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
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
|
;
; jdclrss2.asm - colorspace conversion (64-bit SSE2)
;
; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
; Copyright 2009 D. R. Commander
;
; Based on
; 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 "jcolsamp.inc"
; --------------------------------------------------------------------------
SECTION SEG_TEXT
BITS 64
;
; Convert some rows of samples to the output colorspace.
;
; GLOBAL(void)
; jsimd_ycc_rgb_convert_sse2 (JDIMENSION out_width,
; JSAMPIMAGE input_buf, JDIMENSION input_row,
; JSAMPARRAY output_buf, int num_rows)
;
; r10 = JDIMENSION out_width
; r11 = JSAMPIMAGE input_buf
; r12 = JDIMENSION input_row
; r13 = JSAMPARRAY output_buf
; r14 = int num_rows
%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
%define WK_NUM 2
align 16
global EXTN(jsimd_ycc_rgb_convert_sse2)
EXTN(jsimd_ycc_rgb_convert_sse2):
push rbp
mov rax,rsp ; rax = original rbp
sub rsp, byte 4
and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits
mov [rsp],rax
mov rbp,rsp ; rbp = aligned rbp
lea rsp, [wk(0)]
push rbx
collect_args
mov rcx, r10 ; num_cols
test rcx,rcx
jz near .return
push rcx
mov rdi, r11
mov rcx, r12
mov rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY]
mov rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY]
mov rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY]
lea rsi, [rsi+rcx*SIZEOF_JSAMPROW]
lea rbx, [rbx+rcx*SIZEOF_JSAMPROW]
lea rdx, [rdx+rcx*SIZEOF_JSAMPROW]
pop rcx
mov rdi, r13
mov rax, r14
test rax,rax
jle near .return
.rowloop:
push rax
push rdi
push rdx
push rbx
push rsi
push rcx ; col
mov rsi, JSAMPROW [rsi] ; inptr0
mov rbx, JSAMPROW [rbx] ; inptr1
mov rdx, JSAMPROW [rdx] ; inptr2
mov rdi, JSAMPROW [rdi] ; outptr
.columnloop:
movdqa xmm5, XMMWORD [rbx] ; xmm5=Cb(0123456789ABCDEF)
movdqa xmm1, XMMWORD [rdx] ; xmm1=Cr(0123456789ABCDEF)
pcmpeqw xmm4,xmm4
pcmpeqw xmm7,xmm7
psrlw xmm4,BYTE_BIT
psllw xmm7,7 ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
movdqa xmm0,xmm4 ; xmm0=xmm4={0xFF 0x00 0xFF 0x00 ..}
pand xmm4,xmm5 ; xmm4=Cb(02468ACE)=CbE
psrlw xmm5,BYTE_BIT ; xmm5=Cb(13579BDF)=CbO
pand xmm0,xmm1 ; xmm0=Cr(02468ACE)=CrE
psrlw xmm1,BYTE_BIT ; xmm1=Cr(13579BDF)=CrO
paddw xmm4,xmm7
paddw xmm5,xmm7
paddw xmm0,xmm7
paddw xmm1,xmm7
; (Original)
; R = Y + 1.40200 * Cr
; G = Y - 0.34414 * Cb - 0.71414 * Cr
; B = Y + 1.77200 * Cb
;
; (This implementation)
; R = Y + 0.40200 * Cr + Cr
; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
; B = Y - 0.22800 * Cb + Cb + Cb
movdqa xmm2,xmm4 ; xmm2=CbE
movdqa xmm3,xmm5 ; xmm3=CbO
paddw xmm4,xmm4 ; xmm4=2*CbE
paddw xmm5,xmm5 ; xmm5=2*CbO
movdqa xmm6,xmm0 ; xmm6=CrE
movdqa xmm7,xmm1 ; xmm7=CrO
paddw xmm0,xmm0 ; xmm0=2*CrE
paddw xmm1,xmm1 ; xmm1=2*CrO
pmulhw xmm4,[PW_MF0228] ; xmm4=(2*CbE * -FIX(0.22800))
pmulhw xmm5,[PW_MF0228] ; xmm5=(2*CbO * -FIX(0.22800))
pmulhw xmm0,[PW_F0402] ; xmm0=(2*CrE * FIX(0.40200))
pmulhw xmm1,[PW_F0402] ; xmm1=(2*CrO * FIX(0.40200))
paddw xmm4,[PW_ONE]
paddw xmm5,[PW_ONE]
psraw xmm4,1 ; xmm4=(CbE * -FIX(0.22800))
psraw xmm5,1 ; xmm5=(CbO * -FIX(0.22800))
paddw xmm0,[PW_ONE]
paddw xmm1,[PW_ONE]
psraw xmm0,1 ; xmm0=(CrE * FIX(0.40200))
psraw xmm1,1 ; xmm1=(CrO * FIX(0.40200))
paddw xmm4,xmm2
paddw xmm5,xmm3
paddw xmm4,xmm2 ; xmm4=(CbE * FIX(1.77200))=(B-Y)E
paddw xmm5,xmm3 ; xmm5=(CbO * FIX(1.77200))=(B-Y)O
paddw xmm0,xmm6 ; xmm0=(CrE * FIX(1.40200))=(R-Y)E
paddw xmm1,xmm7 ; xmm1=(CrO * FIX(1.40200))=(R-Y)O
movdqa XMMWORD [wk(0)], xmm4 ; wk(0)=(B-Y)E
movdqa XMMWORD [wk(1)], xmm5 ; wk(1)=(B-Y)O
movdqa xmm4,xmm2
movdqa xmm5,xmm3
punpcklwd xmm2,xmm6
punpckhwd xmm4,xmm6
pmaddwd xmm2,[PW_MF0344_F0285]
pmaddwd xmm4,[PW_MF0344_F0285]
punpcklwd xmm3,xmm7
punpckhwd xmm5,xmm7
pmaddwd xmm3,[PW_MF0344_F0285]
pmaddwd xmm5,[PW_MF0344_F0285]
paddd xmm2,[PD_ONEHALF]
paddd xmm4,[PD_ONEHALF]
psrad xmm2,SCALEBITS
psrad xmm4,SCALEBITS
paddd xmm3,[PD_ONEHALF]
paddd xmm5,[PD_ONEHALF]
psrad xmm3,SCALEBITS
psrad xmm5,SCALEBITS
packssdw xmm2,xmm4 ; xmm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
packssdw xmm3,xmm5 ; xmm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
psubw xmm2,xmm6 ; xmm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
psubw xmm3,xmm7 ; xmm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
movdqa xmm5, XMMWORD [rsi] ; xmm5=Y(0123456789ABCDEF)
pcmpeqw xmm4,xmm4
psrlw xmm4,BYTE_BIT ; xmm4={0xFF 0x00 0xFF 0x00 ..}
pand xmm4,xmm5 ; xmm4=Y(02468ACE)=YE
psrlw xmm5,BYTE_BIT ; xmm5=Y(13579BDF)=YO
paddw xmm0,xmm4 ; xmm0=((R-Y)E+YE)=RE=R(02468ACE)
paddw xmm1,xmm5 ; xmm1=((R-Y)O+YO)=RO=R(13579BDF)
packuswb xmm0,xmm0 ; xmm0=R(02468ACE********)
packuswb xmm1,xmm1 ; xmm1=R(13579BDF********)
paddw xmm2,xmm4 ; xmm2=((G-Y)E+YE)=GE=G(02468ACE)
paddw xmm3,xmm5 ; xmm3=((G-Y)O+YO)=GO=G(13579BDF)
packuswb xmm2,xmm2 ; xmm2=G(02468ACE********)
packuswb xmm3,xmm3 ; xmm3=G(13579BDF********)
paddw xmm4, XMMWORD [wk(0)] ; xmm4=(YE+(B-Y)E)=BE=B(02468ACE)
paddw xmm5, XMMWORD [wk(1)] ; xmm5=(YO+(B-Y)O)=BO=B(13579BDF)
packuswb xmm4,xmm4 ; xmm4=B(02468ACE********)
packuswb xmm5,xmm5 ; xmm5=B(13579BDF********)
%if RGB_PIXELSIZE == 3 ; ---------------
; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
punpcklbw xmmA,xmmC ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
punpcklbw xmmE,xmmB ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
punpcklbw xmmD,xmmF ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
movdqa xmmG,xmmA
movdqa xmmH,xmmA
punpcklwd xmmA,xmmE ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
punpckhwd xmmG,xmmE ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
psrldq xmmH,2 ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
psrldq xmmE,2 ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
movdqa xmmC,xmmD
movdqa xmmB,xmmD
punpcklwd xmmD,xmmH ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
punpckhwd xmmC,xmmH ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
psrldq xmmB,2 ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
movdqa xmmF,xmmE
punpcklwd xmmE,xmmB ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
punpckhwd xmmF,xmmB ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
pshufd xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
movdqa xmmB,xmmE
punpckldq xmmA,xmmD ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
punpckldq xmmE,xmmH ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
punpckhdq xmmD,xmmB ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
pshufd xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
movdqa xmmB,xmmF
punpckldq xmmG,xmmC ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
punpckldq xmmF,xmmH ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
punpckhdq xmmC,xmmB ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
punpcklqdq xmmA,xmmE ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
punpcklqdq xmmD,xmmG ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
punpcklqdq xmmF,xmmC ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
cmp rcx, byte SIZEOF_XMMWORD
jb short .column_st32
test rdi, SIZEOF_XMMWORD-1
jnz short .out1
; --(aligned)-------------------
movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF
add rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; outptr
jmp short .out0
.out1: ; --(unaligned)-----------------
pcmpeqb xmmH,xmmH ; xmmH=(all 1's)
maskmovdqu xmmA,xmmH ; movntdqu XMMWORD [rdi], xmmA
add rdi, byte SIZEOF_XMMWORD ; outptr
maskmovdqu xmmD,xmmH ; movntdqu XMMWORD [rdi], xmmD
add rdi, byte SIZEOF_XMMWORD ; outptr
maskmovdqu xmmF,xmmH ; movntdqu XMMWORD [rdi], xmmF
add rdi, byte SIZEOF_XMMWORD ; outptr
.out0:
sub rcx, byte SIZEOF_XMMWORD
jz near .nextrow
add rsi, byte SIZEOF_XMMWORD ; inptr0
add rbx, byte SIZEOF_XMMWORD ; inptr1
add rdx, byte SIZEOF_XMMWORD ; inptr2
jmp near .columnloop
.column_st32:
pcmpeqb xmmH,xmmH ; xmmH=(all 1's)
lea rcx, [rcx+rcx*2] ; imul ecx, RGB_PIXELSIZE
cmp rcx, byte 2*SIZEOF_XMMWORD
jb short .column_st16
maskmovdqu xmmA,xmmH ; movntdqu XMMWORD [rdi], xmmA
add rdi, byte SIZEOF_XMMWORD ; outptr
maskmovdqu xmmD,xmmH ; movntdqu XMMWORD [rdi], xmmD
add rdi, byte SIZEOF_XMMWORD ; outptr
movdqa xmmA,xmmF
sub rcx, byte 2*SIZEOF_XMMWORD
jmp short .column_st15
.column_st16:
cmp rcx, byte SIZEOF_XMMWORD
jb short .column_st15
maskmovdqu xmmA,xmmH ; movntdqu XMMWORD [rdi], xmmA
add rdi, byte SIZEOF_XMMWORD ; outptr
movdqa xmmA,xmmD
sub rcx, byte SIZEOF_XMMWORD
.column_st15:
mov rax,rcx
xor rcx, byte 0x0F
shl rcx, 2
movd xmmB,ecx
psrlq xmmH,4
pcmpeqb xmmE,xmmE
psrlq xmmH,xmmB
psrlq xmmE,xmmB
punpcklbw xmmE,xmmH
; ----------------
mov rcx,rdi
and rcx, byte SIZEOF_XMMWORD-1
jz short .adj0
add rax,rcx
cmp rax, byte SIZEOF_XMMWORD
ja short .adj0
and rdi, byte (-SIZEOF_XMMWORD) ; align to 16-byte boundary
shl rcx, 3 ; pslldq xmmA,ecx & pslldq xmmE,rcx
movdqa xmmG,xmmA
movdqa xmmC,xmmE
pslldq xmmA, SIZEOF_XMMWORD/2
pslldq xmmE, SIZEOF_XMMWORD/2
movd xmmD,ecx
sub rcx, byte (SIZEOF_XMMWORD/2)*BYTE_BIT
jb short .adj1
movd xmmF,ecx
psllq xmmA,xmmF
psllq xmmE,xmmF
jmp short .adj0
.adj1: neg ecx
movd xmmF,ecx
psrlq xmmA,xmmF
psrlq xmmE,xmmF
psllq xmmG,xmmD
psllq xmmC,xmmD
por xmmA,xmmG
por xmmE,xmmC
.adj0: ; ----------------
maskmovdqu xmmA,xmmE ; movntdqu XMMWORD [rdi], xmmA
%else ; RGB_PIXELSIZE == 4 ; -----------
%ifdef RGBX_FILLER_0XFF
pcmpeqb xmm6,xmm6 ; xmm6=XE=X(02468ACE********)
pcmpeqb xmm7,xmm7 ; xmm7=XO=X(13579BDF********)
%else
pxor xmm6,xmm6 ; xmm6=XE=X(02468ACE********)
pxor xmm7,xmm7 ; xmm7=XO=X(13579BDF********)
%endif
; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
punpcklbw xmmA,xmmC ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
punpcklbw xmmE,xmmG ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
punpcklbw xmmB,xmmD ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
punpcklbw xmmF,xmmH ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
movdqa xmmC,xmmA
punpcklwd xmmA,xmmE ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
punpckhwd xmmC,xmmE ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
movdqa xmmG,xmmB
punpcklwd xmmB,xmmF ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
punpckhwd xmmG,xmmF ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
movdqa xmmD,xmmA
punpckldq xmmA,xmmB ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
punpckhdq xmmD,xmmB ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
movdqa xmmH,xmmC
punpckldq xmmC,xmmG ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
punpckhdq xmmH,xmmG ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
cmp rcx, byte SIZEOF_XMMWORD
jb short .column_st32
test rdi, SIZEOF_XMMWORD-1
jnz short .out1
; --(aligned)-------------------
movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC
movntdq XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH
add rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; outptr
jmp short .out0
.out1: ; --(unaligned)-----------------
pcmpeqb xmmE,xmmE ; xmmE=(all 1's)
maskmovdqu xmmA,xmmE ; movntdqu XMMWORD [rdi], xmmA
add rdi, byte SIZEOF_XMMWORD ; outptr
maskmovdqu xmmD,xmmE ; movntdqu XMMWORD [rdi], xmmD
add rdi, byte SIZEOF_XMMWORD ; outptr
maskmovdqu xmmC,xmmE ; movntdqu XMMWORD [rdi], xmmC
add rdi, byte SIZEOF_XMMWORD ; outptr
maskmovdqu xmmH,xmmE ; movntdqu XMMWORD [rdi], xmmH
add rdi, byte SIZEOF_XMMWORD ; outptr
.out0:
sub rcx, byte SIZEOF_XMMWORD
jz near .nextrow
add rsi, byte SIZEOF_XMMWORD ; inptr0
add rbx, byte SIZEOF_XMMWORD ; inptr1
add rdx, byte SIZEOF_XMMWORD ; inptr2
jmp near .columnloop
.column_st32:
pcmpeqb xmmE,xmmE ; xmmE=(all 1's)
cmp rcx, byte SIZEOF_XMMWORD/2
jb short .column_st16
maskmovdqu xmmA,xmmE ; movntdqu XMMWORD [rdi], xmmA
add rdi, byte SIZEOF_XMMWORD ; outptr
maskmovdqu xmmD,xmmE ; movntdqu XMMWORD [rdi], xmmD
add rdi, byte SIZEOF_XMMWORD ; outptr
movdqa xmmA,xmmC
movdqa xmmD,xmmH
sub rcx, byte SIZEOF_XMMWORD/2
.column_st16:
cmp rcx, byte SIZEOF_XMMWORD/4
jb short .column_st15
maskmovdqu xmmA,xmmE ; movntdqu XMMWORD [rdi], xmmA
add rdi, byte SIZEOF_XMMWORD ; outptr
movdqa xmmA,xmmD
sub rcx, byte SIZEOF_XMMWORD/4
.column_st15:
cmp rcx, byte SIZEOF_XMMWORD/16
jb near .nextrow
mov rax,rcx
xor rcx, byte 0x03
inc rcx
shl rcx, 4
movd xmmF,ecx
psrlq xmmE,xmmF
punpcklbw xmmE,xmmE
; ----------------
mov rcx,rdi
and rcx, byte SIZEOF_XMMWORD-1
jz short .adj0
lea rax, [rcx+rax*4] ; RGB_PIXELSIZE
cmp rax, byte SIZEOF_XMMWORD
ja short .adj0
and rdi, byte (-SIZEOF_XMMWORD) ; align to 16-byte boundary
shl rcx, 3 ; pslldq xmmA,ecx & pslldq xmmE,ecx
movdqa xmmB,xmmA
movdqa xmmG,xmmE
pslldq xmmA, SIZEOF_XMMWORD/2
pslldq xmmE, SIZEOF_XMMWORD/2
movd xmmC,ecx
sub rcx, byte (SIZEOF_XMMWORD/2)*BYTE_BIT
jb short .adj1
movd xmmH,ecx
psllq xmmA,xmmH
psllq xmmE,xmmH
jmp short .adj0
.adj1: neg rcx
movd xmmH,ecx
psrlq xmmA,xmmH
psrlq xmmE,xmmH
psllq xmmB,xmmC
psllq xmmG,xmmC
por xmmA,xmmB
por xmmE,xmmG
.adj0: ; ----------------
maskmovdqu xmmA,xmmE ; movntdqu XMMWORD [rdi], xmmA
%endif ; RGB_PIXELSIZE ; ---------------
.nextrow:
pop rcx
pop rsi
pop rbx
pop rdx
pop rdi
pop rax
add rsi, byte SIZEOF_JSAMPROW
add rbx, byte SIZEOF_JSAMPROW
add rdx, byte SIZEOF_JSAMPROW
add rdi, byte SIZEOF_JSAMPROW ; output_buf
dec rax ; num_rows
jg near .rowloop
sfence ; flush the write buffer
.return:
uncollect_args
pop rbx
mov rsp,rbp ; rsp <- aligned rbp
pop rsp ; rsp <- original rbp
pop rbp
ret
|
