forked from jamesathey/libjpeg-turbo
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathjcsample-sse2-64.asm
329 lines (263 loc) · 8.79 KB
/
jcsample-sse2-64.asm
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
;
; jcsample.asm - downsampling (64-bit SSE2)
;
; Copyright 2009 Pierre Ossman <[email protected]> for Cendio AB
; Copyright (C) 2009, D. R. Commander.
;
; Based on the x86 SIMD extension for IJG JPEG library
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
;
; [TAB8]
%include "jsimdext.inc"
; --------------------------------------------------------------------------
SECTION SEG_TEXT
BITS 64
;
; Downsample pixel values of a single component.
; This version handles the common case of 2:1 horizontal and 1:1 vertical,
; without smoothing.
;
; GLOBAL(void)
; jsimd_h2v1_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor,
; JDIMENSION v_samp_factor, JDIMENSION width_blocks,
; JSAMPARRAY input_data, JSAMPARRAY output_data);
;
; r10 = JDIMENSION image_width
; r11 = int max_v_samp_factor
; r12 = JDIMENSION v_samp_factor
; r13 = JDIMENSION width_blocks
; r14 = JSAMPARRAY input_data
; r15 = JSAMPARRAY output_data
align 16
global EXTN(jsimd_h2v1_downsample_sse2)
EXTN(jsimd_h2v1_downsample_sse2):
push rbp
mov rax,rsp
mov rbp,rsp
collect_args
mov ecx, r13d
shl rcx,3 ; imul rcx,DCTSIZE (rcx = output_cols)
jz near .return
mov edx, r10d
; -- expand_right_edge
push rcx
shl rcx,1 ; output_cols * 2
sub rcx,rdx
jle short .expand_end
mov rax, r11
test rax,rax
jle short .expand_end
cld
mov rsi, r14 ; input_data
.expandloop:
push rax
push rcx
mov rdi, JSAMPROW [rsi]
add rdi,rdx
mov al, JSAMPLE [rdi-1]
rep stosb
pop rcx
pop rax
add rsi, byte SIZEOF_JSAMPROW
dec rax
jg short .expandloop
.expand_end:
pop rcx ; output_cols
; -- h2v1_downsample
mov eax, r12d ; rowctr
test eax,eax
jle near .return
mov rdx, 0x00010000 ; bias pattern
movd xmm7,edx
pcmpeqw xmm6,xmm6
pshufd xmm7,xmm7,0x00 ; xmm7={0, 1, 0, 1, 0, 1, 0, 1}
psrlw xmm6,BYTE_BIT ; xmm6={0xFF 0x00 0xFF 0x00 ..}
mov rsi, r14 ; input_data
mov rdi, r15 ; output_data
.rowloop:
push rcx
push rdi
push rsi
mov rsi, JSAMPROW [rsi] ; inptr
mov rdi, JSAMPROW [rdi] ; outptr
cmp rcx, byte SIZEOF_XMMWORD
jae short .columnloop
.columnloop_r8:
movdqa xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD]
pxor xmm1,xmm1
mov rcx, SIZEOF_XMMWORD
jmp short .downsample
.columnloop:
movdqa xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD]
movdqa xmm1, XMMWORD [rsi+1*SIZEOF_XMMWORD]
.downsample:
movdqa xmm2,xmm0
movdqa xmm3,xmm1
pand xmm0,xmm6
psrlw xmm2,BYTE_BIT
pand xmm1,xmm6
psrlw xmm3,BYTE_BIT
paddw xmm0,xmm2
paddw xmm1,xmm3
paddw xmm0,xmm7
paddw xmm1,xmm7
psrlw xmm0,1
psrlw xmm1,1
packuswb xmm0,xmm1
movdqa XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0
sub rcx, byte SIZEOF_XMMWORD ; outcol
add rsi, byte 2*SIZEOF_XMMWORD ; inptr
add rdi, byte 1*SIZEOF_XMMWORD ; outptr
cmp rcx, byte SIZEOF_XMMWORD
jae short .columnloop
test rcx,rcx
jnz short .columnloop_r8
pop rsi
pop rdi
pop rcx
add rsi, byte SIZEOF_JSAMPROW ; input_data
add rdi, byte SIZEOF_JSAMPROW ; output_data
dec rax ; rowctr
jg near .rowloop
.return:
uncollect_args
pop rbp
ret
; --------------------------------------------------------------------------
;
; Downsample pixel values of a single component.
; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
; without smoothing.
;
; GLOBAL(void)
; jsimd_h2v2_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor,
; JDIMENSION v_samp_factor, JDIMENSION width_blocks,
; JSAMPARRAY input_data, JSAMPARRAY output_data);
;
; r10 = JDIMENSION image_width
; r11 = int max_v_samp_factor
; r12 = JDIMENSION v_samp_factor
; r13 = JDIMENSION width_blocks
; r14 = JSAMPARRAY input_data
; r15 = JSAMPARRAY output_data
align 16
global EXTN(jsimd_h2v2_downsample_sse2)
EXTN(jsimd_h2v2_downsample_sse2):
push rbp
mov rax,rsp
mov rbp,rsp
collect_args
mov ecx, r13d
shl rcx,3 ; imul rcx,DCTSIZE (rcx = output_cols)
jz near .return
mov edx, r10d
; -- expand_right_edge
push rcx
shl rcx,1 ; output_cols * 2
sub rcx,rdx
jle short .expand_end
mov rax, r11
test rax,rax
jle short .expand_end
cld
mov rsi, r14 ; input_data
.expandloop:
push rax
push rcx
mov rdi, JSAMPROW [rsi]
add rdi,rdx
mov al, JSAMPLE [rdi-1]
rep stosb
pop rcx
pop rax
add rsi, byte SIZEOF_JSAMPROW
dec rax
jg short .expandloop
.expand_end:
pop rcx ; output_cols
; -- h2v2_downsample
mov eax, r12d ; rowctr
test rax,rax
jle near .return
mov rdx, 0x00020001 ; bias pattern
movd xmm7,edx
pcmpeqw xmm6,xmm6
pshufd xmm7,xmm7,0x00 ; xmm7={1, 2, 1, 2, 1, 2, 1, 2}
psrlw xmm6,BYTE_BIT ; xmm6={0xFF 0x00 0xFF 0x00 ..}
mov rsi, r14 ; input_data
mov rdi, r15 ; output_data
.rowloop:
push rcx
push rdi
push rsi
mov rdx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW] ; inptr0
mov rsi, JSAMPROW [rsi+1*SIZEOF_JSAMPROW] ; inptr1
mov rdi, JSAMPROW [rdi] ; outptr
cmp rcx, byte SIZEOF_XMMWORD
jae short .columnloop
.columnloop_r8:
movdqa xmm0, XMMWORD [rdx+0*SIZEOF_XMMWORD]
movdqa xmm1, XMMWORD [rsi+0*SIZEOF_XMMWORD]
pxor xmm2,xmm2
pxor xmm3,xmm3
mov rcx, SIZEOF_XMMWORD
jmp short .downsample
.columnloop:
movdqa xmm0, XMMWORD [rdx+0*SIZEOF_XMMWORD]
movdqa xmm1, XMMWORD [rsi+0*SIZEOF_XMMWORD]
movdqa xmm2, XMMWORD [rdx+1*SIZEOF_XMMWORD]
movdqa xmm3, XMMWORD [rsi+1*SIZEOF_XMMWORD]
.downsample:
movdqa xmm4,xmm0
movdqa xmm5,xmm1
pand xmm0,xmm6
psrlw xmm4,BYTE_BIT
pand xmm1,xmm6
psrlw xmm5,BYTE_BIT
paddw xmm0,xmm4
paddw xmm1,xmm5
movdqa xmm4,xmm2
movdqa xmm5,xmm3
pand xmm2,xmm6
psrlw xmm4,BYTE_BIT
pand xmm3,xmm6
psrlw xmm5,BYTE_BIT
paddw xmm2,xmm4
paddw xmm3,xmm5
paddw xmm0,xmm1
paddw xmm2,xmm3
paddw xmm0,xmm7
paddw xmm2,xmm7
psrlw xmm0,2
psrlw xmm2,2
packuswb xmm0,xmm2
movdqa XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0
sub rcx, byte SIZEOF_XMMWORD ; outcol
add rdx, byte 2*SIZEOF_XMMWORD ; inptr0
add rsi, byte 2*SIZEOF_XMMWORD ; inptr1
add rdi, byte 1*SIZEOF_XMMWORD ; outptr
cmp rcx, byte SIZEOF_XMMWORD
jae near .columnloop
test rcx,rcx
jnz near .columnloop_r8
pop rsi
pop rdi
pop rcx
add rsi, byte 2*SIZEOF_JSAMPROW ; input_data
add rdi, byte 1*SIZEOF_JSAMPROW ; output_data
dec rax ; rowctr
jg near .rowloop
.return:
uncollect_args
pop rbp
ret
; For some reason, the OS X linker does not honor the request to align the
; segment unless we do this.
align 16