File: vtgraphicsi.inc

package info (click to toggle)
lazarus 2.0.0%2Bdfsg-2
  • links: PTS, VCS
  • area: main
  • in suites: buster
  • size: 214,460 kB
  • sloc: pascal: 1,862,622; xml: 265,709; cpp: 56,595; sh: 3,008; java: 609; makefile: 535; perl: 297; sql: 222; ansic: 137
file content (725 lines) | stat: -rw-r--r-- 32,045 bytes parent folder | download | duplicates (12)
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
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725

{$ASMMODE INTEL}

procedure AlphaBlendLineConstant(Source, Destination: Pointer; Count: Integer; ConstantAlpha, Bias: Integer);

// Blends a line of Count pixels from Source to Destination using a constant alpha value.
// The layout of a pixel must be BGRA where A is ignored (but is calculated as the other components).
// ConstantAlpha must be in the range 0..255 where 0 means totally transparent (destination pixel only)
// and 255 totally opaque (source pixel only).
// Bias is an additional value which gets added to every component and must be in the range -128..127

asm

{$ifdef CPU64}
// RCX contains Source
// RDX contains Destination
// R8D contains Count
// R9D contains ConstantAlpha
// Bias is on the stack

        //.NOFRAME

        // Load XMM3 with the constant alpha value (replicate it for every component).
        // Expand it to word size.
        MOVD        XMM3, R9D  // ConstantAlpha
        PUNPCKLWD   XMM3, XMM3
        PUNPCKLDQ   XMM3, XMM3

        // Load XMM5 with the bias value.
        MOVD        XMM5, [Bias]
        PUNPCKLWD   XMM5, XMM5
        PUNPCKLDQ   XMM5, XMM5

        // Load XMM4 with 128 to allow for saturated biasing.
        MOV         R10D, 128
        MOVD        XMM4, R10D
        PUNPCKLWD   XMM4, XMM4
        PUNPCKLDQ   XMM4, XMM4

@1:     // The pixel loop calculates an entire pixel in one run.
        // Note: The pixel byte values are expanded into the higher bytes of a word due
        //       to the way unpacking works. We compensate for this with an extra shift.
        MOVD        XMM1, DWORD PTR [RCX]   // data is unaligned
        MOVD        XMM2, DWORD PTR [RDX]   // data is unaligned
        PXOR        XMM0, XMM0    // clear source pixel register for unpacking
        PUNPCKLBW   XMM0, XMM1{[RCX]}    // unpack source pixel byte values into words
        PSRLW       XMM0, 8       // move higher bytes to lower bytes
        PXOR        XMM1, XMM1    // clear target pixel register for unpacking
        PUNPCKLBW   XMM1, XMM2{[RDX]}    // unpack target pixel byte values into words
        MOVQ        XMM2, XMM1    // make a copy of the shifted values, we need them again
        PSRLW       XMM1, 8       // move higher bytes to lower bytes

        // calculation is: target = (alpha * (source - target) + 256 * target) / 256
        PSUBW       XMM0, XMM1    // source - target
        PMULLW      XMM0, XMM3    // alpha * (source - target)
        PADDW       XMM0, XMM2    // add target (in shifted form)
        PSRLW       XMM0, 8       // divide by 256

        // Bias is accounted for by conversion of range 0..255 to -128..127,
        // doing a saturated add and convert back to 0..255.
        PSUBW     XMM0, XMM4
        PADDSW    XMM0, XMM5
        PADDW     XMM0, XMM4
        PACKUSWB  XMM0, XMM0      // convert words to bytes with saturation
        MOVD      DWORD PTR [RDX], XMM0     // store the result
@3:
        ADD       RCX, 4
        ADD       RDX, 4
        DEC       R8D
        JNZ       @1


{$else}
// EAX contains Source
// EDX contains Destination
// ECX contains Count
// ConstantAlpha and Bias are on the stack


        PUSH    ESI                    // save used registers
        PUSH    EDI

        MOV     ESI, EAX               // ESI becomes the actual source pointer
        MOV     EDI, EDX               // EDI becomes the actual target pointer

        // Load MM6 with the constant alpha value (replicate it for every component).
        // Expand it to word size.
        MOV     EAX, [ConstantAlpha]
        DB      $0F, $6E, $F0          /// MOVD      MM6, EAX
        DB      $0F, $61, $F6          /// PUNPCKLWD MM6, MM6
        DB      $0F, $62, $F6          /// PUNPCKLDQ MM6, MM6

        // Load MM5 with the bias value.
        MOV     EAX, [Bias]
        DB      $0F, $6E, $E8          /// MOVD      MM5, EAX
        DB      $0F, $61, $ED          /// PUNPCKLWD MM5, MM5
        DB      $0F, $62, $ED          /// PUNPCKLDQ MM5, MM5

        // Load MM4 with 128 to allow for saturated biasing.
        MOV     EAX, 128
        DB      $0F, $6E, $E0          /// MOVD      MM4, EAX
        DB      $0F, $61, $E4          /// PUNPCKLWD MM4, MM4
        DB      $0F, $62, $E4          /// PUNPCKLDQ MM4, MM4

@1:     // The pixel loop calculates an entire pixel in one run.
        // Note: The pixel byte values are expanded into the higher bytes of a word due
        //       to the way unpacking works. We compensate for this with an extra shift.
        DB      $0F, $EF, $C0          /// PXOR      MM0, MM0,   clear source pixel register for unpacking
        DB      $0F, $60, $06          /// PUNPCKLBW MM0, [ESI], unpack source pixel byte values into words
        DB      $0F, $71, $D0, $08     /// PSRLW     MM0, 8,     move higher bytes to lower bytes
        DB      $0F, $EF, $C9          /// PXOR      MM1, MM1,   clear target pixel register for unpacking
        DB      $0F, $60, $0F          /// PUNPCKLBW MM1, [EDI], unpack target pixel byte values into words
        DB      $0F, $6F, $D1          /// MOVQ      MM2, MM1,   make a copy of the shifted values, we need them again
        DB      $0F, $71, $D1, $08     /// PSRLW     MM1, 8,     move higher bytes to lower bytes

        // calculation is: target = (alpha * (source - target) + 256 * target) / 256
        DB      $0F, $F9, $C1          /// PSUBW     MM0, MM1,   source - target
        DB      $0F, $D5, $C6          /// PMULLW    MM0, MM6,   alpha * (source - target)
        DB      $0F, $FD, $C2          /// PADDW     MM0, MM2,   add target (in shifted form)
        DB      $0F, $71, $D0, $08     /// PSRLW     MM0, 8,     divide by 256

        // Bias is accounted for by conversion of range 0..255 to -128..127,
        // doing a saturated add and convert back to 0..255.
        DB      $0F, $F9, $C4          /// PSUBW     MM0, MM4
        DB      $0F, $ED, $C5          /// PADDSW    MM0, MM5
        DB      $0F, $FD, $C4          /// PADDW     MM0, MM4
        DB      $0F, $67, $C0          /// PACKUSWB  MM0, MM0,   convert words to bytes with saturation
        DB      $0F, $7E, $07          /// MOVD      [EDI], MM0, store the result
@3:
        ADD     ESI, 4
        ADD     EDI, 4
        DEC     ECX
        JNZ     @1
        POP     EDI
        POP     ESI
{$endif}
end;

//----------------------------------------------------------------------------------------------------------------------

procedure AlphaBlendLinePerPixel(Source, Destination: Pointer; Count, Bias: Integer);

// Blends a line of Count pixels from Source to Destination using the alpha value of the source pixels.
// The layout of a pixel must be BGRA.
// Bias is an additional value which gets added to every component and must be in the range -128..127

asm

{$ifdef CPU64}
// RCX contains Source
// RDX contains Destination
// R8D contains Count
// R9D contains Bias

        //.NOFRAME

        // Load XMM5 with the bias value.
        MOVD        XMM5, R9D   // Bias
        PUNPCKLWD   XMM5, XMM5
        PUNPCKLDQ   XMM5, XMM5

        // Load XMM4 with 128 to allow for saturated biasing.
        MOV         R10D, 128
        MOVD        XMM4, R10D
        PUNPCKLWD   XMM4, XMM4
        PUNPCKLDQ   XMM4, XMM4

@1:     // The pixel loop calculates an entire pixel in one run.
        // Note: The pixel byte values are expanded into the higher bytes of a word due
        //       to the way unpacking works. We compensate for this with an extra shift.
        MOVD        XMM1, DWORD PTR [RCX]   // data is unaligned
        MOVD        XMM2, DWORD PTR [RDX]   // data is unaligned
        PXOR        XMM0, XMM0    // clear source pixel register for unpacking
        PUNPCKLBW   XMM0, XMM1{[RCX]}    // unpack source pixel byte values into words
        PSRLW       XMM0, 8       // move higher bytes to lower bytes
        PXOR        XMM1, XMM1    // clear target pixel register for unpacking
        PUNPCKLBW   XMM1, XMM2{[RDX]}    // unpack target pixel byte values into words
        MOVQ        XMM2, XMM1    // make a copy of the shifted values, we need them again
        PSRLW       XMM1, 8       // move higher bytes to lower bytes

        // Load XMM3 with the source alpha value (replicate it for every component).
        // Expand it to word size.
        MOVQ        XMM3, XMM0
        PUNPCKHWD   XMM3, XMM3
        PUNPCKHDQ   XMM3, XMM3

        // calculation is: target = (alpha * (source - target) + 256 * target) / 256
        PSUBW       XMM0, XMM1    // source - target
        PMULLW      XMM0, XMM3    // alpha * (source - target)
        PADDW       XMM0, XMM2    // add target (in shifted form)
        PSRLW       XMM0, 8       // divide by 256

        // Bias is accounted for by conversion of range 0..255 to -128..127,
        // doing a saturated add and convert back to 0..255.
        PSUBW       XMM0, XMM4
        PADDSW      XMM0, XMM5
        PADDW       XMM0, XMM4
        PACKUSWB    XMM0, XMM0    // convert words to bytes with saturation
        MOVD        DWORD PTR [RDX], XMM0   // store the result
@3:
        ADD         RCX, 4
        ADD         RDX, 4
        DEC         R8D
        JNZ         @1


{$else}

// EAX contains Source
// EDX contains Destination
// ECX contains Count
// Bias is on the stack

        PUSH    ESI                    // save used registers
        PUSH    EDI

        MOV     ESI, EAX               // ESI becomes the actual source pointer
        MOV     EDI, EDX               // EDI becomes the actual target pointer

        // Load MM5 with the bias value.
        MOV     EAX, [Bias]
        DB      $0F, $6E, $E8          /// MOVD      MM5, EAX
        DB      $0F, $61, $ED          /// PUNPCKLWD MM5, MM5
        DB      $0F, $62, $ED          /// PUNPCKLDQ MM5, MM5

        // Load MM4 with 128 to allow for saturated biasing.
        MOV     EAX, 128
        DB      $0F, $6E, $E0          /// MOVD      MM4, EAX
        DB      $0F, $61, $E4          /// PUNPCKLWD MM4, MM4
        DB      $0F, $62, $E4          /// PUNPCKLDQ MM4, MM4

@1:     // The pixel loop calculates an entire pixel in one run.
        // Note: The pixel byte values are expanded into the higher bytes of a word due
        //       to the way unpacking works. We compensate for this with an extra shift.
        DB      $0F, $EF, $C0          /// PXOR      MM0, MM0,   clear source pixel register for unpacking
        DB      $0F, $60, $06          /// PUNPCKLBW MM0, [ESI], unpack source pixel byte values into words
        DB      $0F, $71, $D0, $08     /// PSRLW     MM0, 8,     move higher bytes to lower bytes
        DB      $0F, $EF, $C9          /// PXOR      MM1, MM1,   clear target pixel register for unpacking
        DB      $0F, $60, $0F          /// PUNPCKLBW MM1, [EDI], unpack target pixel byte values into words
        DB      $0F, $6F, $D1          /// MOVQ      MM2, MM1,   make a copy of the shifted values, we need them again
        DB      $0F, $71, $D1, $08     /// PSRLW     MM1, 8,     move higher bytes to lower bytes

        // Load MM6 with the source alpha value (replicate it for every component).
        // Expand it to word size.
        DB      $0F, $6F, $F0          /// MOVQ MM6, MM0
        DB      $0F, $69, $F6          /// PUNPCKHWD MM6, MM6
        DB      $0F, $6A, $F6          /// PUNPCKHDQ MM6, MM6

        // calculation is: target = (alpha * (source - target) + 256 * target) / 256
        DB      $0F, $F9, $C1          /// PSUBW     MM0, MM1,   source - target
        DB      $0F, $D5, $C6          /// PMULLW    MM0, MM6,   alpha * (source - target)
        DB      $0F, $FD, $C2          /// PADDW     MM0, MM2,   add target (in shifted form)
        DB      $0F, $71, $D0, $08     /// PSRLW     MM0, 8,     divide by 256

        // Bias is accounted for by conversion of range 0..255 to -128..127,
        // doing a saturated add and convert back to 0..255.
        DB      $0F, $F9, $C4          /// PSUBW     MM0, MM4
        DB      $0F, $ED, $C5          /// PADDSW    MM0, MM5
        DB      $0F, $FD, $C4          /// PADDW     MM0, MM4
        DB      $0F, $67, $C0          /// PACKUSWB  MM0, MM0,   convert words to bytes with saturation
        DB      $0F, $7E, $07          /// MOVD      [EDI], MM0, store the result
@3:
        ADD     ESI, 4
        ADD     EDI, 4
        DEC     ECX
        JNZ     @1
        POP     EDI
        POP     ESI
{$endif}
end;

//----------------------------------------------------------------------------------------------------------------------

procedure AlphaBlendLineMaster(Source, Destination: Pointer; Count: Integer; ConstantAlpha, Bias: Integer);

// Blends a line of Count pixels from Source to Destination using the source pixel and a constant alpha value.
// The layout of a pixel must be BGRA.
// ConstantAlpha must be in the range 0..255.
// Bias is an additional value which gets added to every component and must be in the range -128..127

asm

{$ifdef CPU64}
// RCX contains Source
// RDX contains Destination
// R8D contains Count
// R9D contains ConstantAlpha
// Bias is on the stack

        //.SAVENV XMM6  //todo see how implement in fpc      AlphaBlendLineMaster

        // Load XMM3 with the constant alpha value (replicate it for every component).
        // Expand it to word size.
        MOVD        XMM3, R9D    // ConstantAlpha
        PUNPCKLWD   XMM3, XMM3
        PUNPCKLDQ   XMM3, XMM3

        // Load XMM5 with the bias value.
        MOV         R10D, [Bias]
        MOVD        XMM5, R10D
        PUNPCKLWD   XMM5, XMM5
        PUNPCKLDQ   XMM5, XMM5

        // Load XMM4 with 128 to allow for saturated biasing.
        MOV         R10D, 128
        MOVD        XMM4, R10D
        PUNPCKLWD   XMM4, XMM4
        PUNPCKLDQ   XMM4, XMM4

@1:     // The pixel loop calculates an entire pixel in one run.
        // Note: The pixel byte values are expanded into the higher bytes of a word due
        //       to the way unpacking works. We compensate for this with an extra shift.
        MOVD        XMM1, DWORD PTR [RCX]   // data is unaligned
        MOVD        XMM2, DWORD PTR [RDX]   // data is unaligned
        PXOR        XMM0, XMM0    // clear source pixel register for unpacking
        PUNPCKLBW   XMM0, XMM1{[RCX]}     // unpack source pixel byte values into words
        PSRLW       XMM0, 8       // move higher bytes to lower bytes
        PXOR        XMM1, XMM1    // clear target pixel register for unpacking
        PUNPCKLBW   XMM1, XMM2{[RCX]}     // unpack target pixel byte values into words
        MOVQ        XMM2, XMM1    // make a copy of the shifted values, we need them again
        PSRLW       XMM1, 8       // move higher bytes to lower bytes

        // Load XMM6 with the source alpha value (replicate it for every component).
        // Expand it to word size.
        MOVQ        XMM6, XMM0
        PUNPCKHWD   XMM6, XMM6
        PUNPCKHDQ   XMM6, XMM6
        PMULLW      XMM6, XMM3    // source alpha * master alpha
        PSRLW       XMM6, 8       // divide by 256

        // calculation is: target = (alpha * master alpha * (source - target) + 256 * target) / 256
        PSUBW       XMM0, XMM1    // source - target
        PMULLW      XMM0, XMM6    // alpha * (source - target)
        PADDW       XMM0, XMM2    // add target (in shifted form)
        PSRLW       XMM0, 8       // divide by 256

        // Bias is accounted for by conversion of range 0..255 to -128..127,
        // doing a saturated add and convert back to 0..255.
        PSUBW       XMM0, XMM4
        PADDSW      XMM0, XMM5
        PADDW       XMM0, XMM4
        PACKUSWB    XMM0, XMM0    // convert words to bytes with saturation
        MOVD        DWORD PTR [RDX], XMM0   // store the result
@3:
        ADD         RCX, 4
        ADD         RDX, 4
        DEC         R8D
        JNZ         @1

{$else}

// EAX contains Source
// EDX contains Destination
// ECX contains Count
// ConstantAlpha and Bias are on the stack


        PUSH    ESI                    // save used registers
        PUSH    EDI

        MOV     ESI, EAX               // ESI becomes the actual source pointer
        MOV     EDI, EDX               // EDI becomes the actual target pointer

        // Load MM6 with the constant alpha value (replicate it for every component).
        // Expand it to word size.
        MOV     EAX, [ConstantAlpha]
        DB      $0F, $6E, $F0          /// MOVD      MM6, EAX
        DB      $0F, $61, $F6          /// PUNPCKLWD MM6, MM6
        DB      $0F, $62, $F6          /// PUNPCKLDQ MM6, MM6

        // Load MM5 with the bias value.
        MOV     EAX, [Bias]
        DB      $0F, $6E, $E8          /// MOVD      MM5, EAX
        DB      $0F, $61, $ED          /// PUNPCKLWD MM5, MM5
        DB      $0F, $62, $ED          /// PUNPCKLDQ MM5, MM5

        // Load MM4 with 128 to allow for saturated biasing.
        MOV     EAX, 128
        DB      $0F, $6E, $E0          /// MOVD      MM4, EAX
        DB      $0F, $61, $E4          /// PUNPCKLWD MM4, MM4
        DB      $0F, $62, $E4          /// PUNPCKLDQ MM4, MM4

@1:     // The pixel loop calculates an entire pixel in one run.
        // Note: The pixel byte values are expanded into the higher bytes of a word due
        //       to the way unpacking works. We compensate for this with an extra shift.
        DB      $0F, $EF, $C0          /// PXOR      MM0, MM0,   clear source pixel register for unpacking
        DB      $0F, $60, $06          /// PUNPCKLBW MM0, [ESI], unpack source pixel byte values into words
        DB      $0F, $71, $D0, $08     /// PSRLW     MM0, 8,     move higher bytes to lower bytes
        DB      $0F, $EF, $C9          /// PXOR      MM1, MM1,   clear target pixel register for unpacking
        DB      $0F, $60, $0F          /// PUNPCKLBW MM1, [EDI], unpack target pixel byte values into words
        DB      $0F, $6F, $D1          /// MOVQ      MM2, MM1,   make a copy of the shifted values, we need them again
        DB      $0F, $71, $D1, $08     /// PSRLW     MM1, 8,     move higher bytes to lower bytes

        // Load MM7 with the source alpha value (replicate it for every component).
        // Expand it to word size.
        DB      $0F, $6F, $F8          /// MOVQ      MM7, MM0
        DB      $0F, $69, $FF          /// PUNPCKHWD MM7, MM7
        DB      $0F, $6A, $FF          /// PUNPCKHDQ MM7, MM7
        DB      $0F, $D5, $FE          /// PMULLW    MM7, MM6,   source alpha * master alpha
        DB      $0F, $71, $D7, $08     /// PSRLW     MM7, 8,     divide by 256

        // calculation is: target = (alpha * master alpha * (source - target) + 256 * target) / 256
        DB      $0F, $F9, $C1          /// PSUBW     MM0, MM1,   source - target
        DB      $0F, $D5, $C7          /// PMULLW    MM0, MM7,   alpha * (source - target)
        DB      $0F, $FD, $C2          /// PADDW     MM0, MM2,   add target (in shifted form)
        DB      $0F, $71, $D0, $08     /// PSRLW     MM0, 8,     divide by 256

        // Bias is accounted for by conversion of range 0..255 to -128..127,
        // doing a saturated add and convert back to 0..255.
        DB      $0F, $F9, $C4          /// PSUBW     MM0, MM4
        DB      $0F, $ED, $C5          /// PADDSW    MM0, MM5
        DB      $0F, $FD, $C4          /// PADDW     MM0, MM4
        DB      $0F, $67, $C0          /// PACKUSWB  MM0, MM0,   convert words to bytes with saturation
        DB      $0F, $7E, $07          /// MOVD      [EDI], MM0, store the result
@3:
        ADD     ESI, 4
        ADD     EDI, 4
        DEC     ECX
        JNZ     @1
        POP     EDI
        POP     ESI
{$endif}
end;

//----------------------------------------------------------------------------------------------------------------------

procedure AlphaBlendLineMasterAndColor(Destination: Pointer; Count: Integer; ConstantAlpha, Color: Integer);

// Blends a line of Count pixels in Destination against the given color using a constant alpha value.
// The layout of a pixel must be BGRA and Color must be rrggbb00 (as stored by a COLORREF).
// ConstantAlpha must be in the range 0..255.

asm

{$ifdef CPU64}
// RCX contains Destination
// EDX contains Count
// R8D contains ConstantAlpha
// R9D contains Color
        //.NOFRAME

        // The used formula is: target = (alpha * color + (256 - alpha) * target) / 256.
        // alpha * color (factor 1) and 256 - alpha (factor 2) are constant values which can be calculated in advance.
        // The remaining calculation is therefore: target = (F1 + F2 * target) / 256

        // Load XMM3 with the constant alpha value (replicate it for every component).
        // Expand it to word size. (Every calculation here works on word sized operands.)
        MOVD        XMM3, R8D   // ConstantAlpha
        PUNPCKLWD   XMM3, XMM3
        PUNPCKLDQ   XMM3, XMM3

        // Calculate factor 2.
        MOV         R10D, $100
        MOVD        XMM2, R10D
        PUNPCKLWD   XMM2, XMM2
        PUNPCKLDQ   XMM2, XMM2
        PSUBW       XMM2, XMM3             // XMM2 contains now: 255 - alpha = F2

        // Now calculate factor 1. Alpha is still in XMM3, but the r and b components of Color must be swapped.
        BSWAP       R9D  // Color
        ROR         R9D, 8
        MOVD        XMM1, R9D              // Load the color and convert to word sized values.
        PXOR        XMM4, XMM4
        PUNPCKLBW   XMM1, XMM4
        PMULLW      XMM1, XMM3             // XMM1 contains now: color * alpha = F1

@1:     // The pixel loop calculates an entire pixel in one run.
        MOVD        XMM0, DWORD PTR [RCX]
        PUNPCKLBW   XMM0, XMM4

        PMULLW      XMM0, XMM2             // calculate F1 + F2 * target
        PADDW       XMM0, XMM1
        PSRLW       XMM0, 8                // divide by 256

        PACKUSWB    XMM0, XMM0             // convert words to bytes with saturation
        MOVD        DWORD PTR [RCX], XMM0            // store the result

        ADD         RCX, 4
        DEC         EDX
        JNZ         @1


{$else}

// EAX contains Destination
// EDX contains Count
// ECX contains ConstantAlpha
// Color is passed on the stack


        // The used formula is: target = (alpha * color + (256 - alpha) * target) / 256.
        // alpha * color (factor 1) and 256 - alpha (factor 2) are constant values which can be calculated in advance.
        // The remaining calculation is therefore: target = (F1 + F2 * target) / 256

        // Load MM3 with the constant alpha value (replicate it for every component).
        // Expand it to word size. (Every calculation here works on word sized operands.)
        DB      $0F, $6E, $D9          /// MOVD      MM3, ECX
        DB      $0F, $61, $DB          /// PUNPCKLWD MM3, MM3
        DB      $0F, $62, $DB          /// PUNPCKLDQ MM3, MM3

        // Calculate factor 2.
        MOV     ECX, $100
        DB      $0F, $6E, $D1          /// MOVD      MM2, ECX
        DB      $0F, $61, $D2          /// PUNPCKLWD MM2, MM2
        DB      $0F, $62, $D2          /// PUNPCKLDQ MM2, MM2
        DB      $0F, $F9, $D3          /// PSUBW     MM2, MM3             // MM2 contains now: 255 - alpha = F2

        // Now calculate factor 1. Alpha is still in MM3, but the r and b components of Color must be swapped.
        MOV     ECX, [Color]
        BSWAP   ECX
        ROR     ECX, 8
        DB      $0F, $6E, $C9          /// MOVD      MM1, ECX             // Load the color and convert to word sized values.
        DB      $0F, $EF, $E4          /// PXOR      MM4, MM4
        DB      $0F, $60, $CC          /// PUNPCKLBW MM1, MM4
        DB      $0F, $D5, $CB          /// PMULLW    MM1, MM3             // MM1 contains now: color * alpha = F1

@1:     // The pixel loop calculates an entire pixel in one run.
        DB      $0F, $6E, $00          /// MOVD      MM0, [EAX]
        DB      $0F, $60, $C4          /// PUNPCKLBW MM0, MM4

        DB      $0F, $D5, $C2          /// PMULLW    MM0, MM2             // calculate F1 + F2 * target
        DB      $0F, $FD, $C1          /// PADDW     MM0, MM1
        DB      $0F, $71, $D0, $08     /// PSRLW     MM0, 8               // divide by 256

        DB      $0F, $67, $C0          /// PACKUSWB  MM0, MM0             // convert words to bytes with saturation
        DB      $0F, $7E, $00          /// MOVD      [EAX], MM0           // store the result

        ADD     EAX, 4
        DEC     EDX
        JNZ     @1
{$endif}
end;

//----------------------------------------------------------------------------------------------------------------------

procedure EMMS;

// Reset MMX state to use the FPU for other tasks again.

{$ifdef CPU64}
  inline;
  begin
  end;

{$else}

asm
        DB      $0F, $77               /// EMMS
end;
{$endif}

//----------------------------------------------------------------------------------------------------------------------

function GetBitmapBitsFromDeviceContext(DC: HDC; out Width, Height: Integer): Pointer;

// Helper function used to retrieve the bitmap selected into the given device context. If there is a bitmap then
// the function will return a pointer to its bits otherwise nil is returned.
// Additionally the dimensions of the bitmap are returned.

var
  Bitmap: HBITMAP;
  DIB: TDIBSection;

begin
  Result := nil;
  Width := 0;
  Height := 0;
  Bitmap := GetCurrentObject(DC, OBJ_BITMAP);
  if Bitmap <> 0 then
  begin
    if GetObject(Bitmap, SizeOf(DIB), @DIB) = SizeOf(DIB) then
    begin
      Assert(DIB.dsBm.bmPlanes * DIB.dsBm.bmBitsPixel = 32, 'Alpha blending error: bitmap must use 32 bpp.');
      Result := DIB.dsBm.bmBits;
      Width := DIB.dsBmih.biWidth;
      Height := DIB.dsBmih.biHeight;
    end;
  end;
  Assert(Result <> nil, 'Alpha blending DC error: no bitmap available.');
end;

//----------------------------------------------------------------------------------------------------------------------

function GetBitmapBitsFromBitmap(Bitmap: HBITMAP): Pointer;
var
  DIB: TDIBSection;
begin
  Result := nil;
  if Bitmap <> 0 then
  begin
    if GetObject(Bitmap, SizeOf(DIB), @DIB) = SizeOf(DIB) then
    begin
      Assert(DIB.dsBm.bmPlanes * DIB.dsBm.bmBitsPixel = 32, 'Alpha blending error: bitmap must use 32 bpp.');
      Result := DIB.dsBm.bmBits;
    end;
  end;
end;

function CalculateScanline(Bits: Pointer; Width, Height, Row: Integer): Pointer;

// Helper function to calculate the start address for the given row.

begin
  //todo: Height is always > 0 in LCL
  {
  if Height > 0 then  // bottom-up DIB
    Row := Height - Row - 1;
  }
  // Return DWORD aligned address of the requested scanline.
  Result := Bits + Row * ((Width * 32 + 31) and not 31) div 8;
end;

//----------------------------------------------------------------------------------------------------------------------

procedure AlphaBlend(Source, Destination: HDC; const R: TRect; const Target: TPoint; Mode: TBlendMode; ConstantAlpha, Bias: Integer);

// Optimized alpha blend procedure using MMX instructions to perform as quick as possible.
// For this procedure to work properly it is important that both source and target bitmap use the 32 bit color format.
// R describes the source rectangle to work on.
// Target is the place (upper left corner) in the target bitmap where to blend to. Note that source width + X offset
// must be less or equal to the target width. Similar for the height.
// If Mode is bmConstantAlpha then the blend operation uses the given ConstantAlpha value for all pixels.
// If Mode is bmPerPixelAlpha then each pixel is blended using its individual alpha value (the alpha value of the source).
// If Mode is bmMasterAlpha then each pixel is blended using its individual alpha value multiplied by ConstantAlpha.
// If Mode is bmConstantAlphaAndColor then each destination pixel is blended using ConstantAlpha but also a constant
// color which will be obtained from Bias. In this case no offset value is added, otherwise Bias is used as offset.
// Blending of a color into target only (bmConstantAlphaAndColor) ignores Source (the DC) and Target (the position).
// CAUTION: This procedure does not check whether MMX instructions are actually available! Call it only if MMX is really
//          usable.

var
  Y: Integer;
  SourceRun,
  TargetRun: PByte;

  SourceBits,
  DestBits: Pointer;
  SourceWidth,
  SourceHeight,
  DestWidth,
  DestHeight: Integer;

begin
  if not IsRectEmpty(R) then
  begin
    {$ifdef CPU64}
    //avoid MasterAlpha due to incomplete AlphaBlendLineMaster. See comment in procedure
    if Mode = bmMasterAlpha then
      Mode := bmConstantAlpha;
    {$endif}
    // Note: it is tempting to optimize the special cases for constant alpha 0 and 255 by just ignoring soure
    //       (alpha = 0) or simply do a blit (alpha = 255). But this does not take the bias into account.
    case Mode of
      bmConstantAlpha:
        begin
          // Get a pointer to the bitmap bits for the source and target device contexts.
          // Note: this supposes that both contexts do actually have bitmaps assigned!
          SourceBits := GetBitmapBitsFromDeviceContext(Source, SourceWidth, SourceHeight);
          DestBits := GetBitmapBitsFromDeviceContext(Destination, DestWidth, DestHeight);
          if Assigned(SourceBits) and Assigned(DestBits) then
          begin
            for Y := 0 to R.Bottom - R.Top - 1 do
            begin
              SourceRun := CalculateScanline(SourceBits, SourceWidth, SourceHeight, Y + R.Top);
              Inc(SourceRun, 4 * R.Left);
              TargetRun := CalculateScanline(DestBits, DestWidth, DestHeight, Y + Target.Y);
              Inc(TargetRun, 4 * Target.X);
              AlphaBlendLineConstant(SourceRun, TargetRun, R.Right - R.Left, ConstantAlpha, Bias);
            end;
          end;
          EMMS;
        end;
      bmPerPixelAlpha:
        begin
          SourceBits := GetBitmapBitsFromDeviceContext(Source, SourceWidth, SourceHeight);
          DestBits := GetBitmapBitsFromDeviceContext(Destination, DestWidth, DestHeight);
          if Assigned(SourceBits) and Assigned(DestBits) then
          begin
            for Y := 0 to R.Bottom - R.Top - 1 do
            begin
              SourceRun := CalculateScanline(SourceBits, SourceWidth, SourceHeight, Y + R.Top);
              Inc(SourceRun, 4 * R.Left);
              TargetRun := CalculateScanline(DestBits, DestWidth, DestHeight, Y + Target.Y);
              Inc(TargetRun, 4 * Target.X);
              AlphaBlendLinePerPixel(SourceRun, TargetRun, R.Right - R.Left, Bias);
            end;
          end;
          EMMS;
        end;
      bmMasterAlpha:
        begin
          SourceBits := GetBitmapBitsFromDeviceContext(Source, SourceWidth, SourceHeight);
          DestBits := GetBitmapBitsFromDeviceContext(Destination, DestWidth, DestHeight);
          if Assigned(SourceBits) and Assigned(DestBits) then
          begin
            for Y := 0 to R.Bottom - R.Top - 1 do
            begin
              SourceRun := CalculateScanline(SourceBits, SourceWidth, SourceHeight, Y + R.Top);
              Inc(SourceRun, 4 * Target.X);
              TargetRun := CalculateScanline(DestBits, DestWidth, DestHeight, Y + Target.Y);
              AlphaBlendLineMaster(SourceRun, TargetRun, R.Right - R.Left, ConstantAlpha, Bias);
            end;
          end;
          EMMS;
        end;
      bmConstantAlphaAndColor:
        begin
          // Source is ignored since there is a constant color value.
          DestBits := GetBitmapBitsFromDeviceContext(Destination, DestWidth, DestHeight);
          if Assigned(DestBits) then
          begin
            for Y := 0 to R.Bottom - R.Top - 1 do
            begin
              TargetRun := CalculateScanline(DestBits, DestWidth, DestHeight, Y + R.Top);
              Inc(TargetRun, 4 * R.Left);
              AlphaBlendLineMasterAndColor(TargetRun, R.Right - R.Left, ConstantAlpha, Bias);
            end;
          end;
          EMMS;
        end;
    end;
  end;
end;