predict-a.asm

DM642 H.264 codec DM642 H.264 codec DM642 H.264 codec DM642 H.264 codec

;*****************************************************************************
;* predict-a.asm: h264 encoder library
;*****************************************************************************
;* Copyright (C) 2005 x264 project
;*
;* Authors: Loren Merritt <[e-mail]>
;*
;* This program is free software; you can redistribute it and/or modify
;* it under the terms of the GNU General Public License as published by
;* the Free Software Foundation; either version 2 of the License, or
;* (at your option) any later version.
;*
;* This program is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
;* GNU General Public License for more details.
;*
;* You should have received a copy of the GNU General Public License
;* along with this program; if not, write to the Free Software
;* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
;*****************************************************************************

BITS 64

;=============================================================================
; Macros and other preprocessor constants
;=============================================================================

%include "amd64inc.asm"

%macro STORE8x8 2
    movq        [parm1q + 0*FDEC_STRIDE], %1
    movq        [parm1q + 1*FDEC_STRIDE], %1
    movq        [parm1q + 2*FDEC_STRIDE], %1
    movq        [parm1q + 3*FDEC_STRIDE], %1
    movq        [parm1q + 4*FDEC_STRIDE], %2
    movq        [parm1q + 5*FDEC_STRIDE], %2
    movq        [parm1q + 6*FDEC_STRIDE], %2
    movq        [parm1q + 7*FDEC_STRIDE], %2
%endmacro

%macro STORE16x16 2
    mov         eax, 4
ALIGN 4
.loop:
    movq        [parm1q + 1*FDEC_STRIDE], %1
    movq        [parm1q + 2*FDEC_STRIDE], %1
    movq        [parm1q + 3*FDEC_STRIDE], %1
    movq        [parm1q + 4*FDEC_STRIDE], %1
    movq        [parm1q + 1*FDEC_STRIDE + 8], %2
    movq        [parm1q + 2*FDEC_STRIDE + 8], %2
    movq        [parm1q + 3*FDEC_STRIDE + 8], %2
    movq        [parm1q + 4*FDEC_STRIDE + 8], %2
    dec         eax
    lea         parm1q, [parm1q + 4*FDEC_STRIDE]
    jnz         .loop
    nop
%endmacro


SECTION .rodata align=16

ALIGN 16
pw_2: times 4 dw 2
pw_4: times 4 dw 4
pw_8: times 4 dw 8
pw_3210:
    dw 0
    dw 1
    dw 2
    dw 3
ALIGN 16
pb_1: times 16 db 1
pb_00s_ff:
    times 8 db 0
pb_0s_ff:
    times 7 db 0
    db 0xff

;=============================================================================
; Code
;=============================================================================

SECTION .text

cglobal predict_4x4_ddl_mmxext
cglobal predict_4x4_vl_mmxext
cglobal predict_8x8_v_mmxext
cglobal predict_8x8_dc_mmxext
cglobal predict_8x8_dc_top_mmxext
cglobal predict_8x8_dc_left_mmxext
cglobal predict_8x8_ddl_mmxext
cglobal predict_8x8_ddl_sse2
cglobal predict_8x8_ddr_sse2
cglobal predict_8x8_vl_sse2
cglobal predict_8x8_vr_core_mmxext
cglobal predict_8x8c_v_mmx
cglobal predict_8x8c_dc_core_mmxext
cglobal predict_8x8c_p_core_mmxext
cglobal predict_16x16_p_core_mmxext
cglobal predict_16x16_v_mmx
cglobal predict_16x16_dc_core_mmxext
cglobal predict_16x16_dc_top_mmxext


; dest, left, right, src, tmp
; output: %1 = (t[n-1] + t[n]*2 + t[n+1] + 2) >> 2
%macro PRED8x8_LOWPASS0 6
    mov%6       %5, %2
    pavgb       %2, %3
    pxor        %3, %5
    mov%6       %1, %4
    pand        %3, [pb_1 GLOBAL]
    psubusb     %2, %3
    pavgb       %1, %2
%endmacro
%macro PRED8x8_LOWPASS 5
    PRED8x8_LOWPASS0 %1, %2, %3, %4, %5, q
%endmacro
%macro PRED8x8_LOWPASS_XMM 5
    PRED8x8_LOWPASS0 %1, %2, %3, %4, %5, dqa
%endmacro


;-----------------------------------------------------------------------------
; void predict_4x4_ddl_mmxext( uint8_t *src )
;-----------------------------------------------------------------------------

ALIGN 16
predict_4x4_ddl_mmxext:
    sub         parm1q, FDEC_STRIDE
    movq        mm3, [parm1q]
    movq        mm1, [parm1q-1]
    movq        mm2, mm3
    movq        mm4, [pb_0s_ff GLOBAL]
    psrlq       mm2, 8
    pand        mm4, mm3
    por         mm2, mm4

    PRED8x8_LOWPASS mm0, mm1, mm2, mm3, mm5

%assign Y 1
%rep 4
    psrlq       mm0, 8
    movd        [parm1q+Y*FDEC_STRIDE], mm0
%assign Y (Y+1)
%endrep

    ret

;-----------------------------------------------------------------------------
; void predict_4x4_vl_mmxext( uint8_t *src )
;-----------------------------------------------------------------------------

ALIGN 16
predict_4x4_vl_mmxext:
    movq        mm1, [parm1q-FDEC_STRIDE]
    movq        mm3, mm1
    movq        mm2, mm1
    psrlq       mm3, 8
    psrlq       mm2, 16
    movq        mm4, mm3
    pavgb       mm4, mm1

    PRED8x8_LOWPASS mm0, mm1, mm2, mm3, mm5

    movd        [parm1q+0*FDEC_STRIDE], mm4
    movd        [parm1q+1*FDEC_STRIDE], mm0
    psrlq       mm4, 8
    psrlq       mm0, 8
    movd        [parm1q+2*FDEC_STRIDE], mm4
    movd        [parm1q+3*FDEC_STRIDE], mm0

    ret

;-----------------------------------------------------------------------------
; void predict_8x8_v_mmxext( uint8_t *src, uint8_t *edge )
;-----------------------------------------------------------------------------

ALIGN 16
predict_8x8_v_mmxext:
    movq        mm0, [parm2q+16]
    STORE8x8    mm0, mm0
    ret

;-----------------------------------------------------------------------------
; void predict_8x8_dc_mmxext( uint8_t *src, uint8_t *edge );
;-----------------------------------------------------------------------------

ALIGN 16
predict_8x8_dc_mmxext:
    pxor        mm0, mm0
    pxor        mm1, mm1
    psadbw      mm0, [parm2q+7]
    psadbw      mm1, [parm2q+16]
    paddw       mm0, [pw_8 GLOBAL]
    paddw       mm0, mm1
    psrlw       mm0, 4
    pshufw      mm0, mm0, 0
    packuswb    mm0, mm0
    STORE8x8    mm0, mm0
    ret

;-----------------------------------------------------------------------------
; void predict_8x8_dc_top_mmxext( uint8_t *src, uint8_t *edge );
;-----------------------------------------------------------------------------

ALIGN 16
predict_8x8_dc_top_mmxext:
    pxor        mm0, mm0
    psadbw      mm0, [parm2q+16]
    paddw       mm0, [pw_4 GLOBAL]
    psrlw       mm0, 3
    pshufw      mm0, mm0, 0
    packuswb    mm0, mm0
    STORE8x8    mm0, mm0
    ret

;-----------------------------------------------------------------------------
; void predict_8x8_dc_left_mmxext( uint8_t *src, uint8_t *edge );
;-----------------------------------------------------------------------------

ALIGN 16
predict_8x8_dc_left_mmxext:
    pxor        mm0, mm0
    psadbw      mm0, [parm2q+7]
    paddw       mm0, [pw_4 GLOBAL]
    psrlw       mm0, 3
    pshufw      mm0, mm0, 0
    packuswb    mm0, mm0
    STORE8x8    mm0, mm0
    ret

;-----------------------------------------------------------------------------
; void predict_8x8_ddl_mmxext( uint8_t *src, uint8_t *edge )
;-----------------------------------------------------------------------------

ALIGN 16
predict_8x8_ddl_mmxext:
    movq        mm5, [parm2q+16]
    movq        mm2, [parm2q+17]
    movq        mm3, [parm2q+23]
    movq        mm4, [parm2q+25]
    movq        mm1, mm5
    psllq       mm1, 8
    PRED8x8_LOWPASS mm0, mm1, mm2, mm5, mm7
    PRED8x8_LOWPASS mm1, mm3, mm4, [parm2q+24], mm6

%assign Y 7
%rep 6
    movq        [parm1q+Y*FDEC_STRIDE], mm1
    movq        mm2, mm0
    psllq       mm1, 8
    psrlq       mm2, 56
    psllq       mm0, 8
    por         mm1, mm2
%assign Y (Y-1)
%endrep
    movq        [parm1q+Y*FDEC_STRIDE], mm1
    psllq       mm1, 8
    psrlq       mm0, 56
    por         mm1, mm0
%assign Y (Y-1)
    movq        [parm1q+Y*FDEC_STRIDE], mm1

    ret

;-----------------------------------------------------------------------------
; void predict_8x8_ddl_sse2( uint8_t *src, uint8_t *edge )
;-----------------------------------------------------------------------------

ALIGN 16
predict_8x8_ddl_sse2:
    movdqa      xmm3, [parm2q+16]
    movdqu      xmm2, [parm2q+17]
    movdqa      xmm1, xmm3
    pslldq      xmm1, 1
    PRED8x8_LOWPASS_XMM xmm0, xmm1, xmm2, xmm3, xmm4