     edi, 4
        dec     ecx
        jnz     @B
        ret
endp

if 0

align 4
proc new_mix stdcall, output:dword
           locals
             mixCounter  dd ?
             mixIndex  dd ?
             streamIndex dd ?
             inputCount  dd ?
             main_count  dd ?
             blockCount  dd ?
             mix_out  dd ?
           endl

        call    prepare_playlist

        cmp     [play_count], 0
        je      .exit
        call    FpuSave
        mov     [main_count], 32;
.l00:
        mov     [mix_buff_map], 0x0000FFFF;
        xor     eax, eax
        mov     [mixCounter], eax
        mov     [mixIndex], eax
        mov     [streamIndex], eax;
        mov     ebx, [play_count]
        mov     [inputCount], ebx
.l0:
        mov     ecx, 4
.l1:
        mov     ebx, [streamIndex]
        mov     esi, [play_list+ebx*4]
        mov     eax, [esi+STREAM.work_read]
        add     [esi+STREAM.work_read], 512

        mov     ebx, [mixIndex]
        mov     [mix_input+ebx*4], eax
        inc     [mixCounter]
        inc     [mixIndex]
        inc     [streamIndex]
        dec     [inputCount]
        jz      .m2

        dec     ecx
        jnz     .l1

        cmp     [mixCounter], 4
        jnz     .m2

        stdcall mix_4_1, [mix_input], [mix_input+4], [mix_input+8], [mix_input+12]
        sub     [mixIndex], 4
        mov     ebx, [mixIndex]
        mov     [mix_input+ebx*4], eax
        inc     [mixIndex]
        mov     [mixCounter], 0

        cmp     [inputCount], 0
        jnz     .l0
.m2:
        cmp     [mixIndex], 1
        jne     @f
        stdcall copy_mem, [output], [mix_input]
        jmp     .m3
@@:
        cmp     [mixIndex], 2
        jne     @f
        stdcall mix_2_1, [output], [mix_input], [mix_input+4]
        jmp     .m3
@@:
        cmp     [mixIndex], 3
        jne     @f
        stdcall mix_3_1, [output], [mix_input], [mix_input+4], [mix_input+8]
        jmp     .m3
@@:
        stdcall final_mix, [output], [mix_input], [mix_input+4], [mix_input+8], [mix_input+12]
.m3:
        add     [output], 512

        dec     [main_count]
        jnz     .l00

        call    update_stream
        emms
        call    FpuRestore
        ret
.exit:
        mov     edi, [output]
        mov     ecx, 0x1000
        xor     eax, eax
        cld
        rep stosd
        ret
endp

end if