summaryrefslogtreecommitdiffstats
path: root/media/libjpeg/simd/jfdctfst-sse2.asm
blob: 54856a2363b211a0a220453240911bc29bbe32ad (plain)
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
;
; jfdctfst.asm - fast integer FDCT (SSE2)
;
; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
;
; 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
;
; This file contains a fast, not so accurate integer implementation of
; the forward DCT (Discrete Cosine Transform). The following code is
; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c
; for more details.
;
; [TAB8]

%include "jsimdext.inc"
%include "jdct.inc"

; --------------------------------------------------------------------------

%define CONST_BITS      8       ; 14 is also OK.

%if CONST_BITS == 8
F_0_382 equ      98             ; FIX(0.382683433)
F_0_541 equ     139             ; FIX(0.541196100)
F_0_707 equ     181             ; FIX(0.707106781)
F_1_306 equ     334             ; FIX(1.306562965)
%else
; NASM cannot do compile-time arithmetic on floating-point constants.
%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
F_0_382 equ     DESCALE( 410903207,30-CONST_BITS)       ; FIX(0.382683433)
F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
F_0_707 equ     DESCALE( 759250124,30-CONST_BITS)       ; FIX(0.707106781)
F_1_306 equ     DESCALE(1402911301,30-CONST_BITS)       ; FIX(1.306562965)
%endif

; --------------------------------------------------------------------------
        SECTION SEG_CONST

; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)

%define PRE_MULTIPLY_SCALE_BITS   2
%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)

        alignz  16
        global  EXTN(jconst_fdct_ifast_sse2)

EXTN(jconst_fdct_ifast_sse2):

PW_F0707        times 8 dw  F_0_707 << CONST_SHIFT
PW_F0382        times 8 dw  F_0_382 << CONST_SHIFT
PW_F0541        times 8 dw  F_0_541 << CONST_SHIFT
PW_F1306        times 8 dw  F_1_306 << CONST_SHIFT

        alignz  16

; --------------------------------------------------------------------------
        SECTION SEG_TEXT
        BITS    32
;
; Perform the forward DCT on one block of samples.
;
; GLOBAL(void)
; jsimd_fdct_ifast_sse2 (DCTELEM *data)
;

%define data(b)         (b)+8           ; DCTELEM *data

%define original_ebp    ebp+0
%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
%define WK_NUM          2

        align   16
        global  EXTN(jsimd_fdct_ifast_sse2)

EXTN(jsimd_fdct_ifast_sse2):
        push    ebp
        mov     eax,esp                         ; eax = original ebp
        sub     esp, byte 4
        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
        mov     [esp],eax
        mov     ebp,esp                         ; ebp = aligned ebp
        lea     esp, [wk(0)]
        pushpic ebx
;       push    ecx             ; unused
;       push    edx             ; need not be preserved
;       push    esi             ; unused
;       push    edi             ; unused

        get_GOT ebx             ; get GOT address

        ; ---- Pass 1: process rows.

        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)

        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
        movdqa  xmm1, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
        movdqa  xmm2, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)]

        ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
        ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)

        movdqa    xmm4,xmm0             ; transpose coefficients(phase 1)
        punpcklwd xmm0,xmm1             ; xmm0=(00 10 01 11 02 12 03 13)
        punpckhwd xmm4,xmm1             ; xmm4=(04 14 05 15 06 16 07 17)
        movdqa    xmm5,xmm2             ; transpose coefficients(phase 1)
        punpcklwd xmm2,xmm3             ; xmm2=(20 30 21 31 22 32 23 33)
        punpckhwd xmm5,xmm3             ; xmm5=(24 34 25 35 26 36 27 37)

        movdqa  xmm6, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)]
        movdqa  xmm7, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)]
        movdqa  xmm1, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)]
        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)]

        ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
        ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)

        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)

        movdqa    xmm2,xmm6             ; transpose coefficients(phase 1)
        punpcklwd xmm6,xmm7             ; xmm6=(40 50 41 51 42 52 43 53)
        punpckhwd xmm2,xmm7             ; xmm2=(44 54 45 55 46 56 47 57)
        movdqa    xmm5,xmm1             ; transpose coefficients(phase 1)
        punpcklwd xmm1,xmm3             ; xmm1=(60 70 61 71 62 72 63 73)
        punpckhwd xmm5,xmm3             ; xmm5=(64 74 65 75 66 76 67 77)

        movdqa    xmm7,xmm6             ; transpose coefficients(phase 2)
        punpckldq xmm6,xmm1             ; xmm6=(40 50 60 70 41 51 61 71)
        punpckhdq xmm7,xmm1             ; xmm7=(42 52 62 72 43 53 63 73)
        movdqa    xmm3,xmm2             ; transpose coefficients(phase 2)
        punpckldq xmm2,xmm5             ; xmm2=(44 54 64 74 45 55 65 75)
        punpckhdq xmm3,xmm5             ; xmm3=(46 56 66 76 47 57 67 77)

        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(20 30 21 31 22 32 23 33)
        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(24 34 25 35 26 36 27 37)
        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=(42 52 62 72 43 53 63 73)
        movdqa  XMMWORD [wk(1)], xmm2   ; wk(1)=(44 54 64 74 45 55 65 75)

        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
        punpckldq xmm0,xmm1             ; xmm0=(00 10 20 30 01 11 21 31)
        punpckhdq xmm7,xmm1             ; xmm7=(02 12 22 32 03 13 23 33)
        movdqa    xmm2,xmm4             ; transpose coefficients(phase 2)
        punpckldq xmm4,xmm5             ; xmm4=(04 14 24 34 05 15 25 35)
        punpckhdq xmm2,xmm5             ; xmm2=(06 16 26 36 07 17 27 37)

        movdqa     xmm1,xmm0            ; transpose coefficients(phase 3)
        punpcklqdq xmm0,xmm6            ; xmm0=(00 10 20 30 40 50 60 70)=data0
        punpckhqdq xmm1,xmm6            ; xmm1=(01 11 21 31 41 51 61 71)=data1
        movdqa     xmm5,xmm2            ; transpose coefficients(phase 3)
        punpcklqdq xmm2,xmm3            ; xmm2=(06 16 26 36 46 56 66 76)=data6
        punpckhqdq xmm5,xmm3            ; xmm5=(07 17 27 37 47 57 67 77)=data7

        movdqa  xmm6,xmm1
        movdqa  xmm3,xmm0
        psubw   xmm1,xmm2               ; xmm1=data1-data6=tmp6
        psubw   xmm0,xmm5               ; xmm0=data0-data7=tmp7
        paddw   xmm6,xmm2               ; xmm6=data1+data6=tmp1
        paddw   xmm3,xmm5               ; xmm3=data0+data7=tmp0

        movdqa  xmm2, XMMWORD [wk(0)]   ; xmm2=(42 52 62 72 43 53 63 73)
        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(44 54 64 74 45 55 65 75)
        movdqa  XMMWORD [wk(0)], xmm1   ; wk(0)=tmp6
        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp7

        movdqa     xmm1,xmm7            ; transpose coefficients(phase 3)
        punpcklqdq xmm7,xmm2            ; xmm7=(02 12 22 32 42 52 62 72)=data2
        punpckhqdq xmm1,xmm2            ; xmm1=(03 13 23 33 43 53 63 73)=data3
        movdqa     xmm0,xmm4            ; transpose coefficients(phase 3)
        punpcklqdq xmm4,xmm5            ; xmm4=(04 14 24 34 44 54 64 74)=data4
        punpckhqdq xmm0,xmm5            ; xmm0=(05 15 25 35 45 55 65 75)=data5

        movdqa  xmm2,xmm1
        movdqa  xmm5,xmm7
        paddw   xmm1,xmm4               ; xmm1=data3+data4=tmp3
        paddw   xmm7,xmm0               ; xmm7=data2+data5=tmp2
        psubw   xmm2,xmm4               ; xmm2=data3-data4=tmp4
        psubw   xmm5,xmm0               ; xmm5=data2-data5=tmp5

        ; -- Even part

        movdqa  xmm4,xmm3
        movdqa  xmm0,xmm6
        psubw   xmm3,xmm1               ; xmm3=tmp13
        psubw   xmm6,xmm7               ; xmm6=tmp12
        paddw   xmm4,xmm1               ; xmm4=tmp10
        paddw   xmm0,xmm7               ; xmm0=tmp11

        paddw   xmm6,xmm3
        psllw   xmm6,PRE_MULTIPLY_SCALE_BITS
        pmulhw  xmm6,[GOTOFF(ebx,PW_F0707)] ; xmm6=z1

        movdqa  xmm1,xmm4
        movdqa  xmm7,xmm3
        psubw   xmm4,xmm0               ; xmm4=data4
        psubw   xmm3,xmm6               ; xmm3=data6
        paddw   xmm1,xmm0               ; xmm1=data0
        paddw   xmm7,xmm6               ; xmm7=data2

        movdqa  xmm0, XMMWORD [wk(0)]   ; xmm0=tmp6
        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=tmp7
        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=data4
        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=data6

        ; -- Odd part

        paddw   xmm2,xmm5               ; xmm2=tmp10
        paddw   xmm5,xmm0               ; xmm5=tmp11
        paddw   xmm0,xmm6               ; xmm0=tmp12, xmm6=tmp7

        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS

        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
        pmulhw  xmm5,[GOTOFF(ebx,PW_F0707)] ; xmm5=z3

        movdqa  xmm4,xmm2               ; xmm4=tmp10
        psubw   xmm2,xmm0
        pmulhw  xmm2,[GOTOFF(ebx,PW_F0382)] ; xmm2=z5
        pmulhw  xmm4,[GOTOFF(ebx,PW_F0541)] ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
        pmulhw  xmm0,[GOTOFF(ebx,PW_F1306)] ; xmm0=MULTIPLY(tmp12,FIX_1_306562)
        paddw   xmm4,xmm2               ; xmm4=z2
        paddw   xmm0,xmm2               ; xmm0=z4

        movdqa  xmm3,xmm6
        psubw   xmm6,xmm5               ; xmm6=z13
        paddw   xmm3,xmm5               ; xmm3=z11

        movdqa  xmm2,xmm6
        movdqa  xmm5,xmm3
        psubw   xmm6,xmm4               ; xmm6=data3
        psubw   xmm3,xmm0               ; xmm3=data7
        paddw   xmm2,xmm4               ; xmm2=data5
        paddw   xmm5,xmm0               ; xmm5=data1

        ; ---- Pass 2: process columns.

;       mov     edx, POINTER [data(eax)]        ; (DCTELEM *)

        ; xmm1=(00 10 20 30 40 50 60 70), xmm7=(02 12 22 32 42 52 62 72)
        ; xmm5=(01 11 21 31 41 51 61 71), xmm6=(03 13 23 33 43 53 63 73)

        movdqa    xmm4,xmm1             ; transpose coefficients(phase 1)
        punpcklwd xmm1,xmm5             ; xmm1=(00 01 10 11 20 21 30 31)
        punpckhwd xmm4,xmm5             ; xmm4=(40 41 50 51 60 61 70 71)
        movdqa    xmm0,xmm7             ; transpose coefficients(phase 1)
        punpcklwd xmm7,xmm6             ; xmm7=(02 03 12 13 22 23 32 33)
        punpckhwd xmm0,xmm6             ; xmm0=(42 43 52 53 62 63 72 73)

        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=col4
        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=col6

        ; xmm5=(04 14 24 34 44 54 64 74), xmm6=(06 16 26 36 46 56 66 76)
        ; xmm2=(05 15 25 35 45 55 65 75), xmm3=(07 17 27 37 47 57 67 77)

        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=(02 03 12 13 22 23 32 33)
        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(42 43 52 53 62 63 72 73)

        movdqa    xmm7,xmm5             ; transpose coefficients(phase 1)
        punpcklwd xmm5,xmm2             ; xmm5=(04 05 14 15 24 25 34 35)
        punpckhwd xmm7,xmm2             ; xmm7=(44 45 54 55 64 65 74 75)
        movdqa    xmm0,xmm6             ; transpose coefficients(phase 1)
        punpcklwd xmm6,xmm3             ; xmm6=(06 07 16 17 26 27 36 37)
        punpckhwd xmm0,xmm3             ; xmm0=(46 47 56 57 66 67 76 77)

        movdqa    xmm2,xmm5             ; transpose coefficients(phase 2)
        punpckldq xmm5,xmm6             ; xmm5=(04 05 06 07 14 15 16 17)
        punpckhdq xmm2,xmm6             ; xmm2=(24 25 26 27 34 35 36 37)
        movdqa    xmm3,xmm7             ; transpose coefficients(phase 2)
        punpckldq xmm7,xmm0             ; xmm7=(44 45 46 47 54 55 56 57)
        punpckhdq xmm3,xmm0             ; xmm3=(64 65 66 67 74 75 76 77)

        movdqa  xmm6, XMMWORD [wk(0)]   ; xmm6=(02 03 12 13 22 23 32 33)
        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(42 43 52 53 62 63 72 73)
        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(24 25 26 27 34 35 36 37)
        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=(44 45 46 47 54 55 56 57)

        movdqa    xmm2,xmm1             ; transpose coefficients(phase 2)
        punpckldq xmm1,xmm6             ; xmm1=(00 01 02 03 10 11 12 13)
        punpckhdq xmm2,xmm6             ; xmm2=(20 21 22 23 30 31 32 33)
        movdqa    xmm7,xmm4             ; transpose coefficients(phase 2)
        punpckldq xmm4,xmm0             ; xmm4=(40 41 42 43 50 51 52 53)
        punpckhdq xmm7,xmm0             ; xmm7=(60 61 62 63 70 71 72 73)

        movdqa     xmm6,xmm1            ; transpose coefficients(phase 3)
        punpcklqdq xmm1,xmm5            ; xmm1=(00 01 02 03 04 05 06 07)=data0
        punpckhqdq xmm6,xmm5            ; xmm6=(10 11 12 13 14 15 16 17)=data1
        movdqa     xmm0,xmm7            ; transpose coefficients(phase 3)
        punpcklqdq xmm7,xmm3            ; xmm7=(60 61 62 63 64 65 66 67)=data6
        punpckhqdq xmm0,xmm3            ; xmm0=(70 71 72 73 74 75 76 77)=data7

        movdqa  xmm5,xmm6
        movdqa  xmm3,xmm1
        psubw   xmm6,xmm7               ; xmm6=data1-data6=tmp6
        psubw   xmm1,xmm0               ; xmm1=data0-data7=tmp7
        paddw   xmm5,xmm7               ; xmm5=data1+data6=tmp1
        paddw   xmm3,xmm0               ; xmm3=data0+data7=tmp0

        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=(24 25 26 27 34 35 36 37)
        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(44 45 46 47 54 55 56 57)
        movdqa  XMMWORD [wk(0)], xmm6   ; wk(0)=tmp6
        movdqa  XMMWORD [wk(1)], xmm1   ; wk(1)=tmp7

        movdqa     xmm6,xmm2            ; transpose coefficients(phase 3)
        punpcklqdq xmm2,xmm7            ; xmm2=(20 21 22 23 24 25 26 27)=data2
        punpckhqdq xmm6,xmm7            ; xmm6=(30 31 32 33 34 35 36 37)=data3
        movdqa     xmm1,xmm4            ; transpose coefficients(phase 3)
        punpcklqdq xmm4,xmm0            ; xmm4=(40 41 42 43 44 45 46 47)=data4
        punpckhqdq xmm1,xmm0            ; xmm1=(50 51 52 53 54 55 56 57)=data5

        movdqa  xmm7,xmm6
        movdqa  xmm0,xmm2
        paddw   xmm6,xmm4               ; xmm6=data3+data4=tmp3
        paddw   xmm2,xmm1               ; xmm2=data2+data5=tmp2
        psubw   xmm7,xmm4               ; xmm7=data3-data4=tmp4
        psubw   xmm0,xmm1               ; xmm0=data2-data5=tmp5

        ; -- Even part

        movdqa  xmm4,xmm3
        movdqa  xmm1,xmm5
        psubw   xmm3,xmm6               ; xmm3=tmp13
        psubw   xmm5,xmm2               ; xmm5=tmp12
        paddw   xmm4,xmm6               ; xmm4=tmp10
        paddw   xmm1,xmm2               ; xmm1=tmp11

        paddw   xmm5,xmm3
        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
        pmulhw  xmm5,[GOTOFF(ebx,PW_F0707)] ; xmm5=z1

        movdqa  xmm6,xmm4
        movdqa  xmm2,xmm3
        psubw   xmm4,xmm1               ; xmm4=data4
        psubw   xmm3,xmm5               ; xmm3=data6
        paddw   xmm6,xmm1               ; xmm6=data0
        paddw   xmm2,xmm5               ; xmm2=data2

        movdqa  XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)], xmm4
        movdqa  XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)], xmm3
        movdqa  XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)], xmm6
        movdqa  XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)], xmm2

        ; -- Odd part

        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=tmp6
        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp7

        paddw   xmm7,xmm0               ; xmm7=tmp10
        paddw   xmm0,xmm1               ; xmm0=tmp11
        paddw   xmm1,xmm5               ; xmm1=tmp12, xmm5=tmp7

        psllw   xmm7,PRE_MULTIPLY_SCALE_BITS
        psllw   xmm1,PRE_MULTIPLY_SCALE_BITS

        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
        pmulhw  xmm0,[GOTOFF(ebx,PW_F0707)] ; xmm0=z3

        movdqa  xmm4,xmm7               ; xmm4=tmp10
        psubw   xmm7,xmm1
        pmulhw  xmm7,[GOTOFF(ebx,PW_F0382)] ; xmm7=z5
        pmulhw  xmm4,[GOTOFF(ebx,PW_F0541)] ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
        pmulhw  xmm1,[GOTOFF(ebx,PW_F1306)] ; xmm1=MULTIPLY(tmp12,FIX_1_306562)
        paddw   xmm4,xmm7               ; xmm4=z2
        paddw   xmm1,xmm7               ; xmm1=z4

        movdqa  xmm3,xmm5
        psubw   xmm5,xmm0               ; xmm5=z13
        paddw   xmm3,xmm0               ; xmm3=z11

        movdqa  xmm6,xmm5
        movdqa  xmm2,xmm3
        psubw   xmm5,xmm4               ; xmm5=data3
        psubw   xmm3,xmm1               ; xmm3=data7
        paddw   xmm6,xmm4               ; xmm6=data5
        paddw   xmm2,xmm1               ; xmm2=data1

        movdqa  XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)], xmm5
        movdqa  XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)], xmm3
        movdqa  XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)], xmm6
        movdqa  XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)], xmm2

;       pop     edi             ; unused
;       pop     esi             ; unused
;       pop     edx             ; need not be preserved
;       pop     ecx             ; unused
        poppic  ebx
        mov     esp,ebp         ; esp <- aligned ebp
        pop     esp             ; esp <- original ebp
        pop     ebp
        ret

; For some reason, the OS X linker does not honor the request to align the
; segment unless we do this.
        align   16