1 ;******************************************************************************
2 ;* x86 optimized Format Conversion Utils
3 ;* Copyright (c) 2008 Loren Merritt
5 ;* This file is part of FFmpeg.
7 ;* FFmpeg is free software; you can redistribute it and/or
8 ;* modify it under the terms of the GNU Lesser General Public
9 ;* License as published by the Free Software Foundation; either
10 ;* version 2.1 of the License, or (at your option) any later version.
12 ;* FFmpeg is distributed in the hope that it will be useful,
13 ;* but WITHOUT ANY WARRANTY; without even the implied warranty of
14 ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 ;* Lesser General Public License for more details.
17 ;* You should have received a copy of the GNU Lesser General Public
18 ;* License along with FFmpeg; if not, write to the Free Software
19 ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 ;******************************************************************************
23 %include "x86util.asm"
36 %macro FLOAT_TO_INT16_INTERLEAVE6 1
37 ; void float_to_int16_interleave6_sse(int16_t *dst, const float **src, int len)
38 cglobal float_to_int16_interleave6_%1, 2,7,0, dst, src, src1, src2, src3, src4, src5
43 %define lend dword r2m
45 mov src1q, [srcq+1*gprsize]
46 mov src2q, [srcq+2*gprsize]
47 mov src3q, [srcq+3*gprsize]
48 mov src4q, [srcq+4*gprsize]
49 mov src5q, [srcq+5*gprsize]
58 cvtps2pi mm1, [srcq+src1q]
59 cvtps2pi mm2, [srcq+src2q]
60 cvtps2pi mm3, [srcq+src3q]
61 cvtps2pi mm4, [srcq+src4q]
62 cvtps2pi mm5, [srcq+src5q]
83 %endmacro ; FLOAT_TO_INT16_INTERLEAVE6
85 %define pswapd PSWAPD_SSE
86 FLOAT_TO_INT16_INTERLEAVE6 sse
87 %define cvtps2pi pf2id
88 %define pswapd PSWAPD_3DN1
89 FLOAT_TO_INT16_INTERLEAVE6 3dnow
91 FLOAT_TO_INT16_INTERLEAVE6 3dn2
94 ;-----------------------------------------------------------------------------
95 ; void ff_float_interleave6(float *dst, const float **src, unsigned int len);
96 ;-----------------------------------------------------------------------------
98 %macro FLOAT_INTERLEAVE6 2
99 cglobal float_interleave6_%1, 2,7,%2, dst, src, src1, src2, src3, src4, src5
104 %define lend dword r2m
106 mov src1q, [srcq+1*gprsize]
107 mov src2q, [srcq+2*gprsize]
108 mov src3q, [srcq+3*gprsize]
109 mov src4q, [srcq+4*gprsize]
110 mov src5q, [srcq+5*gprsize]
120 movaps m1, [srcq+src1q]
121 movaps m2, [srcq+src2q]
122 movaps m3, [srcq+src3q]
123 movaps m4, [srcq+src4q]
124 movaps m5, [srcq+src5q]
147 movq m1, [srcq+src1q]
148 movq m2, [srcq+src2q]
149 movq m3, [srcq+src3q]
150 movq m4, [srcq+src4q]
151 movq m5, [srcq+src5q]
153 SBUTTERFLY dq, 0, 1, 6
154 SBUTTERFLY dq, 2, 3, 6
155 SBUTTERFLY dq, 4, 5, 6
174 FLOAT_INTERLEAVE6 mmx, 0
176 FLOAT_INTERLEAVE6 sse, 7
178 ;-----------------------------------------------------------------------------
179 ; void ff_float_interleave2(float *dst, const float **src, unsigned int len);
180 ;-----------------------------------------------------------------------------
182 %macro FLOAT_INTERLEAVE2 2
183 cglobal float_interleave2_%1, 3,4,%2, dst, src, len, src1
184 mov src1q, [srcq+gprsize]
189 MOVPS m1, [srcq+src1q ]
190 MOVPS m3, [srcq +mmsize]
191 MOVPS m4, [srcq+src1q+mmsize]
202 MOVPS [dstq+1*mmsize], m2
203 MOVPS [dstq+2*mmsize], m3
204 MOVPS [dstq+3*mmsize], m1
218 %define PUNPCKLDQ punpckldq
219 %define PUNPCKHDQ punpckhdq
220 FLOAT_INTERLEAVE2 mmx, 0
223 %define PUNPCKLDQ unpcklps
224 %define PUNPCKHDQ unpckhps
225 FLOAT_INTERLEAVE2 sse, 5