1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
|
# Pure Data Packet mmx routine.
# Copyright (c) by Tom Schouten <pdp@zzz.kotnet.org>
#
# 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., 675 Mass Ave, Cambridge, MA 02139, USA.
#
# TODO: COUPLED CASCADE SECOND ORDER SECTION
#
# s1[k] = ar * s1[k-1] + ai * s2[k-1] + x[k]
# s2[k] = ar * s2[k-1] - ai * s1[k-1]
# y[k] = c0 * x[k] + c1 * s1[k-1] + c2 * s2[k-1]
# MACRO: df2
#
# computes a coupled cascade
#
# input: %mm0 == input
# %mm1 == state 1
# %mm2 == state 2
# (%esi) == cascade coefs (ar ai c0 c1 c2) in s0.15
# output: %mm0 == output
# %mm1 == state 1
# %mm2 == state 2
.macro coupled
pmovq %mm1, %mm3 # mm3 == s1[k-1]
pmovq %mm1, %mm4 # mm4 == s1[k-1]
pmovq %mm2, %mm5 # mm5 == s2[k-1]
pmovq %mm2, %mm6 # mm5 == s2[k-1]
pmulhw (%esi), %mm1 # mm1 == s1[k-1] * ar
pmulhw 8(%esi), %mm3 # mm3 == s1[k-1] * ai
pmulhw 24(%esi), %mm4 # mm4 == s1[k-1] * c1
pmulhw (%esi), %mm2 # mm2 == s2[k-1] * ar
pmulhw 8(%esi), %mm5 # mm5 == s2[k-1] * ai
pmulhw 32(%esi), %mm6 # mm6 == s2[k-1] * c2
paddw %mm5, %mm1 # mm1 == s1[k-1] * ar + s2[k-1] * ai
psubw %mm3, %mm2 # mm2 == s2[k-1] * ar - s1[k-1] * ai == s2[k]
paddw %mm0, %mm1 # mm1 == s1[k]
pmulhw 16(%esi), %mm0 # mm0 == x[k] * c0
paddw %mm6, %mm4 # mm4 == s1[k-1] * c1 + s2[k-1] * c2
paddw %mm4, %mm0 # mm0 == y[k]
.endm
# in order to use the 4 line parallel cascade routine on horizontal
# lines, we need to reorder (rotate or transpose) the matrix, since
# images are scanline encoded, and we want to work in parallell
# on 4 lines.
#
# since the 4 lines are independent, it doesnt matter in which order
# the the vector elements are present.
#
# this allows us to use the same routine for left->right and right->left
# processing.
#
# some comments on the non-abelean group of square isometries consisting of
# (I) identity
# (H) horizontal axis mirror
# (V) vertical axis mirror
# (T) transpose (diagonal axis mirror)
# (A) antitranspose (antidiagonal axis mirror)
# (R1) 90deg anticlockwize rotation
# (R2) 180deg rotation
# (R3) 90deg clockwize rotation
#
#
# we basicly have two options: (R1,R3) or (T,A)
# we opt for T and A because they are self inverting, which improves locality
#
# use antitranspose for right to left an transpose
# for left to right (little endian)
# antitranspose 4x4
# input
# %mm3 == {d0 d1 d2 d3}
# %mm2 == {c0 c1 c2 c3}
# %mm1 == {b0 b1 b2 b3}
# %mm0 == {a0 a1 a2 a3}
# output
# %mm3 == {a3 b3 c3 d3}
# %mm2 == {a2 b2 c2 d2}
# %mm1 == {a1 b1 c1 d1}
# %mm0 == {a0 b0 c0 d0}
.macro antitranspose_4x4:
movq %mm3, %mm4
punpcklwd %mm1, %mm4 # mm4 <- {b2 d2 b3 d3}
movq %mm3, %mm5
punpckhwd %mm1, %mm5 # mm5 <- {b0 d0 b1 d1}
movq %mm2, %mm6
punpcklwd %mm0, %mm6 # mm6 <- {a2 c2 a3 c3}
movq %mm2, %mm7
punpckhwd %mm0, %mm7 # mm7 <- {a0 c0 a1 c1}
movq %mm4, %mm3
punpcklwd %mm6, %mm3 # mm3 <- {a3 b3 c3 d3}
movq %mm4, %mm2
punpckhwd %mm6, %mm2 # mm2 <- {a2 b2 c2 d2}
movq %mm5, %mm1
punpcklwd %mm7, %mm1 # mm1 <- {a1 b1 c1 d1}
movq %mm5, %mm0
punpckhwd %mm7, %mm0 # mm0 <- {a0 b0 c0 d0}
.endm
# transpose 4x4
# input
# %mm3 == {d3 d2 d1 d0}
# %mm2 == {c3 c2 c1 c0}
# %mm1 == {b3 b2 b1 b0}
# %mm0 == {a3 a2 a1 a0}
# output
# %mm3 == {d3 c3 b3 a3}
# %mm2 == {d2 c2 b2 a2}
# %mm1 == {d1 c1 b1 a1}
# %mm0 == {d0 c0 b0 a0}
.macro transpose_4x4:
movq %mm0, %mm4
punpcklwd %mm2, %mm4 # mm4 <- {c1 a1 c0 a0}
movq %mm0, %mm5
punpckhwd %mm2, %mm5 # mm5 <- {c3 a3 c2 a2}
movq %mm1, %mm6
punpcklwd %mm3, %mm6 # mm6 <- {d1 b1 d0 b0}
movq %mm1, %mm7
punpckhwd %mm3, %mm7 # mm7 <- {d3 b3 d2 b2}
movq %mm4, %mm0
punpcklwd %mm6, %mm0 # mm0 <- {d0 c0 b0 a0}
movq %mm4, %mm1
punpckhwd %mm6, %mm1 # mm1 <- {d1 c1 b1 a1}
movq %mm5, %mm2
punpcklwd %mm7, %mm2 # mm2 <- {d2 c2 b2 a2}
movq %mm5, %mm3
punpckhwd %mm7, %mm3 # mm3 <- {d3 c3 b3 a3}
.endm
.globl pixel_cascade_vertb_s16
.type pixel_cascade_vertb_s16,@function
# pixel_cascade_vertbr_s16(char *pixel_array, int nb_rows, int linewidth, short int coef[20], short int state[8])
pixel_cascade_vertb_s16:
pushl %ebp
movl %esp, %ebp
push %ebx
push %esi
push %edi
movl 8(%ebp), %ebx # pixel array offset
movl 12(%ebp), %ecx # nb of 4x4 pixblocks
movl 16(%ebp), %edx # line with
movl 20(%ebp), %esi # coefs
movl 24(%ebp), %edi # state
shll $1, %edx # short int addressing
subl %edx, %ebx
movq 0(%edi), %mm1 # s1[k-1]
movq 8(%edi), %mm2 # s2[k-1]
.align 16
.cascade_vertb_line_loop:
movq (%ebx,%edx,1), %mm3
movq %mm3, %mm0
addl %edx, %ebx
coupled
movq %mm0, (%ebx)
movq (%ebx,%edx,1), %mm3
movq %mm3, %mm0
addl %edx, %ebx
coupled
movq %mm0, (%ebx)
movq (%ebx,%edx,1), %mm3
movq %mm3, %mm0
addl %edx, %ebx
coupled
movq %mm0, (%ebx)
movq (%ebx,%edx,1), %mm3
movq %mm3, %mm0
addl %edx, %ebx
coupled
movq %mm0, (%ebx)
decl %ecx
jnz .cascade_vertb_line_loop
movq %mm1, 0(%edi) # s1[k-1]
movq %mm2, 8(%edi) # s2[k-1]
emms
pop %edi
pop %esi
pop %ebx
leave
ret
.globl pixel_cascade_horlr_s16
.type pixel_cascade_horlr_s16,@function
# pixel_cascade_hor_s16(char *pixel_array, int nb_rows, int linewidth, short int coef[20], short int state[8])
pixel_cascade_horlr_s16:
pushl %ebp
movl %esp, %ebp
push %ebx
push %esi
push %edi
movl 8(%ebp), %ebx # pixel array offset
movl 12(%ebp), %ecx # nb of 4x4 pixblocks
movl 16(%ebp), %edx # line with
movl 20(%ebp), %esi # coefs
movl 24(%ebp), %edi # state
shll $1, %edx # short int addressing
movl %edx, %eax
shll $1, %eax
addl %edx, %eax # eax = 3 * edx
.align 16
.cascade_horlr_line_loop:
movq (%edi), %mm1
movq 8(%edi), %mm2
movq (%ebx), %mm0
movq (%ebx,%edx,1), %mm1
movq (%ebx,%edx,2), %mm2
movq (%ebx,%eax,1), %mm3
transpose_4x4
movq %mm1, (%ebx,%edx,1)
movq %mm2, (%ebx,%edx,2)
movq %mm3, (%ebx,%eax,1)
coupled
movq %mm0, (%ebx)
movq (%ebx,%edx,1), %mm3
movq %mm3, %mm0
coupled
movq %mm0, (%ebx, %edx,1)
movq (%ebx,%edx,2), %mm3
movq %mm3, %mm0
coupled
movq %mm0, (%ebx, %edx,2)
movq (%ebx,%eax,1), %mm3
movq %mm3, %mm0
coupled
movq %mm1, 0(%edi) # s1[k-1]
movq %mm2, 8(%edi) # s2[k-1]
movq %mm0, %mm3
movq (%ebx), %mm0
movq (%ebx,%edx,1), %mm1
movq (%ebx,%edx,2), %mm2
transpose_4x4
movq %mm0, (%ebx)
movq %mm1, (%ebx,%edx,1)
movq %mm2, (%ebx,%edx,2)
movq %mm3, (%ebx,%eax,1)
addl $8, %ebx
decl %ecx
jnz .cascade_horlr_line_loop
emms
pop %edi
pop %esi
pop %ebx
leave
ret
|