yading@11: ;******************************************************************************
yading@11: ;* x86 optimized dithering format conversion
yading@11: ;* Copyright (c) 2012 Justin Ruggles <justin.ruggles@gmail.com>
yading@11: ;*
yading@11: ;* This file is part of FFmpeg.
yading@11: ;*
yading@11: ;* FFmpeg is free software; you can redistribute it and/or
yading@11: ;* modify it under the terms of the GNU Lesser General Public
yading@11: ;* License as published by the Free Software Foundation; either
yading@11: ;* version 2.1 of the License, or (at your option) any later version.
yading@11: ;*
yading@11: ;* FFmpeg is distributed in the hope that it will be useful,
yading@11: ;* but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@11: ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@11: ;* Lesser General Public License for more details.
yading@11: ;*
yading@11: ;* You should have received a copy of the GNU Lesser General Public
yading@11: ;* License along with FFmpeg; if not, write to the Free Software
yading@11: ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@11: ;******************************************************************************
yading@11: 
yading@11: %include "libavutil/x86/x86util.asm"
yading@11: 
yading@11: SECTION_RODATA 32
yading@11: 
yading@11: ; 1.0f / (2.0f * INT32_MAX)
yading@11: pf_dither_scale: times 8 dd 2.32830643762e-10
yading@11: 
yading@11: pf_s16_scale: times 4 dd 32753.0
yading@11: 
yading@11: SECTION_TEXT
yading@11: 
yading@11: ;------------------------------------------------------------------------------
yading@11: ; void ff_quantize(int16_t *dst, float *src, float *dither, int len);
yading@11: ;------------------------------------------------------------------------------
yading@11: 
yading@11: INIT_XMM sse2
yading@11: cglobal quantize, 4,4,3, dst, src, dither, len
yading@11:     lea         lenq, [2*lend]
yading@11:     add         dstq, lenq
yading@11:     lea         srcq, [srcq+2*lenq]
yading@11:     lea      ditherq, [ditherq+2*lenq]
yading@11:     neg         lenq
yading@11:     mova          m2, [pf_s16_scale]
yading@11: .loop:
yading@11:     mulps         m0, m2, [srcq+2*lenq]
yading@11:     mulps         m1, m2, [srcq+2*lenq+mmsize]
yading@11:     addps         m0, [ditherq+2*lenq]
yading@11:     addps         m1, [ditherq+2*lenq+mmsize]
yading@11:     cvtps2dq      m0, m0
yading@11:     cvtps2dq      m1, m1
yading@11:     packssdw      m0, m1
yading@11:     mova     [dstq+lenq], m0
yading@11:     add         lenq, mmsize
yading@11:     jl .loop
yading@11:     REP_RET
yading@11: 
yading@11: ;------------------------------------------------------------------------------
yading@11: ; void ff_dither_int_to_float_rectangular(float *dst, int *src, int len)
yading@11: ;------------------------------------------------------------------------------
yading@11: 
yading@11: %macro DITHER_INT_TO_FLOAT_RECTANGULAR 0
yading@11: cglobal dither_int_to_float_rectangular, 3,3,3, dst, src, len
yading@11:     lea         lenq, [4*lend]
yading@11:     add         srcq, lenq
yading@11:     add         dstq, lenq
yading@11:     neg         lenq
yading@11:     mova          m0, [pf_dither_scale]
yading@11: .loop:
yading@11:     cvtdq2ps      m1, [srcq+lenq]
yading@11:     cvtdq2ps      m2, [srcq+lenq+mmsize]
yading@11:     mulps         m1, m1, m0
yading@11:     mulps         m2, m2, m0
yading@11:     mova  [dstq+lenq], m1
yading@11:     mova  [dstq+lenq+mmsize], m2
yading@11:     add         lenq, 2*mmsize
yading@11:     jl .loop
yading@11:     REP_RET
yading@11: %endmacro
yading@11: 
yading@11: INIT_XMM sse2
yading@11: DITHER_INT_TO_FLOAT_RECTANGULAR
yading@11: INIT_YMM avx
yading@11: DITHER_INT_TO_FLOAT_RECTANGULAR
yading@11: 
yading@11: ;------------------------------------------------------------------------------
yading@11: ; void ff_dither_int_to_float_triangular(float *dst, int *src0, int len)
yading@11: ;------------------------------------------------------------------------------
yading@11: 
yading@11: %macro DITHER_INT_TO_FLOAT_TRIANGULAR 0
yading@11: cglobal dither_int_to_float_triangular, 3,4,5, dst, src0, len, src1
yading@11:     lea         lenq, [4*lend]
yading@11:     lea        src1q, [src0q+2*lenq]
yading@11:     add        src0q, lenq
yading@11:     add         dstq, lenq
yading@11:     neg         lenq
yading@11:     mova          m0, [pf_dither_scale]
yading@11: .loop:
yading@11:     cvtdq2ps      m1, [src0q+lenq]
yading@11:     cvtdq2ps      m2, [src0q+lenq+mmsize]
yading@11:     cvtdq2ps      m3, [src1q+lenq]
yading@11:     cvtdq2ps      m4, [src1q+lenq+mmsize]
yading@11:     addps         m1, m1, m3
yading@11:     addps         m2, m2, m4
yading@11:     mulps         m1, m1, m0
yading@11:     mulps         m2, m2, m0
yading@11:     mova  [dstq+lenq], m1
yading@11:     mova  [dstq+lenq+mmsize], m2
yading@11:     add         lenq, 2*mmsize
yading@11:     jl .loop
yading@11:     REP_RET
yading@11: %endmacro
yading@11: 
yading@11: INIT_XMM sse2
yading@11: DITHER_INT_TO_FLOAT_TRIANGULAR
yading@11: INIT_YMM avx
yading@11: DITHER_INT_TO_FLOAT_TRIANGULAR