libtoa 0.4.0

TOA compression file format library. Modern compression with built-in error correction
Documentation
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
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
#[cfg(target_arch = "x86_64")]
use crate::reed_solomon::get_generator_poly;
use crate::{
    ErrorCorrection, Result, SimdOverride, Write,
    circular_buffer::CircularBuffer,
    reed_solomon::{code_255_191, code_255_223, code_255_239},
};

#[cfg(target_arch = "x86_64")]
const ECC_BATCH_SIZE_AVX: usize = 32;

#[cfg(target_arch = "x86_64")]
const ECC_BATCH_SIZE_SSE: usize = 16;

#[cfg(target_arch = "aarch64")]
const ECC_BATCH_SIZE_NEON: usize = 16;

type EncodeFunction<W> = fn(&mut ECCEncoder<W>, &[u8]) -> Result<usize>;

fn encode_none<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    encoder.inner.write_all(data)?;
    Ok(data.len())
}

fn encode_standard<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    encoder.encode_scalar::<_, 239, 16>(data, code_255_239::encode)
}

fn encode_paranoid<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    encoder.encode_scalar::<_, 223, 32>(data, code_255_223::encode)
}

fn encode_extreme<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    encoder.encode_scalar::<_, 191, 64>(data, code_255_191::encode)
}

#[cfg(target_arch = "x86_64")]
fn encode_standard_sse2_gfni<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 16, 239, 16>(data, |writer, batch_codewords| {
            encode_simd_batch_sse2_gfni::<_, 16, 239, 16>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "x86_64")]
fn encode_paranoid_sse2_gfni<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 16, 223, 32>(data, |writer, batch_codewords| {
            encode_simd_batch_sse2_gfni::<_, 16, 223, 32>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "x86_64")]
fn encode_extreme_sse2_gfni<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 16, 191, 64>(data, |writer, batch_codewords| {
            encode_simd_batch_sse2_gfni::<_, 16, 191, 64>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "x86_64")]
fn encode_standard_ssse3<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 16, 239, 16>(data, |writer, batch_codewords| {
            encode_simd_batch_ssse3::<_, 16, 239, 16>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "x86_64")]
fn encode_paranoid_ssse3<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 16, 223, 32>(data, |writer, batch_codewords| {
            encode_simd_batch_ssse3::<_, 16, 223, 32>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "x86_64")]
fn encode_extreme_ssse3<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 16, 191, 64>(data, |writer, batch_codewords| {
            encode_simd_batch_ssse3::<_, 16, 191, 64>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "x86_64")]
fn encode_standard_avx2<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 32, 239, 16>(data, |writer, batch_codewords| {
            encode_simd_batch_avx2::<_, 32, 239, 16>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "x86_64")]
fn encode_paranoid_avx2<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 32, 223, 32>(data, |writer, batch_codewords| {
            encode_simd_batch_avx2::<_, 32, 223, 32>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "x86_64")]
fn encode_extreme_avx2<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 32, 191, 64>(data, |writer, batch_codewords| {
            encode_simd_batch_avx2::<_, 32, 191, 64>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "x86_64")]
fn encode_standard_avx2_gfni<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 32, 239, 16>(data, |writer, batch_codewords| {
            encode_simd_batch_avx2_gfni::<_, 32, 239, 16>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "x86_64")]
fn encode_paranoid_avx2_gfni<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 32, 223, 32>(data, |writer, batch_codewords| {
            encode_simd_batch_avx2_gfni::<_, 32, 223, 32>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "x86_64")]
fn encode_extreme_avx2_gfni<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 32, 191, 64>(data, |writer, batch_codewords| {
            encode_simd_batch_avx2_gfni::<_, 32, 191, 64>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "aarch64")]
fn encode_standard_neon<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 16, 239, 16>(data, |writer, batch_codewords| {
            encode_simd_batch_neon::<_, 16, 239, 16>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "aarch64")]
fn encode_paranoid_neon<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 16, 223, 32>(data, |writer, batch_codewords| {
            encode_simd_batch_neon::<_, 16, 223, 32>(writer, batch_codewords)
        })
    }
}

#[cfg(target_arch = "aarch64")]
fn encode_extreme_neon<W: Write>(encoder: &mut ECCEncoder<W>, data: &[u8]) -> Result<usize> {
    unsafe {
        encoder.encode_batch::<_, 16, 191, 64>(data, |writer, batch_codewords| {
            encode_simd_batch_neon::<_, 16, 191, 64>(writer, batch_codewords)
        })
    }
}

/// Error Correction Code Writer that applies Reed-Solomon encoding to compressed data.
pub struct ECCEncoder<W> {
    inner: W,
    encode_fn: EncodeFunction<W>,
    encode_fn_simd: Option<EncodeFunction<W>>,
    buffer: CircularBuffer,
    uses_buffer: bool,
    batch_size: usize,
}

impl<W: Write> ECCEncoder<W> {
    #[cfg(target_arch = "x86_64")]
    fn get_simd_function(error_correction: ErrorCorrection) -> (Option<EncodeFunction<W>>, usize) {
        match error_correction {
            ErrorCorrection::None => (None, 1),
            ErrorCorrection::Standard => {
                if is_x86_feature_detected!("avx2") && is_x86_feature_detected!("gfni") {
                    (
                        Some(encode_standard_avx2_gfni as EncodeFunction<W>),
                        ECC_BATCH_SIZE_AVX,
                    )
                } else if is_x86_feature_detected!("avx2") {
                    (
                        Some(encode_standard_avx2 as EncodeFunction<W>),
                        ECC_BATCH_SIZE_AVX,
                    )
                } else if is_x86_feature_detected!("sse2") && is_x86_feature_detected!("gfni") {
                    (
                        Some(encode_standard_sse2_gfni as EncodeFunction<W>),
                        ECC_BATCH_SIZE_SSE,
                    )
                } else if is_x86_feature_detected!("ssse3") {
                    (
                        Some(encode_standard_ssse3 as EncodeFunction<W>),
                        ECC_BATCH_SIZE_SSE,
                    )
                } else {
                    (None, 1)
                }
            }
            ErrorCorrection::Paranoid => {
                if is_x86_feature_detected!("avx2") && is_x86_feature_detected!("gfni") {
                    (
                        Some(encode_paranoid_avx2_gfni as EncodeFunction<W>),
                        ECC_BATCH_SIZE_AVX,
                    )
                } else if is_x86_feature_detected!("avx2") {
                    (
                        Some(encode_paranoid_avx2 as EncodeFunction<W>),
                        ECC_BATCH_SIZE_AVX,
                    )
                } else if is_x86_feature_detected!("sse2") && is_x86_feature_detected!("gfni") {
                    (
                        Some(encode_paranoid_sse2_gfni as EncodeFunction<W>),
                        ECC_BATCH_SIZE_SSE,
                    )
                } else if is_x86_feature_detected!("ssse3") {
                    (
                        Some(encode_paranoid_ssse3 as EncodeFunction<W>),
                        ECC_BATCH_SIZE_SSE,
                    )
                } else {
                    (None, 1)
                }
            }
            ErrorCorrection::Extreme => {
                if is_x86_feature_detected!("avx2") && is_x86_feature_detected!("gfni") {
                    (
                        Some(encode_extreme_avx2_gfni as EncodeFunction<W>),
                        ECC_BATCH_SIZE_AVX,
                    )
                } else if is_x86_feature_detected!("avx2") {
                    (
                        Some(encode_extreme_avx2 as EncodeFunction<W>),
                        ECC_BATCH_SIZE_AVX,
                    )
                } else if is_x86_feature_detected!("sse2") && is_x86_feature_detected!("gfni") {
                    (
                        Some(encode_extreme_sse2_gfni as EncodeFunction<W>),
                        ECC_BATCH_SIZE_SSE,
                    )
                } else if is_x86_feature_detected!("ssse3") {
                    (
                        Some(encode_extreme_ssse3 as EncodeFunction<W>),
                        ECC_BATCH_SIZE_SSE,
                    )
                } else {
                    (None, 1)
                }
            }
        }
    }

    #[cfg(all(feature = "std", target_arch = "aarch64"))]
    fn get_simd_function(error_correction: ErrorCorrection) -> (Option<EncodeFunction<W>>, usize) {
        use std::arch::is_aarch64_feature_detected;

        match error_correction {
            ErrorCorrection::None => (None, 1),
            ErrorCorrection::Standard => {
                if is_aarch64_feature_detected!("neon") {
                    (
                        Some(encode_standard_neon as EncodeFunction<W>),
                        ECC_BATCH_SIZE_NEON,
                    )
                } else {
                    (None, 1)
                }
            }
            ErrorCorrection::Paranoid => {
                if is_aarch64_feature_detected!("neon") {
                    (
                        Some(encode_paranoid_neon as EncodeFunction<W>),
                        ECC_BATCH_SIZE_NEON,
                    )
                } else {
                    (None, 1)
                }
            }
            ErrorCorrection::Extreme => {
                if is_aarch64_feature_detected!("neon") {
                    (
                        Some(encode_extreme_neon as EncodeFunction<W>),
                        ECC_BATCH_SIZE_NEON,
                    )
                } else {
                    (None, 1)
                }
            }
        }
    }

    #[cfg(not(any(target_arch = "x86_64", all(target_arch = "aarch64", feature = "std"))))]
    fn get_simd_function(_error_correction: ErrorCorrection) -> (Option<EncodeFunction<W>>, usize) {
        (None, 1)
    }

    fn apply_simd_override(
        error_correction: ErrorCorrection,
        override_setting: SimdOverride,
    ) -> (Option<EncodeFunction<W>>, usize) {
        match override_setting {
            SimdOverride::Auto => Self::get_simd_function(error_correction),
            SimdOverride::ForceScalar => (None, 1),
            #[cfg(target_arch = "x86_64")]
            SimdOverride::ForceSsse3 => {
                if is_x86_feature_detected!("ssse3") {
                    match error_correction {
                        ErrorCorrection::None => (None, 1),
                        ErrorCorrection::Standard => (
                            Some(encode_standard_ssse3 as EncodeFunction<W>),
                            ECC_BATCH_SIZE_SSE,
                        ),
                        ErrorCorrection::Paranoid => (
                            Some(encode_paranoid_ssse3 as EncodeFunction<W>),
                            ECC_BATCH_SIZE_SSE,
                        ),
                        ErrorCorrection::Extreme => (
                            Some(encode_extreme_ssse3 as EncodeFunction<W>),
                            ECC_BATCH_SIZE_SSE,
                        ),
                    }
                } else {
                    eprintln!("Warning: SSSE3 requested but not available, falling back to scalar");
                    (None, 1)
                }
            }
            #[cfg(target_arch = "x86_64")]
            SimdOverride::ForceSse2Gfni => {
                if is_x86_feature_detected!("ssse3") && is_x86_feature_detected!("gfni") {
                    match error_correction {
                        ErrorCorrection::None => (None, 1),
                        ErrorCorrection::Standard => (
                            Some(encode_standard_sse2_gfni as EncodeFunction<W>),
                            ECC_BATCH_SIZE_SSE,
                        ),
                        ErrorCorrection::Paranoid => (
                            Some(encode_paranoid_sse2_gfni as EncodeFunction<W>),
                            ECC_BATCH_SIZE_SSE,
                        ),
                        ErrorCorrection::Extreme => (
                            Some(encode_extreme_sse2_gfni as EncodeFunction<W>),
                            ECC_BATCH_SIZE_SSE,
                        ),
                    }
                } else {
                    eprintln!(
                        "Warning: SSSE3+GFNI requested but not available, falling back to scalar"
                    );
                    (None, 1)
                }
            }
            #[cfg(target_arch = "x86_64")]
            SimdOverride::ForceAvx2 => {
                if is_x86_feature_detected!("avx2") {
                    match error_correction {
                        ErrorCorrection::None => (None, 1),
                        ErrorCorrection::Standard => (
                            Some(encode_standard_avx2 as EncodeFunction<W>),
                            ECC_BATCH_SIZE_AVX,
                        ),
                        ErrorCorrection::Paranoid => (
                            Some(encode_paranoid_avx2 as EncodeFunction<W>),
                            ECC_BATCH_SIZE_AVX,
                        ),
                        ErrorCorrection::Extreme => (
                            Some(encode_extreme_avx2 as EncodeFunction<W>),
                            ECC_BATCH_SIZE_AVX,
                        ),
                    }
                } else {
                    eprintln!("Warning: AVX2 requested but not available, falling back to scalar");
                    (None, 1)
                }
            }
            #[cfg(target_arch = "x86_64")]
            SimdOverride::ForceAvx2Gfni => {
                if is_x86_feature_detected!("avx2") && is_x86_feature_detected!("gfni") {
                    match error_correction {
                        ErrorCorrection::None => (None, 1),
                        ErrorCorrection::Standard => (
                            Some(encode_standard_avx2_gfni as EncodeFunction<W>),
                            ECC_BATCH_SIZE_AVX,
                        ),
                        ErrorCorrection::Paranoid => (
                            Some(encode_paranoid_avx2_gfni as EncodeFunction<W>),
                            ECC_BATCH_SIZE_AVX,
                        ),
                        ErrorCorrection::Extreme => (
                            Some(encode_extreme_avx2_gfni as EncodeFunction<W>),
                            ECC_BATCH_SIZE_AVX,
                        ),
                    }
                } else {
                    eprintln!(
                        "Warning: AVX2+GFNI requested but not available, falling back to scalar"
                    );
                    (None, 1)
                }
            }
            #[cfg(target_arch = "aarch64")]
            SimdOverride::ForceNeon => {
                #[cfg(feature = "std")]
                {
                    if std::arch::is_aarch64_feature_detected!("neon") {
                        match error_correction {
                            ErrorCorrection::None => (None, 1),
                            ErrorCorrection::Standard => (
                                Some(encode_standard_neon as EncodeFunction<W>),
                                ECC_BATCH_SIZE_NEON,
                            ),
                            ErrorCorrection::Paranoid => (
                                Some(encode_paranoid_neon as EncodeFunction<W>),
                                ECC_BATCH_SIZE_NEON,
                            ),
                            ErrorCorrection::Extreme => (
                                Some(encode_extreme_neon as EncodeFunction<W>),
                                ECC_BATCH_SIZE_NEON,
                            ),
                        }
                    } else {
                        eprintln!(
                            "Warning: NEON requested but not available, falling back to scalar"
                        );
                        (None, 1)
                    }
                }
                #[cfg(not(feature = "std"))]
                {
                    eprintln!(
                        "Warning: NEON detection not available in no_std, falling back to scalar"
                    );
                    (None, 1)
                }
            }
        }
    }

    /// Create a new ECCWriter with the specified error correction level and SIMD override.
    pub fn new(inner: W, error_correction: ErrorCorrection, simd_override: SimdOverride) -> Self {
        let (encode_fn_simd, simd_batch_size) =
            Self::apply_simd_override(error_correction, simd_override);

        let (encode_fn, uses_buffer, codeword_data_len) = match error_correction {
            ErrorCorrection::None => (encode_none as EncodeFunction<W>, false, 0),
            ErrorCorrection::Standard => (encode_standard as EncodeFunction<W>, true, 239),
            ErrorCorrection::Paranoid => (encode_paranoid as EncodeFunction<W>, true, 223),
            ErrorCorrection::Extreme => (encode_extreme as EncodeFunction<W>, true, 191),
        };

        let batch_size = if encode_fn_simd.is_some() {
            simd_batch_size * codeword_data_len
        } else {
            codeword_data_len
        };

        let buffer = if uses_buffer {
            CircularBuffer::with_capacity(batch_size * 2)
        } else {
            CircularBuffer::with_capacity(0)
        };

        Self {
            inner,
            encode_fn,
            encode_fn_simd,
            buffer,
            uses_buffer,
            batch_size,
        }
    }

    fn encode_and_write_data(&mut self, data: &[u8]) -> Result<usize> {
        if let Some(simd_fn) = self.encode_fn_simd {
            simd_fn(self, data)
        } else {
            (self.encode_fn)(self, data)
        }
    }

    #[inline(always)]
    fn encode_scalar<F, const DATA_LEN: usize, const PARITY_LEN: usize>(
        &mut self,
        data: &[u8],
        encode_rs_fn: F,
    ) -> Result<usize>
    where
        F: Fn(&[u8; DATA_LEN]) -> [u8; PARITY_LEN],
    {
        let mut input_processed = 0;

        while self.buffer.available_data() >= DATA_LEN
            || (data.is_empty() && self.buffer.available_data() > 0)
        {
            let mut codeword_data = [0u8; DATA_LEN];
            let copied = self.buffer.copy_to(&mut codeword_data);

            // Zero-padding is already in place from initialization.
            let parity = encode_rs_fn(&codeword_data);
            self.inner.write_all(&codeword_data)?;
            self.inner.write_all(&parity)?;

            self.buffer.consume(copied.min(DATA_LEN));

            if copied < DATA_LEN {
                break;
            }
        }

        if data.is_empty() {
            return Ok(0);
        }

        let mut pos = 0;
        while pos + DATA_LEN <= data.len() {
            let mut codeword_data = [0u8; DATA_LEN];
            codeword_data.copy_from_slice(&data[pos..pos + DATA_LEN]);

            let parity = encode_rs_fn(&codeword_data);
            self.inner.write_all(&codeword_data)?;
            self.inner.write_all(&parity)?;

            pos += DATA_LEN;
            input_processed += DATA_LEN;
        }

        Ok(input_processed)
    }

    unsafe fn encode_batch<F, const BATCH: usize, const DATA_LEN: usize, const PARITY_LEN: usize>(
        &mut self,
        data: &[u8],
        simd_encode_fn: F,
    ) -> Result<usize>
    where
        F: Fn(&mut W, &[[u8; DATA_LEN]; BATCH]) -> Result<()>,
    {
        let batch_data_size = BATCH * DATA_LEN;
        let mut input_processed = 0;

        if self.buffer.available_data() >= batch_data_size {
            let mut batch_codewords = [[0u8; DATA_LEN]; BATCH];

            if self
                .buffer
                .fill_batch_from_buffer::<BATCH, DATA_LEN>(&mut batch_codewords, batch_data_size)
            {
                simd_encode_fn(&mut self.inner, &batch_codewords)?;

                self.buffer.consume(batch_data_size);
                return Ok(0);
            }
        }

        // Try to process aligned data directly without copying.
        let (left, aligned, _right) = unsafe { data.align_to::<[u8; DATA_LEN]>() };

        if left.is_empty() && aligned.len() >= BATCH {
            // Data is perfectly aligned, and we have enough for at least one batch!
            let batches_possible = aligned.len() / BATCH;

            for batch_idx in 0..batches_possible {
                let batch_start = batch_idx * BATCH;
                let batch_slice = &aligned[batch_start..batch_start + BATCH];

                // Safe transmute: we know the slice has exactly BATCH elements of [u8; DATA_LEN]
                let batch_codewords: &[[u8; DATA_LEN]; BATCH] =
                    unsafe { &*(batch_slice.as_ptr() as *const [[u8; DATA_LEN]; BATCH]) };

                simd_encode_fn(&mut self.inner, batch_codewords)?;

                input_processed += batch_data_size;
            }

            return Ok(input_processed);
        }

        // Fallback to copying approach for misaligned or insufficient data.
        let mut pos = 0;
        while pos + batch_data_size <= data.len() {
            let mut batch_codewords = [[0u8; DATA_LEN]; BATCH];

            for (i, codeword) in batch_codewords.iter_mut().enumerate() {
                let start = pos + i * DATA_LEN;
                let end = start + DATA_LEN;
                codeword.copy_from_slice(&data[start..end]);
            }

            simd_encode_fn(&mut self.inner, &batch_codewords)?;

            pos += batch_data_size;
            input_processed += batch_data_size;
        }

        Ok(input_processed)
    }

    /// Finish writing and flush any remaining data.
    pub fn finish(mut self) -> Result<W> {
        if self.buffer.available_data() > 0 {
            // Process remaining buffered data.
            (self.encode_fn)(&mut self, &[])?;
        }

        Ok(self.inner)
    }
}

impl<W: Write> Write for ECCEncoder<W> {
    fn write(&mut self, buf: &[u8]) -> Result<usize> {
        if !self.uses_buffer {
            return self.inner.write(buf);
        }

        if buf.is_empty() {
            return Ok(0);
        }

        let mut buf_pos = 0;

        // If we have buffered data, try to fill it to a processable amount.
        if self.buffer.available_data() > 0 {
            let needed = self.batch_size - self.buffer.available_data();
            let available = buf.len().min(needed);

            self.buffer.append(&buf[..available]);
            buf_pos += available;

            // Try to process the buffer if it's full enough, else return.
            match self.buffer.available_data() >= self.batch_size {
                true => {
                    self.encode_and_write_data(&[])?;
                }
                false => {
                    // Not enough data to encode yet.
                    return Ok(available);
                }
            }
        }

        let remaining = buf.len() - buf_pos;

        if remaining >= self.batch_size {
            // We have enough for a complete batch, process directly.
            let processed = self.encode_and_write_data(&buf[buf_pos..])?;
            if buf_pos + processed < buf.len() {
                self.buffer.append(&buf[buf_pos + processed..]);
            }
        } else {
            // Not enough for a complete batch, store in buffer.
            if remaining > 0 {
                self.buffer.append(&buf[buf_pos..]);
            }
        }

        Ok(buf.len())
    }

    fn flush(&mut self) -> Result<()> {
        self.inner.flush()
    }
}

#[cfg(target_arch = "x86_64")]
#[target_feature(enable = "avx2,gfni")]
unsafe fn encode_simd_batch_avx2_gfni<
    R: Write,
    const BATCH: usize,
    const DATA_LEN: usize,
    const PARITY_LEN: usize,
>(
    mut writer: R,
    batch_codewords: &[[u8; DATA_LEN]; BATCH],
) -> Result<()> {
    use core::arch::x86_64::*;

    let transposed_data = crate::transpose_for_simd::<BATCH, DATA_LEN>(batch_codewords);
    let gen_poly = get_generator_poly::<PARITY_LEN>();

    // LFSR-based encoding with SIMD.
    let mut remainder = [[0u8; BATCH]; PARITY_LEN];

    // Process each data byte position (from highest to lowest).
    for data_bytes in transposed_data.iter().rev() {
        let data_ptr = data_bytes.as_ptr() as *const __m256i;
        let data_vec = unsafe { _mm256_loadu_si256(data_ptr) };

        // XOR with feedback from the highest remainder position.
        let feedback_ptr = remainder[PARITY_LEN - 1].as_ptr() as *const __m256i;
        let feedback_vec = unsafe { _mm256_loadu_si256(feedback_ptr) };
        let feedback = _mm256_xor_si256(data_vec, feedback_vec);

        // Shift remainder right.
        for i in (1..PARITY_LEN).rev() {
            remainder[i] = remainder[i - 1];
        }
        remainder[0] = [0u8; BATCH];

        // Apply generator polynomial multiplication with GFNI.
        for (i, &g_coeff) in gen_poly[..PARITY_LEN].iter().enumerate() {
            if g_coeff != 0 {
                let g_vec = _mm256_set1_epi8(g_coeff as i8);
                let product = _mm256_gf2p8mul_epi8(feedback, g_vec);

                let current_ptr = remainder[i].as_ptr() as *const __m256i;
                let current = unsafe { _mm256_loadu_si256(current_ptr) };
                let result = _mm256_xor_si256(current, product);
                let result_ptr = remainder[i].as_mut_ptr() as *mut __m256i;
                unsafe { _mm256_storeu_si256(result_ptr, result) };
            }
        }
    }

    let (data_codewords, parity_codewords) =
        crate::transpose_from_simd::<BATCH, DATA_LEN, PARITY_LEN>(&transposed_data, &remainder);

    for i in 0..BATCH {
        writer.write_all(&data_codewords[i])?;
        writer.write_all(&parity_codewords[i])?;
    }

    Ok(())
}

#[cfg(target_arch = "x86_64")]
#[target_feature(enable = "avx,avx2")]
unsafe fn encode_simd_batch_avx2<
    R: Write,
    const BATCH: usize,
    const DATA_LEN: usize,
    const PARITY_LEN: usize,
>(
    mut writer: R,
    batch_codewords: &[[u8; DATA_LEN]; BATCH],
) -> Result<()> {
    use core::arch::x86_64::*;

    use crate::reed_solomon::simd::{RS_255_191_TABLES, RS_255_223_TABLES, RS_255_239_TABLES};

    // Define a macro to reduce repetition in applying tables.
    macro_rules! apply_tables_for_parity {
        ($parity:expr, $tables:expr, $table_len:expr, $feedback:expr, $remainder:expr) => {
            for i in 0..$parity {
                if i < $table_len {
                    let table = &$tables[0][i];
                    unsafe {
                        apply_avx2_gf_multiplication::<BATCH>($feedback, &mut $remainder[i], table);
                    }
                }
            }
        };
    }

    let transposed_data = crate::transpose_for_simd::<BATCH, DATA_LEN>(batch_codewords);

    // LFSR-based encoding with SIMD and lookup tables.
    let mut remainder = [[0u8; BATCH]; PARITY_LEN];

    // Process each data byte position (from highest to lowest).
    for data_bytes in transposed_data.iter().rev() {
        let data_ptr = data_bytes.as_ptr() as *const __m256i;
        let data_vec = unsafe { _mm256_loadu_si256(data_ptr) };

        // XOR with feedback from the highest remainder position.
        let feedback_ptr = remainder[PARITY_LEN - 1].as_ptr() as *const __m256i;
        let feedback_vec = unsafe { _mm256_loadu_si256(feedback_ptr) };
        let feedback = _mm256_xor_si256(data_vec, feedback_vec);

        // Shift remainder right.
        for i in (1..PARITY_LEN).rev() {
            remainder[i] = remainder[i - 1];
        }
        remainder[0] = [0u8; BATCH];

        // Apply generator polynomial multiplication using lookup tables.
        match PARITY_LEN {
            16 => apply_tables_for_parity!(PARITY_LEN, RS_255_239_TABLES, 17, &feedback, remainder),
            32 => apply_tables_for_parity!(PARITY_LEN, RS_255_223_TABLES, 33, &feedback, remainder),
            64 => apply_tables_for_parity!(PARITY_LEN, RS_255_191_TABLES, 65, &feedback, remainder),
            _ => unreachable!(),
        }
    }

    let (data_codewords, parity_codewords) =
        crate::transpose_from_simd::<BATCH, DATA_LEN, PARITY_LEN>(&transposed_data, &remainder);

    for i in 0..BATCH {
        writer.write_all(&data_codewords[i])?;
        writer.write_all(&parity_codewords[i])?;
    }

    Ok(())
}

#[cfg(target_arch = "x86_64")]
#[target_feature(enable = "avx,avx2")]
unsafe fn apply_avx2_gf_multiplication<const BATCH: usize>(
    feedback: &core::arch::x86_64::__m256i,
    remainder_row: &mut [u8; BATCH],
    table: &crate::reed_solomon::simd::GfFourBitTables,
) {
    use core::arch::x86_64::*;

    // Extract low and high nibbles from feedback vector.
    let low_nibble_mask = _mm256_set1_epi8(0x0F_u8 as i8);
    let low_nibbles = _mm256_and_si256(*feedback, low_nibble_mask);
    let high_nibbles = _mm256_srli_epi16::<4>(*feedback);
    let high_nibbles = _mm256_and_si256(high_nibbles, low_nibble_mask);

    // Perform table lookups for low nibbles.
    // Note: We need to duplicate the 16-byte table to fill the 32-byte AVX2 register.
    let low_table_ptr = table.low_four.as_ptr() as *const __m128i;
    let low_table_128 = unsafe { _mm_loadu_si128(low_table_ptr) };
    let low_table = _mm256_broadcastsi128_si256(low_table_128);
    let low_products = _mm256_shuffle_epi8(low_table, low_nibbles);

    // Perform table lookups for high nibbles.
    let high_table_ptr = table.high_four.as_ptr() as *const __m128i;
    let high_table_128 = unsafe { _mm_loadu_si128(high_table_ptr) };
    let high_table = _mm256_broadcastsi128_si256(high_table_128);
    let high_products = _mm256_shuffle_epi8(high_table, high_nibbles);

    // Combine low and high products.
    let products = _mm256_xor_si256(low_products, high_products);

    // XOR with current remainder.
    let current_ptr = remainder_row.as_ptr() as *const __m256i;
    let current = unsafe { _mm256_loadu_si256(current_ptr) };
    let result = _mm256_xor_si256(current, products);
    let result_ptr = remainder_row.as_mut_ptr() as *mut __m256i;
    unsafe { _mm256_storeu_si256(result_ptr, result) };
}

#[cfg(target_arch = "x86_64")]
#[target_feature(enable = "sse2,gfni")]
unsafe fn encode_simd_batch_sse2_gfni<
    R: Write,
    const BATCH: usize,
    const DATA_LEN: usize,
    const PARITY_LEN: usize,
>(
    mut writer: R,
    batch_codewords: &[[u8; DATA_LEN]; BATCH],
) -> Result<()> {
    use core::arch::x86_64::*;

    let transposed_data = crate::transpose_for_simd::<BATCH, DATA_LEN>(batch_codewords);
    let gen_poly = get_generator_poly::<PARITY_LEN>();

    // LFSR-based encoding with SIMD.
    let mut remainder = [[0u8; BATCH]; PARITY_LEN];

    // Process each data byte position (from highest to lowest).
    for data_bytes in transposed_data.iter().rev() {
        let data_ptr = data_bytes.as_ptr() as *const __m128i;
        let data_vec = unsafe { _mm_loadu_si128(data_ptr) };

        // XOR with feedback from the highest remainder position.
        let feedback_ptr = remainder[PARITY_LEN - 1].as_ptr() as *const __m128i;
        let feedback_vec = unsafe { _mm_loadu_si128(feedback_ptr) };
        let feedback = _mm_xor_si128(data_vec, feedback_vec);

        // Shift remainder right.
        for i in (1..PARITY_LEN).rev() {
            remainder[i] = remainder[i - 1];
        }
        remainder[0] = [0u8; BATCH];

        // Apply generator polynomial multiplication with GFNI.
        for (i, &g_coeff) in gen_poly[..PARITY_LEN].iter().enumerate() {
            if g_coeff != 0 {
                let g_vec = _mm_set1_epi8(g_coeff as i8);
                let product = _mm_gf2p8mul_epi8(feedback, g_vec);

                let current_ptr = remainder[i].as_ptr() as *const __m128i;
                let current = unsafe { _mm_loadu_si128(current_ptr) };
                let result = _mm_xor_si128(current, product);
                let result_ptr = remainder[i].as_mut_ptr() as *mut __m128i;
                unsafe { _mm_storeu_si128(result_ptr, result) };
            }
        }
    }

    let (data_codewords, parity_codewords) =
        crate::transpose_from_simd::<BATCH, DATA_LEN, PARITY_LEN>(&transposed_data, &remainder);

    for i in 0..BATCH {
        writer.write_all(&data_codewords[i])?;
        writer.write_all(&parity_codewords[i])?;
    }

    Ok(())
}

#[cfg(target_arch = "x86_64")]
#[target_feature(enable = "sse2,ssse3")]
unsafe fn encode_simd_batch_ssse3<
    R: Write,
    const BATCH: usize,
    const DATA_LEN: usize,
    const PARITY_LEN: usize,
>(
    mut writer: R,
    batch_codewords: &[[u8; DATA_LEN]; BATCH],
) -> Result<()> {
    use core::arch::x86_64::*;

    use crate::reed_solomon::simd::{RS_255_191_TABLES, RS_255_223_TABLES, RS_255_239_TABLES};

    // Define a macro to reduce repetition in applying tables.
    macro_rules! apply_tables_for_parity {
        ($parity:expr, $tables:expr, $table_len:expr, $feedback:expr, $remainder:expr) => {
            for i in 0..$parity {
                if i < $table_len {
                    let table = &$tables[0][i];
                    unsafe {
                        apply_ssse3_gf_multiplication::<BATCH>(
                            $feedback,
                            &mut $remainder[i],
                            table,
                        );
                    }
                }
            }
        };
    }

    let transposed_data = crate::transpose_for_simd::<BATCH, DATA_LEN>(batch_codewords);

    // LFSR-based encoding with SIMD and lookup tables.
    let mut remainder = [[0u8; BATCH]; PARITY_LEN];

    // Process each data byte position (from highest to lowest).
    for data_bytes in transposed_data.iter().rev() {
        let data_ptr = data_bytes.as_ptr() as *const __m128i;
        let data_vec = unsafe { _mm_loadu_si128(data_ptr) };

        // XOR with feedback from the highest remainder position.
        let feedback_ptr = remainder[PARITY_LEN - 1].as_ptr() as *const __m128i;
        let feedback_vec = unsafe { _mm_loadu_si128(feedback_ptr) };
        let feedback = _mm_xor_si128(data_vec, feedback_vec);

        // Shift remainder right.
        for i in (1..PARITY_LEN).rev() {
            remainder[i] = remainder[i - 1];
        }
        remainder[0] = [0u8; BATCH];

        // Apply generator polynomial multiplication using lookup tables.
        match PARITY_LEN {
            16 => apply_tables_for_parity!(PARITY_LEN, RS_255_239_TABLES, 17, &feedback, remainder),
            32 => apply_tables_for_parity!(PARITY_LEN, RS_255_223_TABLES, 33, &feedback, remainder),
            64 => apply_tables_for_parity!(PARITY_LEN, RS_255_191_TABLES, 65, &feedback, remainder),
            _ => unreachable!(),
        }
    }

    let (data_codewords, parity_codewords) =
        crate::transpose_from_simd::<BATCH, DATA_LEN, PARITY_LEN>(&transposed_data, &remainder);

    for i in 0..BATCH {
        writer.write_all(&data_codewords[i])?;
        writer.write_all(&parity_codewords[i])?;
    }

    Ok(())
}

#[cfg(target_arch = "x86_64")]
#[target_feature(enable = "sse2,ssse3")]
unsafe fn apply_ssse3_gf_multiplication<const BATCH: usize>(
    feedback: &core::arch::x86_64::__m128i,
    remainder_row: &mut [u8; BATCH],
    table: &crate::reed_solomon::simd::GfFourBitTables,
) {
    use core::arch::x86_64::*;

    // Extract low and high nibbles from feedback vector.
    let low_nibble_mask = _mm_set1_epi8(0x0F_u8 as i8);
    let low_nibbles = _mm_and_si128(*feedback, low_nibble_mask);

    // For high nibbles, first shift right by 4 bits per byte.
    let high_nibbles = _mm_srli_epi16::<4>(*feedback);
    let high_nibbles = _mm_and_si128(high_nibbles, low_nibble_mask);

    // Perform table lookups for low nibbles using SSSE3 shuffle.
    let low_table_ptr = table.low_four.as_ptr() as *const __m128i;
    let low_table = unsafe { _mm_loadu_si128(low_table_ptr) };
    let low_products = _mm_shuffle_epi8(low_table, low_nibbles);

    // Perform table lookups for high nibbles using SSSE3 shuffle.
    let high_table_ptr = table.high_four.as_ptr() as *const __m128i;
    let high_table = unsafe { _mm_loadu_si128(high_table_ptr) };
    let high_products = _mm_shuffle_epi8(high_table, high_nibbles);

    // Combine low and high products.
    let products = _mm_xor_si128(low_products, high_products);

    // XOR with current remainder.
    let current_ptr = remainder_row.as_ptr() as *const __m128i;
    let current = unsafe { _mm_loadu_si128(current_ptr) };
    let result = _mm_xor_si128(current, products);
    let result_ptr = remainder_row.as_mut_ptr() as *mut __m128i;
    unsafe { _mm_storeu_si128(result_ptr, result) };
}

#[cfg(target_arch = "aarch64")]
#[target_feature(enable = "neon")]
unsafe fn encode_simd_batch_neon<
    R: Write,
    const BATCH: usize,
    const DATA_LEN: usize,
    const PARITY_LEN: usize,
>(
    mut writer: R,
    batch_codewords: &[[u8; DATA_LEN]; BATCH],
) -> Result<()> {
    use core::arch::aarch64::*;

    use crate::reed_solomon::simd::{RS_255_191_TABLES, RS_255_223_TABLES, RS_255_239_TABLES};

    // Define a macro to reduce repetition in applying tables.
    macro_rules! apply_tables_for_parity {
        ($parity:expr, $tables:expr, $table_len:expr, $feedback:expr, $remainder:expr) => {
            for i in 0..$parity {
                if i < $table_len {
                    let table = &$tables[0][i];
                    unsafe {
                        apply_neon_gf_multiplication::<BATCH>($feedback, &mut $remainder[i], table);
                    }
                }
            }
        };
    }

    let transposed_data = crate::transpose_for_simd::<BATCH, DATA_LEN>(batch_codewords);

    // LFSR-based encoding with SIMD and lookup tables.
    let mut remainder = [[0u8; BATCH]; PARITY_LEN];

    // Process each data byte position (from highest to lowest).
    for data_bytes in transposed_data.iter().rev() {
        let data_vec = unsafe { vld1q_u8(data_bytes.as_ptr()) };

        // XOR with feedback from the highest remainder position.
        let feedback_vec = unsafe { vld1q_u8(remainder[PARITY_LEN - 1].as_ptr()) };
        let feedback = veorq_u8(data_vec, feedback_vec);

        // Shift remainder right.
        for i in (1..PARITY_LEN).rev() {
            remainder[i] = remainder[i - 1];
        }
        remainder[0] = [0u8; BATCH];

        // Apply generator polynomial multiplication using lookup tables.
        match PARITY_LEN {
            16 => apply_tables_for_parity!(PARITY_LEN, RS_255_239_TABLES, 17, &feedback, remainder),
            32 => apply_tables_for_parity!(PARITY_LEN, RS_255_223_TABLES, 33, &feedback, remainder),
            64 => apply_tables_for_parity!(PARITY_LEN, RS_255_191_TABLES, 65, &feedback, remainder),
            _ => unreachable!(),
        }
    }

    let (data_codewords, parity_codewords) =
        crate::transpose_from_simd::<BATCH, DATA_LEN, PARITY_LEN>(&transposed_data, &remainder);

    for i in 0..BATCH {
        writer.write_all(&data_codewords[i])?;
        writer.write_all(&parity_codewords[i])?;
    }

    Ok(())
}

#[cfg(target_arch = "aarch64")]
#[target_feature(enable = "neon")]
unsafe fn apply_neon_gf_multiplication<const BATCH: usize>(
    feedback: &core::arch::aarch64::uint8x16_t,
    remainder_row: &mut [u8; BATCH],
    table: &crate::reed_solomon::simd::GfFourBitTables,
) {
    use core::arch::aarch64::*;

    // Extract low and high nibbles from feedback vector.
    let low_nibble_mask = vdupq_n_u8(0x0F);
    let low_nibbles = vandq_u8(*feedback, low_nibble_mask);
    let high_nibbles = vshrq_n_u8::<4>(*feedback);

    // Perform table lookups for low nibbles.
    let low_table = unsafe { vld1q_u8(table.low_four.as_ptr()) };
    let low_products = vqtbl1q_u8(low_table, low_nibbles);

    // Perform table lookups for high nibbles.
    let high_table = unsafe { vld1q_u8(table.high_four.as_ptr()) };
    let high_products = vqtbl1q_u8(high_table, high_nibbles);

    // Combine low and high products.
    let products = veorq_u8(low_products, high_products);

    // XOR with current remainder.
    let current = unsafe { vld1q_u8(remainder_row.as_ptr()) };
    let result = veorq_u8(current, products);
    unsafe { vst1q_u8(remainder_row.as_mut_ptr(), result) };
}

#[cfg(test)]
mod tests {
    use alloc::vec::Vec;

    use super::*;

    #[test]
    fn test_ec_encoder_none_passthrough() {
        let mut output = Vec::new();
        let mut ec_encoder =
            ECCEncoder::new(&mut output, ErrorCorrection::None, SimdOverride::Auto);

        let test_data = b"Hello, World!";
        ec_encoder.write_all(test_data).unwrap();

        let _final_output = ec_encoder.finish().unwrap();

        assert_eq!(output, test_data);
    }

    #[test]
    fn test_ec_encoder_standard_encoding() {
        let mut output = Vec::new();
        let mut ec_encoder =
            ECCEncoder::new(&mut output, ErrorCorrection::Standard, SimdOverride::Auto);

        let test_data = b"Hello, Reed-Solomon encoding!";
        ec_encoder.write_all(test_data).unwrap();

        let _final_output = ec_encoder.finish().unwrap();

        assert_eq!(output.len(), 255);

        assert_eq!(&output[..test_data.len()], test_data);

        for (i, &x) in output[test_data.len()..239].iter().enumerate() {
            assert_eq!(x, 0, "Padding should be zero at position {i}");
        }

        assert_eq!(output[239..].len(), 16);
    }

    #[test]
    fn test_ec_encoder_multiple_codewords() {
        let mut output = Vec::new();
        let mut ec_encoder =
            ECCEncoder::new(&mut output, ErrorCorrection::Standard, SimdOverride::Auto);

        let mut test_data = Vec::new();
        test_data.extend_from_slice(b"A".repeat(300).as_slice());

        ec_encoder.write_all(&test_data).unwrap();
        let _final_output = ec_encoder.finish().unwrap();

        // We should have 2 codewords: 2 * 255 = 510 bytes.
        assert_eq!(output.len(), 2 * 255);

        assert_eq!(&output[..239], &test_data[0..239]);

        let second_codeword_data_size = 300 - 239;

        assert_eq!(
            &output[255..255 + second_codeword_data_size],
            &test_data[239..]
        );

        for (i, &x) in output[(255 + second_codeword_data_size)..(255 + 239)]
            .iter()
            .enumerate()
        {
            assert_eq!(x, 0, "Padding should be zero at position {i}");
        }
    }

    #[test]
    fn test_all_simd_paths_on_current_arch() {
        fn test_simd_path_consistency(
            error_correction: ErrorCorrection,
            simd_override: SimdOverride,
            test_name: &str,
        ) -> bool {
            let test_data =
                b"Hello, World! This is a test of SIMD consistency across all paths.".repeat(20);

            let mut scalar_output = Vec::new();
            let mut scalar_encoder = ECCEncoder::new(
                &mut scalar_output,
                error_correction,
                SimdOverride::ForceScalar,
            );
            scalar_encoder.write_all(&test_data).unwrap();
            scalar_encoder.finish().unwrap();

            let mut simd_output = Vec::new();
            let mut simd_encoder =
                ECCEncoder::new(&mut simd_output, error_correction, simd_override);
            simd_encoder.write_all(&test_data).unwrap();
            simd_encoder.finish().unwrap();

            let matches = scalar_output == simd_output;
            if matches {
                println!("✓ {test_name} - outputs match");
            } else {
                println!("✗ {test_name} - outputs differ!");
                println!("  Scalar output length: {}", scalar_output.len());
                println!("  SIMD output length: {}", simd_output.len());
                println!("  Scalar hash: {}", blake3::hash(&scalar_output));
                println!("  SIMD hash: {}", blake3::hash(&simd_output));
            }
            matches
        }

        let error_corrections = [
            ErrorCorrection::Standard,
            ErrorCorrection::Paranoid,
            ErrorCorrection::Extreme,
        ];

        let mut all_passed = true;

        for &ec in &error_corrections {
            let ec_name = match ec {
                ErrorCorrection::None => "None",
                ErrorCorrection::Standard => "Standard",
                ErrorCorrection::Paranoid => "Paranoid",
                ErrorCorrection::Extreme => "Extreme",
            };

            #[cfg(target_arch = "x86_64")]
            {
                if is_x86_feature_detected!("sse2") && is_x86_feature_detected!("gfni") {
                    let test_name = format!("SSE2 + GFNI vs Scalar - {ec_name}");
                    all_passed &=
                        test_simd_path_consistency(ec, SimdOverride::ForceSse2Gfni, &test_name);
                } else {
                    println!("⊗ SSE2 + GFNI not available on this CPU - {ec_name}");
                }

                if is_x86_feature_detected!("ssse3") {
                    let test_name = format!("SSSE3 vs Scalar - {ec_name}");
                    all_passed &=
                        test_simd_path_consistency(ec, SimdOverride::ForceSsse3, &test_name);
                } else {
                    println!("⊗ SSSE3 not available on this CPU - {ec_name}");
                }

                if is_x86_feature_detected!("avx2") {
                    let test_name = format!("AVX2 (pure) vs Scalar - {ec_name}");
                    all_passed &=
                        test_simd_path_consistency(ec, SimdOverride::ForceAvx2, &test_name);
                } else {
                    println!("⊗ AVX2 not available on this CPU - {ec_name}");
                }

                if is_x86_feature_detected!("avx2") && is_x86_feature_detected!("gfni") {
                    let test_name = format!("AVX2 + GFNI vs Scalar - {ec_name}");
                    all_passed &=
                        test_simd_path_consistency(ec, SimdOverride::ForceAvx2Gfni, &test_name);
                } else {
                    println!("⊗ AVX2 + GFNI not available on this CPU - {ec_name}");
                }
            }

            #[cfg(all(target_arch = "aarch64", feature = "std"))]
            {
                // Test NEON if available
                if std::arch::is_aarch64_feature_detected!("neon") {
                    let test_name = format!("NEON vs Scalar - {ec_name}");
                    all_passed &=
                        test_simd_path_consistency(ec, SimdOverride::ForceNeon, &test_name);
                } else {
                    println!("⊗ NEON not available on this CPU - {ec_name}");
                }
            }

            #[cfg(not(any(
                target_arch = "x86_64",
                all(target_arch = "aarch64", feature = "std")
            )))]
            {
                println!("⊗ No SIMD paths available on this architecture - {ec_name}");
            }
        }

        assert!(
            all_passed,
            "One or more SIMD paths produced different outputs than scalar reference"
        );
    }
}