Subversion Repositories Kolibri OS

Rev

Details | Last modification | View Log | RSS feed

Rev Author Line No. Line
6148 serge 1 
;*****************************************************************************
2 
;* MMX/SSE2/AVX-optimized 10-bit H.264 weighted prediction code
3 
;*****************************************************************************
4 
;* Copyright (C) 2005-2011 x264 project
5 
;*
6 
;* Authors: Daniel Kang
7 
;*
8 
;* This file is part of FFmpeg.
9 
;*
10 
;* FFmpeg is free software; you can redistribute it and/or
11 
;* modify it under the terms of the GNU Lesser General Public
12 
;* License as published by the Free Software Foundation; either
13 
;* version 2.1 of the License, or (at your option) any later version.
14 
;*
15 
;* FFmpeg is distributed in the hope that it will be useful,
16 
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
17 
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
18 
;* Lesser General Public License for more details.
19 
;*
20 
;* You should have received a copy of the GNU Lesser General Public
21 
;* License along with FFmpeg; if not, write to the Free Software
22 
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 
;******************************************************************************
24 
 
25 
%include "libavutil/x86/x86util.asm"
26 
 
27 
SECTION_RODATA 32
28 
 
29 
pw_pixel_max: times 8 dw ((1 << 10)-1)
30 
sq_1: dq 1
31 
      dq 0
32 
 
33 
cextern pw_1
34 
 
35 
SECTION .text
36 
 
37 
;-----------------------------------------------------------------------------
38 
; void h264_weight(uint8_t *dst, int stride, int height, int log2_denom,
39 
;                  int weight, int offset);
40 
;-----------------------------------------------------------------------------
41 
%macro WEIGHT_PROLOGUE 0
42 
.prologue:
43 
    PROLOGUE 0,6,8
44 
    movifnidn  r0, r0mp
45 
    movifnidn r1d, r1m
46 
    movifnidn r2d, r2m
47 
    movifnidn r4d, r4m
48 
    movifnidn r5d, r5m
49 
%endmacro
50 
 
51 
%macro WEIGHT_SETUP 0
52 
    mova       m0, [pw_1]
53 
    movd       m2, r3m
54 
    pslld      m0, m2       ; 1<
55 
    SPLATW     m0, m0
56 
    shl        r5, 19       ; *8, move to upper half of dword
57 
    lea        r5, [r5+r4*2+0x10000]
58 
    movd       m3, r5d      ; weight<<1 | 1+(offset<<(3))
59 
    pshufd     m3, m3, 0
60 
    mova       m4, [pw_pixel_max]
61 
    paddw      m2, [sq_1]   ; log2_denom+1
62 
%if notcpuflag(sse4)
63 
    pxor       m7, m7
64 
%endif
65 
%endmacro
66 
 
67 
%macro WEIGHT_OP 1-2
68 
%if %0==1
69 
    mova        m5, [r0+%1]
70 
    punpckhwd   m6, m5, m0
71 
    punpcklwd   m5, m0
72 
%else
73 
    movq        m5, [r0+%1]
74 
    movq        m6, [r0+%2]
75 
    punpcklwd   m5, m0
76 
    punpcklwd   m6, m0
77 
%endif
78 
    pmaddwd     m5, m3
79 
    pmaddwd     m6, m3
80 
    psrad       m5, m2
81 
    psrad       m6, m2
82 
%if cpuflag(sse4)
83 
    packusdw    m5, m6
84 
    pminsw      m5, m4
85 
%else
86 
    packssdw    m5, m6
87 
    CLIPW       m5, m7, m4
88 
%endif
89 
%endmacro
90 
 
91 
%macro WEIGHT_FUNC_DBL 0
92 
cglobal h264_weight_16_10
93 
    WEIGHT_PROLOGUE
94 
    WEIGHT_SETUP
95 
.nextrow:
96 
    WEIGHT_OP  0
97 
    mova [r0   ], m5
98 
    WEIGHT_OP 16
99 
    mova [r0+16], m5
100 
    add       r0, r1
101 
    dec       r2d
102 
    jnz .nextrow
103 
    REP_RET
104 
%endmacro
105 
 
106 
INIT_XMM sse2
107 
WEIGHT_FUNC_DBL
108 
INIT_XMM sse4
109 
WEIGHT_FUNC_DBL
110 
 
111 
 
112 
%macro WEIGHT_FUNC_MM 0
113 
cglobal h264_weight_8_10
114 
    WEIGHT_PROLOGUE
115 
    WEIGHT_SETUP
116 
.nextrow:
117 
    WEIGHT_OP   0
118 
    mova     [r0], m5
119 
    add        r0, r1
120 
    dec        r2d
121 
    jnz .nextrow
122 
    REP_RET
123 
%endmacro
124 
 
125 
INIT_XMM sse2
126 
WEIGHT_FUNC_MM
127 
INIT_XMM sse4
128 
WEIGHT_FUNC_MM
129 
 
130 
 
131 
%macro WEIGHT_FUNC_HALF_MM 0
132 
cglobal h264_weight_4_10
133 
    WEIGHT_PROLOGUE
134 
    sar         r2d, 1
135 
    WEIGHT_SETUP
136 
    lea         r3, [r1*2]
137 
.nextrow:
138 
    WEIGHT_OP    0, r1
139 
    movh      [r0], m5
140 
    movhps [r0+r1], m5
141 
    add         r0, r3
142 
    dec         r2d
143 
    jnz .nextrow
144 
    REP_RET
145 
%endmacro
146 
 
147 
INIT_XMM sse2
148 
WEIGHT_FUNC_HALF_MM
149 
INIT_XMM sse4
150 
WEIGHT_FUNC_HALF_MM
151 
 
152 
 
153 
;-----------------------------------------------------------------------------
154 
; void h264_biweight(uint8_t *dst, uint8_t *src, int stride, int height,
155 
;                    int log2_denom, int weightd, int weights, int offset);
156 
;-----------------------------------------------------------------------------
157 
%if ARCH_X86_32
158 
DECLARE_REG_TMP 3
159 
%else
160 
DECLARE_REG_TMP 7
161 
%endif
162 
 
163 
%macro BIWEIGHT_PROLOGUE 0
164 
.prologue:
165 
    PROLOGUE 0,8,8
166 
    movifnidn  r0, r0mp
167 
    movifnidn  r1, r1mp
168 
    movifnidn r2d, r2m
169 
    movifnidn r5d, r5m
170 
    movifnidn r6d, r6m
171 
    movifnidn t0d, r7m
172 
%endmacro
173 
 
174 
%macro BIWEIGHT_SETUP 0
175 
    lea        t0, [t0*4+1] ; (offset<<2)+1
176 
    or         t0, 1
177 
    shl        r6, 16
178 
    or         r5, r6
179 
    movd       m4, r5d      ; weightd | weights
180 
    movd       m5, t0d      ; (offset+1)|1
181 
    movd       m6, r4m      ; log2_denom
182 
    pslld      m5, m6       ; (((offset<<2)+1)|1)<
183 
    paddd      m6, [sq_1]
184 
    pshufd     m4, m4, 0
185 
    pshufd     m5, m5, 0
186 
    mova       m3, [pw_pixel_max]
187 
    movifnidn r3d, r3m
188 
%if notcpuflag(sse4)
189 
    pxor       m7, m7
190 
%endif
191 
%endmacro
192 
 
193 
%macro BIWEIGHT 1-2
194 
%if %0==1
195 
    mova       m0, [r0+%1]
196 
    mova       m1, [r1+%1]
197 
    punpckhwd  m2, m0, m1
198 
    punpcklwd  m0, m1
199 
%else
200 
    movq       m0, [r0+%1]
201 
    movq       m1, [r1+%1]
202 
    punpcklwd  m0, m1
203 
    movq       m2, [r0+%2]
204 
    movq       m1, [r1+%2]
205 
    punpcklwd  m2, m1
206 
%endif
207 
    pmaddwd    m0, m4
208 
    pmaddwd    m2, m4
209 
    paddd      m0, m5
210 
    paddd      m2, m5
211 
    psrad      m0, m6
212 
    psrad      m2, m6
213 
%if cpuflag(sse4)
214 
    packusdw   m0, m2
215 
    pminsw     m0, m3
216 
%else
217 
    packssdw   m0, m2
218 
    CLIPW      m0, m7, m3
219 
%endif
220 
%endmacro
221 
 
222 
%macro BIWEIGHT_FUNC_DBL 0
223 
cglobal h264_biweight_16_10
224 
    BIWEIGHT_PROLOGUE
225 
    BIWEIGHT_SETUP
226 
.nextrow:
227 
    BIWEIGHT   0
228 
    mova [r0   ], m0
229 
    BIWEIGHT  16
230 
    mova [r0+16], m0
231 
    add       r0, r2
232 
    add       r1, r2
233 
    dec       r3d
234 
    jnz .nextrow
235 
    REP_RET
236 
%endmacro
237 
 
238 
INIT_XMM sse2
239 
BIWEIGHT_FUNC_DBL
240 
INIT_XMM sse4
241 
BIWEIGHT_FUNC_DBL
242 
 
243 
%macro BIWEIGHT_FUNC 0
244 
cglobal h264_biweight_8_10
245 
    BIWEIGHT_PROLOGUE
246 
    BIWEIGHT_SETUP
247 
.nextrow:
248 
    BIWEIGHT  0
249 
    mova   [r0], m0
250 
    add      r0, r2
251 
    add      r1, r2
252 
    dec      r3d
253 
    jnz .nextrow
254 
    REP_RET
255 
%endmacro
256 
 
257 
INIT_XMM sse2
258 
BIWEIGHT_FUNC
259 
INIT_XMM sse4
260 
BIWEIGHT_FUNC
261 
 
262 
%macro BIWEIGHT_FUNC_HALF 0
263 
cglobal h264_biweight_4_10
264 
    BIWEIGHT_PROLOGUE
265 
    BIWEIGHT_SETUP
266 
    sar        r3d, 1
267 
    lea        r4, [r2*2]
268 
.nextrow:
269 
    BIWEIGHT     0, r2
270 
    movh   [r0   ], m0
271 
    movhps [r0+r2], m0
272 
    add         r0, r4
273 
    add         r1, r4
274 
    dec         r3d
275 
    jnz .nextrow
276 
    REP_RET
277 
%endmacro
278 
 
279 
INIT_XMM sse2
280 
BIWEIGHT_FUNC_HALF
281 
INIT_XMM sse4
282 
BIWEIGHT_FUNC_HALF