summaryrefslogtreecommitdiff
path: root/media/libjpeg/simd/i386/jidctflt-3dn.asm
blob: 87951910d8e555d6a0f33d3b28b1373b6c56595a (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
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
;
; jidctflt.asm - floating-point IDCT (3DNow! & MMX)
;
; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
; Copyright (C) 2016, 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
;
; This file contains a floating-point implementation of the inverse DCT
; (Discrete Cosine Transform). The following code is based directly on
; the IJG's original jidctflt.c; see the jidctflt.c for more details.

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

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

    alignz      32
    GLOBAL_DATA(jconst_idct_float_3dnow)

EXTN(jconst_idct_float_3dnow):

PD_1_414        times 2 dd 1.414213562373095048801689
PD_1_847        times 2 dd 1.847759065022573512256366
PD_1_082        times 2 dd 1.082392200292393968799446
PD_2_613        times 2 dd 2.613125929752753055713286
PD_RNDINT_MAGIC times 2 dd 100663296.0  ; (float)(0x00C00000 << 3)
PB_CENTERJSAMP  times 8 db CENTERJSAMPLE

    alignz      32

; --------------------------------------------------------------------------
    SECTION     SEG_TEXT
    BITS        32
;
; Perform dequantization and inverse DCT on one block of coefficients.
;
; GLOBAL(void)
; jsimd_idct_float_3dnow(void *dct_table, JCOEFPTR coef_block,
;                        JSAMPARRAY output_buf, JDIMENSION output_col)
;

%define dct_table(b)   (b) + 8          ; void *dct_table
%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
%define output_col(b)  (b) + 20         ; JDIMENSION output_col

%define original_ebp   ebp + 0
%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_MMWORD
                                        ; mmword wk[WK_NUM]
%define WK_NUM         2
%define workspace      wk(0) - DCTSIZE2 * SIZEOF_FAST_FLOAT
                                        ; FAST_FLOAT workspace[DCTSIZE2]

    align       32
    GLOBAL_FUNCTION(jsimd_idct_float_3dnow)

EXTN(jsimd_idct_float_3dnow):
    push        ebp
    mov         eax, esp                    ; eax = original ebp
    sub         esp, byte 4
    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
    mov         [esp], eax
    mov         ebp, esp                    ; ebp = aligned ebp
    lea         esp, [workspace]
    push        ebx
;   push        ecx                     ; need not be preserved
;   push        edx                     ; need not be preserved
    push        esi
    push        edi

    get_GOT     ebx                     ; get GOT address

    ; ---- Pass 1: process columns from input, store into work array.

;   mov         eax, [original_ebp]
    mov         edx, POINTER [dct_table(eax)]    ; quantptr
    mov         esi, JCOEFPTR [coef_block(eax)]  ; inptr
    lea         edi, [workspace]                 ; FAST_FLOAT *wsptr
    mov         ecx, DCTSIZE/2                   ; ctr
    alignx      16, 7
.columnloop:
%ifndef NO_ZERO_COLUMN_TEST_FLOAT_3DNOW
    mov         eax, dword [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
    or          eax, dword [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
    jnz         short .columnDCT

    pushpic     ebx                     ; save GOT address
    mov         ebx, dword [DWBLOCK(3,0,esi,SIZEOF_JCOEF)]
    mov         eax, dword [DWBLOCK(4,0,esi,SIZEOF_JCOEF)]
    or          ebx, dword [DWBLOCK(5,0,esi,SIZEOF_JCOEF)]
    or          eax, dword [DWBLOCK(6,0,esi,SIZEOF_JCOEF)]
    or          ebx, dword [DWBLOCK(7,0,esi,SIZEOF_JCOEF)]
    or          eax, ebx
    poppic      ebx                     ; restore GOT address
    jnz         short .columnDCT

    ; -- AC terms all zero

    movd        mm0, dword [DWBLOCK(0,0,esi,SIZEOF_JCOEF)]

    punpcklwd   mm0, mm0
    psrad       mm0, (DWORD_BIT-WORD_BIT)
    pi2fd       mm0, mm0

    pfmul       mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]

    movq        mm1, mm0
    punpckldq   mm0, mm0
    punpckhdq   mm1, mm1

    movq        MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm0
    movq        MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm0
    movq        MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm0
    movq        MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0
    movq        MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm1
    movq        MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm1
    movq        MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm1
    movq        MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm1
    jmp         near .nextcolumn
    alignx      16, 7
%endif
.columnDCT:

    ; -- Even part

    movd        mm0, dword [DWBLOCK(0,0,esi,SIZEOF_JCOEF)]
    movd        mm1, dword [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
    movd        mm2, dword [DWBLOCK(4,0,esi,SIZEOF_JCOEF)]
    movd        mm3, dword [DWBLOCK(6,0,esi,SIZEOF_JCOEF)]

    punpcklwd   mm0, mm0
    punpcklwd   mm1, mm1
    psrad       mm0, (DWORD_BIT-WORD_BIT)
    psrad       mm1, (DWORD_BIT-WORD_BIT)
    pi2fd       mm0, mm0
    pi2fd       mm1, mm1

    pfmul       mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
    pfmul       mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)]

    punpcklwd   mm2, mm2
    punpcklwd   mm3, mm3
    psrad       mm2, (DWORD_BIT-WORD_BIT)
    psrad       mm3, (DWORD_BIT-WORD_BIT)
    pi2fd       mm2, mm2
    pi2fd       mm3, mm3

    pfmul       mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
    pfmul       mm3, MMWORD [MMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)]

    movq        mm4, mm0
    movq        mm5, mm1
    pfsub       mm0, mm2                ; mm0=tmp11
    pfsub       mm1, mm3
    pfadd       mm4, mm2                ; mm4=tmp10
    pfadd       mm5, mm3                ; mm5=tmp13

    pfmul       mm1, [GOTOFF(ebx,PD_1_414)]
    pfsub       mm1, mm5                ; mm1=tmp12

    movq        mm6, mm4
    movq        mm7, mm0
    pfsub       mm4, mm5                ; mm4=tmp3
    pfsub       mm0, mm1                ; mm0=tmp2
    pfadd       mm6, mm5                ; mm6=tmp0
    pfadd       mm7, mm1                ; mm7=tmp1

    movq        MMWORD [wk(1)], mm4     ; tmp3
    movq        MMWORD [wk(0)], mm0     ; tmp2

    ; -- Odd part

    movd        mm2, dword [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
    movd        mm3, dword [DWBLOCK(3,0,esi,SIZEOF_JCOEF)]
    movd        mm5, dword [DWBLOCK(5,0,esi,SIZEOF_JCOEF)]
    movd        mm1, dword [DWBLOCK(7,0,esi,SIZEOF_JCOEF)]

    punpcklwd   mm2, mm2
    punpcklwd   mm3, mm3
    psrad       mm2, (DWORD_BIT-WORD_BIT)
    psrad       mm3, (DWORD_BIT-WORD_BIT)
    pi2fd       mm2, mm2
    pi2fd       mm3, mm3

    pfmul       mm2, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
    pfmul       mm3, MMWORD [MMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)]

    punpcklwd   mm5, mm5
    punpcklwd   mm1, mm1
    psrad       mm5, (DWORD_BIT-WORD_BIT)
    psrad       mm1, (DWORD_BIT-WORD_BIT)
    pi2fd       mm5, mm5
    pi2fd       mm1, mm1

    pfmul       mm5, MMWORD [MMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
    pfmul       mm1, MMWORD [MMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)]

    movq        mm4, mm2
    movq        mm0, mm5
    pfadd       mm2, mm1                ; mm2=z11
    pfadd       mm5, mm3                ; mm5=z13
    pfsub       mm4, mm1                ; mm4=z12
    pfsub       mm0, mm3                ; mm0=z10

    movq        mm1, mm2
    pfsub       mm2, mm5
    pfadd       mm1, mm5                ; mm1=tmp7

    pfmul       mm2, [GOTOFF(ebx,PD_1_414)]  ; mm2=tmp11

    movq        mm3, mm0
    pfadd       mm0, mm4
    pfmul       mm0, [GOTOFF(ebx,PD_1_847)]  ; mm0=z5
    pfmul       mm3, [GOTOFF(ebx,PD_2_613)]  ; mm3=(z10 * 2.613125930)
    pfmul       mm4, [GOTOFF(ebx,PD_1_082)]  ; mm4=(z12 * 1.082392200)
    pfsubr      mm3, mm0                     ; mm3=tmp12
    pfsub       mm4, mm0                     ; mm4=tmp10

    ; -- Final output stage

    pfsub       mm3, mm1                ; mm3=tmp6
    movq        mm5, mm6
    movq        mm0, mm7
    pfadd       mm6, mm1                ; mm6=data0=(00 01)
    pfadd       mm7, mm3                ; mm7=data1=(10 11)
    pfsub       mm5, mm1                ; mm5=data7=(70 71)
    pfsub       mm0, mm3                ; mm0=data6=(60 61)
    pfsub       mm2, mm3                ; mm2=tmp5

    movq        mm1, mm6                ; transpose coefficients
    punpckldq   mm6, mm7                ; mm6=(00 10)
    punpckhdq   mm1, mm7                ; mm1=(01 11)
    movq        mm3, mm0                ; transpose coefficients
    punpckldq   mm0, mm5                ; mm0=(60 70)
    punpckhdq   mm3, mm5                ; mm3=(61 71)

    movq        MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm6
    movq        MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm1
    movq        MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0
    movq        MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm3

    movq        mm7, MMWORD [wk(0)]     ; mm7=tmp2
    movq        mm5, MMWORD [wk(1)]     ; mm5=tmp3

    pfadd       mm4, mm2                ; mm4=tmp4
    movq        mm6, mm7
    movq        mm1, mm5
    pfadd       mm7, mm2                ; mm7=data2=(20 21)
    pfadd       mm5, mm4                ; mm5=data4=(40 41)
    pfsub       mm6, mm2                ; mm6=data5=(50 51)
    pfsub       mm1, mm4                ; mm1=data3=(30 31)

    movq        mm0, mm7                ; transpose coefficients
    punpckldq   mm7, mm1                ; mm7=(20 30)
    punpckhdq   mm0, mm1                ; mm0=(21 31)
    movq        mm3, mm5                ; transpose coefficients
    punpckldq   mm5, mm6                ; mm5=(40 50)
    punpckhdq   mm3, mm6                ; mm3=(41 51)

    movq        MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm7
    movq        MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm0
    movq        MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm5
    movq        MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm3

.nextcolumn:
    add         esi, byte 2*SIZEOF_JCOEF               ; coef_block
    add         edx, byte 2*SIZEOF_FLOAT_MULT_TYPE     ; quantptr
    add         edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT  ; wsptr
    dec         ecx                                    ; ctr
    jnz         near .columnloop

    ; -- Prefetch the next coefficient block

    prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32]
    prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32]
    prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32]
    prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32]

    ; ---- Pass 2: process rows from work array, store into output array.

    mov         eax, [original_ebp]
    lea         esi, [workspace]                   ; FAST_FLOAT *wsptr
    mov         edi, JSAMPARRAY [output_buf(eax)]  ; (JSAMPROW *)
    mov         eax, JDIMENSION [output_col(eax)]
    mov         ecx, DCTSIZE/2                     ; ctr
    alignx      16, 7
.rowloop:

    ; -- Even part

    movq        mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
    movq        mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)]
    movq        mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)]
    movq        mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)]

    movq        mm4, mm0
    movq        mm5, mm1
    pfsub       mm0, mm2                ; mm0=tmp11
    pfsub       mm1, mm3
    pfadd       mm4, mm2                ; mm4=tmp10
    pfadd       mm5, mm3                ; mm5=tmp13

    pfmul       mm1, [GOTOFF(ebx,PD_1_414)]
    pfsub       mm1, mm5                ; mm1=tmp12

    movq        mm6, mm4
    movq        mm7, mm0
    pfsub       mm4, mm5                ; mm4=tmp3
    pfsub       mm0, mm1                ; mm0=tmp2
    pfadd       mm6, mm5                ; mm6=tmp0
    pfadd       mm7, mm1                ; mm7=tmp1

    movq        MMWORD [wk(1)], mm4     ; tmp3
    movq        MMWORD [wk(0)], mm0     ; tmp2

    ; -- Odd part

    movq        mm2, MMWORD [MMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
    movq        mm3, MMWORD [MMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)]
    movq        mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)]
    movq        mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)]

    movq        mm4, mm2
    movq        mm0, mm5
    pfadd       mm2, mm1                ; mm2=z11
    pfadd       mm5, mm3                ; mm5=z13
    pfsub       mm4, mm1                ; mm4=z12
    pfsub       mm0, mm3                ; mm0=z10

    movq        mm1, mm2
    pfsub       mm2, mm5
    pfadd       mm1, mm5                ; mm1=tmp7

    pfmul       mm2, [GOTOFF(ebx,PD_1_414)]  ; mm2=tmp11

    movq        mm3, mm0
    pfadd       mm0, mm4
    pfmul       mm0, [GOTOFF(ebx,PD_1_847)]  ; mm0=z5
    pfmul       mm3, [GOTOFF(ebx,PD_2_613)]  ; mm3=(z10 * 2.613125930)
    pfmul       mm4, [GOTOFF(ebx,PD_1_082)]  ; mm4=(z12 * 1.082392200)
    pfsubr      mm3, mm0                     ; mm3=tmp12
    pfsub       mm4, mm0                     ; mm4=tmp10

    ; -- Final output stage

    pfsub       mm3, mm1                ; mm3=tmp6
    movq        mm5, mm6
    movq        mm0, mm7
    pfadd       mm6, mm1                ; mm6=data0=(00 10)
    pfadd       mm7, mm3                ; mm7=data1=(01 11)
    pfsub       mm5, mm1                ; mm5=data7=(07 17)
    pfsub       mm0, mm3                ; mm0=data6=(06 16)
    pfsub       mm2, mm3                ; mm2=tmp5

    movq        mm1, [GOTOFF(ebx,PD_RNDINT_MAGIC)]  ; mm1=[PD_RNDINT_MAGIC]
    pcmpeqd     mm3, mm3
    psrld       mm3, WORD_BIT           ; mm3={0xFFFF 0x0000 0xFFFF 0x0000}

    pfadd       mm6, mm1                ; mm6=roundint(data0/8)=(00 ** 10 **)
    pfadd       mm7, mm1                ; mm7=roundint(data1/8)=(01 ** 11 **)
    pfadd       mm0, mm1                ; mm0=roundint(data6/8)=(06 ** 16 **)
    pfadd       mm5, mm1                ; mm5=roundint(data7/8)=(07 ** 17 **)

    pand        mm6, mm3                ; mm6=(00 -- 10 --)
    pslld       mm7, WORD_BIT           ; mm7=(-- 01 -- 11)
    pand        mm0, mm3                ; mm0=(06 -- 16 --)
    pslld       mm5, WORD_BIT           ; mm5=(-- 07 -- 17)
    por         mm6, mm7                ; mm6=(00 01 10 11)
    por         mm0, mm5                ; mm0=(06 07 16 17)

    movq        mm1, MMWORD [wk(0)]     ; mm1=tmp2
    movq        mm3, MMWORD [wk(1)]     ; mm3=tmp3

    pfadd       mm4, mm2                ; mm4=tmp4
    movq        mm7, mm1
    movq        mm5, mm3
    pfadd       mm1, mm2                ; mm1=data2=(02 12)
    pfadd       mm3, mm4                ; mm3=data4=(04 14)
    pfsub       mm7, mm2                ; mm7=data5=(05 15)
    pfsub       mm5, mm4                ; mm5=data3=(03 13)

    movq        mm2, [GOTOFF(ebx,PD_RNDINT_MAGIC)]  ; mm2=[PD_RNDINT_MAGIC]
    pcmpeqd     mm4, mm4
    psrld       mm4, WORD_BIT           ; mm4={0xFFFF 0x0000 0xFFFF 0x0000}

    pfadd       mm3, mm2                ; mm3=roundint(data4/8)=(04 ** 14 **)
    pfadd       mm7, mm2                ; mm7=roundint(data5/8)=(05 ** 15 **)
    pfadd       mm1, mm2                ; mm1=roundint(data2/8)=(02 ** 12 **)
    pfadd       mm5, mm2                ; mm5=roundint(data3/8)=(03 ** 13 **)

    pand        mm3, mm4                ; mm3=(04 -- 14 --)
    pslld       mm7, WORD_BIT           ; mm7=(-- 05 -- 15)
    pand        mm1, mm4                ; mm1=(02 -- 12 --)
    pslld       mm5, WORD_BIT           ; mm5=(-- 03 -- 13)
    por         mm3, mm7                ; mm3=(04 05 14 15)
    por         mm1, mm5                ; mm1=(02 03 12 13)

    movq        mm2, [GOTOFF(ebx,PB_CENTERJSAMP)]  ; mm2=[PB_CENTERJSAMP]

    packsswb    mm6, mm3                ; mm6=(00 01 10 11 04 05 14 15)
    packsswb    mm1, mm0                ; mm1=(02 03 12 13 06 07 16 17)
    paddb       mm6, mm2
    paddb       mm1, mm2

    movq        mm4, mm6                ; transpose coefficients(phase 2)
    punpcklwd   mm6, mm1                ; mm6=(00 01 02 03 10 11 12 13)
    punpckhwd   mm4, mm1                ; mm4=(04 05 06 07 14 15 16 17)

    movq        mm7, mm6                ; transpose coefficients(phase 3)
    punpckldq   mm6, mm4                ; mm6=(00 01 02 03 04 05 06 07)
    punpckhdq   mm7, mm4                ; mm7=(10 11 12 13 14 15 16 17)

    pushpic     ebx                     ; save GOT address

    mov         edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
    mov         ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
    movq        MMWORD [edx+eax*SIZEOF_JSAMPLE], mm6
    movq        MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm7

    poppic      ebx                     ; restore GOT address

    add         esi, byte 2*SIZEOF_FAST_FLOAT  ; wsptr
    add         edi, byte 2*SIZEOF_JSAMPROW
    dec         ecx                            ; ctr
    jnz         near .rowloop

    femms                               ; empty MMX/3DNow! state

    pop         edi
    pop         esi
;   pop         edx                     ; need not be preserved
;   pop         ecx                     ; need not be preserved
    pop         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       32