aboutsummaryrefslogtreecommitdiffstats
path: root/mistral/lab.cc
blob: 0719b37bc63ce7fee95ec11ca160b496e58b4f55 (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
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
/*
 *  nextpnr -- Next Generation Place and Route
 *
 *  Copyright (C) 2021  gatecat <gatecat@ds0.me>
 *
 *  Permission to use, copy, modify, and/or distribute this software for any
 *  purpose with or without fee is hereby granted, provided that the above
 *  copyright notice and this permission notice appear in all copies.
 *
 *  THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
 *  WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
 *  MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
 *  ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
 *  WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
 *  ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
 *  OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 *
 */

#include "design_utils.h"
#include "log.h"
#include "nextpnr.h"
#include "util.h"

NEXTPNR_NAMESPACE_BEGIN

// This file contains functions related to our custom LAB structure, including creating the LAB bels; checking the
// legality of LABs; and manipulating LUT inputs and equations

// LAB/ALM structure creation functions
namespace {
static void create_alm(Arch *arch, int x, int y, int z, uint32_t lab_idx)
{
    auto &lab = arch->labs.at(lab_idx);
    auto &alm = lab.alms.at(z);
    // Create the combinational part of ALMs.
    // There are two of these, for the two LUT outputs, and these also contain the carry chain and associated logic
    // Each one has all 8 ALM inputs as input pins. In many cases only a subset of these are used; depending on mode;
    // and the bel-cell pin mappings are used to handle this post-placement without losing flexibility
    for (int i = 0; i < 2; i++) {
        // Carry/share wires are a bit tricky due to all the different permutations
        WireId carry_in, share_in;
        WireId carry_out, share_out;
        if (z == 0 && i == 0) {
            if (y == 0) {
                // Base case
                carry_in = arch->add_wire(x, y, id_CARRY_START);
                share_in = arch->add_wire(x, y, id_CARRY_START);
            } else {
                // Output of last tile
                carry_in = arch->add_wire(x, y - 1, id_COUT);
                share_in = arch->add_wire(x, y - 1, id_SHAREOUT);
            }
        } else {
            // Output from last combinational unit
            carry_in = arch->add_wire(x, y, arch->id(stringf("CARRY[%d]", (z * 2 + i) - 1)));
            share_in = arch->add_wire(x, y, arch->id(stringf("SHARE[%d]", (z * 2 + i) - 1)));
        }
        if (z == 9 && i == 1) {
            carry_out = arch->add_wire(x, y, id_COUT);
            share_out = arch->add_wire(x, y, id_SHAREOUT);
        } else {
            carry_out = arch->add_wire(x, y, arch->id(stringf("CARRY[%d]", z * 2 + i)));
            share_out = arch->add_wire(x, y, arch->id(stringf("SHARE[%d]", z * 2 + i)));
        }

        BelId bel = arch->add_bel(x, y, arch->id(stringf("ALM%d_COMB%d", z, i)), id_MISTRAL_COMB);
        // LUT/MUX inputs
        arch->add_bel_pin(bel, id_A, PORT_IN, arch->get_port(CycloneV::LAB, x, y, z, CycloneV::A));
        arch->add_bel_pin(bel, id_B, PORT_IN, arch->get_port(CycloneV::LAB, x, y, z, CycloneV::B));
        arch->add_bel_pin(bel, id_C, PORT_IN, arch->get_port(CycloneV::LAB, x, y, z, CycloneV::C));
        arch->add_bel_pin(bel, id_D, PORT_IN, arch->get_port(CycloneV::LAB, x, y, z, CycloneV::D));
        arch->add_bel_pin(bel, id_E0, PORT_IN, arch->get_port(CycloneV::LAB, x, y, z, CycloneV::E0));
        arch->add_bel_pin(bel, id_E1, PORT_IN, arch->get_port(CycloneV::LAB, x, y, z, CycloneV::E1));
        arch->add_bel_pin(bel, id_F0, PORT_IN, arch->get_port(CycloneV::LAB, x, y, z, CycloneV::F0));
        arch->add_bel_pin(bel, id_F1, PORT_IN, arch->get_port(CycloneV::LAB, x, y, z, CycloneV::F1));
        // Carry/share chain
        arch->add_bel_pin(bel, id_CIN, PORT_IN, carry_in);
        arch->add_bel_pin(bel, id_SHAREIN, PORT_IN, share_in);
        arch->add_bel_pin(bel, id_COUT, PORT_OUT, carry_out);
        arch->add_bel_pin(bel, id_SHAREOUT, PORT_OUT, share_out);
        // Combinational output
        alm.comb_out[i] = arch->add_wire(x, y, arch->id(stringf("COMBOUT[%d]", z * 2 + i)));
        arch->add_bel_pin(bel, id_COMBOUT, PORT_OUT, alm.comb_out[i]);
        // Assign indexing
        alm.lut_bels.at(i) = bel;
        auto &b = arch->bel_data(bel);
        b.lab_data.lab = lab_idx;
        b.lab_data.alm = z;
        b.lab_data.idx = i;
    }
    // Create the control set and E/F selection - which is per pair of FF
    for (int i = 0; i < 2; i++) {
        // Wires
        alm.sel_clk[i] = arch->add_wire(x, y, arch->id(stringf("CLK%c[%d]", i ? 'B' : 'T', z)));
        alm.sel_ena[i] = arch->add_wire(x, y, arch->id(stringf("ENA%c[%d]", i ? 'B' : 'T', z)));
        alm.sel_aclr[i] = arch->add_wire(x, y, arch->id(stringf("ACLR%c[%d]", i ? 'B' : 'T', z)));
        alm.sel_ef[i] = arch->add_wire(x, y, arch->id(stringf("%cEF[%d]", i ? 'B' : 'T', z)));
        // Muxes - three CLK/ENA per LAB, two ACLR
        for (int j = 0; j < 3; j++) {
            arch->add_pip(lab.clk_wires[j], alm.sel_clk[i]);
            arch->add_pip(lab.ena_wires[j], alm.sel_ena[i]);
            if (j < 2)
                arch->add_pip(lab.aclr_wires[j], alm.sel_aclr[i]);
        }
        // E/F pips
        arch->add_pip(arch->get_port(CycloneV::LAB, x, y, z, i ? CycloneV::E1 : CycloneV::E0), alm.sel_ef[i]);
        arch->add_pip(arch->get_port(CycloneV::LAB, x, y, z, i ? CycloneV::F1 : CycloneV::F0), alm.sel_ef[i]);
    }

    // Create the flipflops and associated routing
    const CycloneV::port_type_t outputs[4] = {CycloneV::FFT0, CycloneV::FFT1, CycloneV::FFB0, CycloneV::FFB1};
    const CycloneV::port_type_t l_outputs[4] = {CycloneV::FFT1L, CycloneV::FFB1L};

    for (int i = 0; i < 4; i++) {
        // FF input, selected by *PKREG*
        alm.ff_in[i] = arch->add_wire(x, y, arch->id(stringf("FFIN[%d]", (z * 4) + i)));
        arch->add_pip(alm.comb_out[i / 2], alm.ff_in[i]);
        arch->add_pip(alm.sel_ef[i / 2], alm.ff_in[i]);
        // FF bel
        BelId bel = arch->add_bel(x, y, arch->id(stringf("ALM%d_FF%d", z, i)), id_MISTRAL_FF);
        arch->add_bel_pin(bel, id_CLK, PORT_IN, alm.sel_clk[i / 2]);
        arch->add_bel_pin(bel, id_ENA, PORT_IN, alm.sel_ena[i / 2]);
        arch->add_bel_pin(bel, id_ACLR, PORT_IN, alm.sel_aclr[i / 2]);
        arch->add_bel_pin(bel, id_SCLR, PORT_IN, lab.sclr_wire);
        arch->add_bel_pin(bel, id_SLOAD, PORT_IN, lab.sload_wire);
        arch->add_bel_pin(bel, id_DATAIN, PORT_IN, alm.ff_in[i]);
        arch->add_bel_pin(bel, id_SDATA, PORT_IN, alm.sel_ef[i / 2]);

        // FF output
        alm.ff_out[i] = arch->add_wire(x, y, arch->id(stringf("FFOUT[%d]", (z * 4) + i)));
        arch->add_bel_pin(bel, id_Q, PORT_OUT, alm.ff_out[i]);
        // Output mux (*DFF*)
        WireId out = arch->get_port(CycloneV::LAB, x, y, z, outputs[i]);
        arch->add_pip(alm.ff_out[i], out);
        arch->add_pip(alm.comb_out[i / 2], out);
        // 'L' output mux where applicable
        if (i == 1 || i == 3) {
            WireId l_out = arch->get_port(CycloneV::LAB, x, y, z, l_outputs[i / 2]);
            arch->add_pip(alm.ff_out[i], l_out);
            arch->add_pip(alm.comb_out[i / 2], l_out);
        }

        lab.alms.at(z).ff_bels.at(i) = bel;
        auto &b = arch->bel_data(bel);
        b.lab_data.lab = lab_idx;
        b.lab_data.alm = z;
        b.lab_data.idx = i;
    }
}
} // namespace

void Arch::create_lab(int x, int y)
{
    uint32_t lab_idx = labs.size();
    labs.emplace_back();

    auto &lab = labs.back();

    // Create common control set configuration. This is actually a subset of what's possible, but errs on the side of
    // caution due to incomplete documentation

    // Clocks - hardcode to CLKA choices, as both CLKA and CLKB coming from general routing causes unexpected
    // permutations
    for (int i = 0; i < 3; i++) {
        lab.clk_wires[i] = add_wire(x, y, id(stringf("CLK%d", i)));
        add_pip(get_port(CycloneV::LAB, x, y, -1, CycloneV::CLKIN, 0), lab.clk_wires[i]);  // dedicated routing
        add_pip(get_port(CycloneV::LAB, x, y, -1, CycloneV::DATAIN, 0), lab.clk_wires[i]); // general routing
    }

    // Enables - while it looks from the config like there are choices for these, it seems like EN0_SEL actually selects
    // SCLR not ENA0 and EN1_SEL actually selects SLOAD?
    lab.ena_wires[0] = get_port(CycloneV::LAB, x, y, -1, CycloneV::DATAIN, 2);
    lab.ena_wires[1] = get_port(CycloneV::LAB, x, y, -1, CycloneV::DATAIN, 3);
    lab.ena_wires[2] = get_port(CycloneV::LAB, x, y, -1, CycloneV::DATAIN, 0);

    // ACLRs - only consider general routing for now
    lab.aclr_wires[0] = get_port(CycloneV::LAB, x, y, -1, CycloneV::DATAIN, 3);
    lab.aclr_wires[1] = get_port(CycloneV::LAB, x, y, -1, CycloneV::DATAIN, 2);

    // SCLR and SLOAD - as above it seems like these might be selectable using the "EN*_SEL" bits but play it safe for
    // now
    lab.sclr_wire = get_port(CycloneV::LAB, x, y, -1, CycloneV::DATAIN, 3);
    lab.sload_wire = get_port(CycloneV::LAB, x, y, -1, CycloneV::DATAIN, 1);

    for (int i = 0; i < 10; i++) {
        create_alm(this, x, y, i, lab_idx);
    }
}

// Cell handling and annotation functions
namespace {
ControlSig get_ctrlsig(const Context *ctx, const CellInfo *cell, IdString port, bool explicit_const = false)
{
    ControlSig result;
    result.net = get_net_or_empty(cell, port);
    if (result.net == nullptr && explicit_const) {
        // For ENA, 1 (and 0) are explicit control set choices even though they aren't routed, as "no ENA" still
        // consumes a clock+ENA pair
        CellPinState st = PIN_1;
        result.net = ctx->nets.at((st == PIN_1) ? ctx->id("$PACKER_VCC_NET") : ctx->id("$PACKER_GND_NET")).get();
    }
    if (cell->pin_data.count(port))
        result.inverted = cell->pin_data.at(port).state == PIN_INV;
    else
        result.inverted = false;
    return result;
}
} // namespace

bool Arch::is_comb_cell(IdString cell_type) const
{
    // Return true if a cell is a combinational cell type, to be a placed at a MISTRAL_COMB location
    switch (cell_type.index) {
    case ID_MISTRAL_ALUT6:
    case ID_MISTRAL_ALUT5:
    case ID_MISTRAL_ALUT4:
    case ID_MISTRAL_ALUT3:
    case ID_MISTRAL_ALUT2:
    case ID_MISTRAL_NOT:
    case ID_MISTRAL_CONST:
    case ID_MISTRAL_ALUT_ARITH:
        return true;
    default:
        return false;
    }
}

void Arch::assign_comb_info(CellInfo *cell) const
{
    cell->combInfo.is_carry = false;
    cell->combInfo.is_shared = false;
    cell->combInfo.is_extended = false;

    if (cell->type == id_MISTRAL_ALUT_ARITH) {
        cell->combInfo.is_carry = true;
        cell->combInfo.lut_input_count = 5;
        cell->combInfo.lut_bits_count = 32;
        // This is a special case in terms of naming
        int i = 0;
        for (auto pin : {id_A, id_B, id_C, id_D0, id_D1}) {
            cell->combInfo.lut_in[i++] = get_net_or_empty(cell, pin);
        }
        cell->combInfo.comb_out = get_net_or_empty(cell, id_SO);
    } else {
        cell->combInfo.lut_input_count = 0;
        switch (cell->type.index) {
        case ID_MISTRAL_ALUT6:
            ++cell->combInfo.lut_input_count;
            cell->combInfo.lut_in[5] = get_net_or_empty(cell, id_F);
            [[fallthrough]];
        case ID_MISTRAL_ALUT5:
            ++cell->combInfo.lut_input_count;
            cell->combInfo.lut_in[4] = get_net_or_empty(cell, id_E);
            [[fallthrough]];
        case ID_MISTRAL_ALUT4:
            ++cell->combInfo.lut_input_count;
            cell->combInfo.lut_in[3] = get_net_or_empty(cell, id_D);
            [[fallthrough]];
        case ID_MISTRAL_ALUT3:
            ++cell->combInfo.lut_input_count;
            cell->combInfo.lut_in[2] = get_net_or_empty(cell, id_C);
            [[fallthrough]];
        case ID_MISTRAL_ALUT2:
            ++cell->combInfo.lut_input_count;
            cell->combInfo.lut_in[1] = get_net_or_empty(cell, id_B);
            [[fallthrough]];
        case ID_MISTRAL_BUF: // used to route through to FFs etc
        case ID_MISTRAL_NOT: // used for inverters that map to LUTs
            ++cell->combInfo.lut_input_count;
            cell->combInfo.lut_in[0] = get_net_or_empty(cell, id_A);
            [[fallthrough]];
        case ID_MISTRAL_CONST:
            // MISTRAL_CONST is a nextpnr-inserted cell type for 0-input, constant-generating LUTs
            break;
        default:
            log_error("unexpected combinational cell type %s\n", getCtx()->nameOf(cell->type));
        }
        // Note that this relationship won't hold for extended mode, when that is supported
        cell->combInfo.lut_bits_count = (1 << cell->combInfo.lut_input_count);
    }
    cell->combInfo.used_lut_input_count = 0;
    for (int i = 0; i < cell->combInfo.lut_input_count; i++)
        if (cell->combInfo.lut_in[i])
            ++cell->combInfo.used_lut_input_count;
}

void Arch::assign_ff_info(CellInfo *cell) const
{
    cell->ffInfo.ctrlset.clk = get_ctrlsig(getCtx(), cell, id_CLK);
    cell->ffInfo.ctrlset.ena = get_ctrlsig(getCtx(), cell, id_ENA, true);
    cell->ffInfo.ctrlset.aclr = get_ctrlsig(getCtx(), cell, id_ACLR);
    cell->ffInfo.ctrlset.sclr = get_ctrlsig(getCtx(), cell, id_SCLR);
    cell->ffInfo.ctrlset.sload = get_ctrlsig(getCtx(), cell, id_SLOAD);
    cell->ffInfo.sdata = get_net_or_empty(cell, id_SDATA);
    cell->ffInfo.datain = get_net_or_empty(cell, id_DATAIN);
}

// Validity checking functions
bool Arch::is_alm_legal(uint32_t lab, uint8_t alm) const
{
    auto &alm_data = labs.at(lab).alms.at(alm);
    // Get cells into an array for fast access
    std::array<const CellInfo *, 2> luts{getBoundBelCell(alm_data.lut_bels[0]), getBoundBelCell(alm_data.lut_bels[1])};
    std::array<const CellInfo *, 4> ffs{getBoundBelCell(alm_data.ff_bels[0]), getBoundBelCell(alm_data.ff_bels[1]),
                                        getBoundBelCell(alm_data.ff_bels[2]), getBoundBelCell(alm_data.ff_bels[3])};
    int used_lut_bits = 0;

    int total_lut_inputs = 0;
    // TODO: for more complex modes like extended/arithmetic, it might not always be possible for any LUT input to map
    // to any of the ALM half inputs particularly shared and extended mode will need more thought and probably for this
    // to be revisited
    for (int i = 0; i < 2; i++) {
        if (!luts[i])
            continue;
        total_lut_inputs += luts[i]->combInfo.lut_input_count;
        used_lut_bits += luts[i]->combInfo.lut_bits_count;
    }
    // An ALM only has 64 bits of storage. In theory some of these cases might be legal because of overlap between the
    // two functions, but the current placer is unlikely to stumble upon these cases frequently without anything to
    // guide it, and the cost of checking them here almost certainly outweighs any marginal benefit in supporting them,
    // at least for now.
    if (used_lut_bits > 64)
        return false;

    if (total_lut_inputs > 8) {
        NPNR_ASSERT(luts[0] && luts[1]); // something has gone badly wrong if this fails!
        // Make sure that LUT inputs are not overprovisioned
        int shared_lut_inputs = 0;
        // Even though this N^2 search looks inefficient, it's unlikely a set lookup or similar is going to be much
        // better given the low N.
        for (int i = 0; i < luts[1]->combInfo.lut_input_count; i++) {
            const NetInfo *sig = luts[1]->combInfo.lut_in[i];
            for (int j = 0; j < luts[0]->combInfo.lut_input_count; j++) {
                if (sig == luts[0]->combInfo.lut_in[j]) {
                    ++shared_lut_inputs;
                    break;
                }
            }
        }
        if ((total_lut_inputs - shared_lut_inputs) > 8)
            return false;
    }

    // For each ALM half; check FF control set sharing and input routeability
    for (int i = 0; i < 2; i++) {
        // There are two ways to route from the fabric into FF data - either routing through a LUT or using the E/F
        // signals and SLOAD=1 (*PKREF*)
        bool route_thru_lut_avail = !luts[i] && (total_lut_inputs < 8) && (used_lut_bits < 64);
        // E/F is available if this LUT is using 3 or fewer inputs - this is conservative and sharing can probably
        // improve this situation
        bool ef_available = (!luts[i] || luts[i]->combInfo.used_lut_input_count <= 3);
        // Control set checking
        bool found_ff = false;

        FFControlSet ctrlset;
        for (int j = 0; j < 2; j++) {
            const CellInfo *ff = ffs[i * 2 + j];
            if (!ff)
                continue;
            if (found_ff) {
                // Two FFs in the same half with an incompatible control set
                if (ctrlset != ff->ffInfo.ctrlset)
                    return false;
            } else {
                ctrlset = ff->ffInfo.ctrlset;
            }
            // SDATA must use the E/F input
            // TODO: rare case of two FFs with the same SDATA in the same ALM half
            if (ff->ffInfo.sdata) {
                if (!ef_available)
                    return false;
                ef_available = false;
            }
            // Find a way of routing the input through fabric, if it's not driven by the LUT
            if (ff->ffInfo.datain && (!luts[i] || (ff->ffInfo.datain != luts[i]->combInfo.comb_out))) {
                if (route_thru_lut_avail)
                    route_thru_lut_avail = false;
                else if (ef_available)
                    ef_available = false;
                else
                    return false;
            }
            found_ff = true;
        }
    }

    return true;
}

namespace {
bool check_assign_sig(ControlSig &sig_set, const ControlSig &sig)
{
    if (sig.net == nullptr) {
        return true;
    } else if (sig_set == sig) {
        return true;
    } else if (sig_set.net == nullptr) {
        sig_set = sig;
        return true;
    } else {
        return false;
    }
};

template <size_t N> bool check_assign_sig(std::array<ControlSig, N> &sig_set, const ControlSig &sig)
{
    if (sig.net == nullptr)
        return true;
    for (size_t i = 0; i < N; i++)
        if (sig_set[i] == sig) {
            return true;
        } else if (sig_set[i].net == nullptr) {
            sig_set[i] = sig;
            return true;
        }
    return false;
};
}; // namespace

bool Arch::is_lab_ctrlset_legal(uint32_t lab) const
{
    // Strictly speaking the constraint is up to 2 unique CLK and 3 CLK+ENA pairs. For now we simplify this to 1 CLK and
    // 3 ENA though.
    ControlSig clk{}, sload{}, sclr{};
    std::array<ControlSig, 2> aclr{};
    std::array<ControlSig, 3> ena{};

    for (uint8_t alm = 0; alm < 10; alm++) {
        for (uint8_t i = 0; i < 4; i++) {
            const CellInfo *ff = getBoundBelCell(labs.at(lab).alms.at(alm).ff_bels.at(i));
            if (ff == nullptr)
                continue;

            if (!check_assign_sig(clk, ff->ffInfo.ctrlset.clk))
                return false;
            if (!check_assign_sig(sload, ff->ffInfo.ctrlset.sload))
                return false;
            if (!check_assign_sig(sclr, ff->ffInfo.ctrlset.sclr))
                return false;
            if (!check_assign_sig(aclr, ff->ffInfo.ctrlset.aclr))
                return false;
            if (!check_assign_sig(ena, ff->ffInfo.ctrlset.ena))
                return false;
        }
    }

    // Check for overuse of the shared, LAB-wide datain signals
    std::array<ControlSig, 4> datain{};
    if (clk.net != nullptr && !clk.net->is_global)
        if (!check_assign_sig(datain[0], clk)) // CLK only needs DATAIN[0] if it's not global
            return false;
    if (!check_assign_sig(datain[1], sload))
        return false;
    if (!check_assign_sig(datain[3], sclr))
        return false;
    for (const auto &aclr_sig : aclr) {
        // Check both possibilities that ACLR can map to
        // TODO: ACLR could be global, too
        if (check_assign_sig(datain[3], aclr_sig))
            continue;
        if (check_assign_sig(datain[2], aclr_sig))
            continue;
        // Failed to find any free ACLR-capable DATAIN
        return false;
    }
    for (const auto &ena_sig : ena) {
        // Check all 3 possibilities that ACLR can map to
        // TODO: ACLR could be global, too
        if (check_assign_sig(datain[2], ena_sig))
            continue;
        if (check_assign_sig(datain[3], ena_sig))
            continue;
        if (check_assign_sig(datain[0], ena_sig))
            continue;
        // Failed to find any free ENA-capable DATAIN
        return false;
    }
    return true;
}

void Arch::lab_pre_route()
{
    log_info("Preparing LABs for routing...\n");
    for (uint32_t lab = 0; lab < labs.size(); lab++) {
        assign_control_sets(lab);
        for (uint8_t alm = 0; alm < 10; alm++) {
            reassign_alm_inputs(lab, alm);
        }
    }
}

void Arch::assign_control_sets(uint32_t lab)
{
    // TODO: set up reservations for checkPipAvailForNet for control set signals
    // This will be needed because clock and CE are routed together and must be kept together, there isn't free choice
    // e.g. CLK0 & ENA0 must be use for one control set, and CLK1 & ENA1 for another, they can't be mixed and matched
}

namespace {
// Gets the name of logical LUT pin i for a given cell
static IdString get_lut_pin(CellInfo *cell, int i)
{
    const std::array<IdString, 6> log_pins{id_A, id_B, id_C, id_D, id_E, id_F};
    const std::array<IdString, 5> log_pins_arith{id_A, id_B, id_C, id_D0, id_D1};
    return (cell->type == id_MISTRAL_ALUT_ARITH) ? log_pins_arith.at(i) : log_pins.at(i);
}

static void assign_lut6_inputs(CellInfo *cell, int lut)
{
    std::array<IdString, 6> phys_pins{id_A, id_B, id_C, id_D, (lut == 1) ? id_E1 : id_E0, (lut == 1) ? id_F1 : id_F0};
    int phys_idx = 0;
    for (int i = 0; i < 6; i++) {
        IdString log = get_lut_pin(cell, i);
        if (!cell->ports.count(log) || cell->ports.at(log).net == nullptr)
            continue;
        cell->pin_data[log].bel_pins.clear();
        cell->pin_data[log].bel_pins.push_back(phys_pins.at(++phys_idx));
    }
}
} // namespace

void Arch::reassign_alm_inputs(uint32_t lab, uint8_t alm)
{
    // Based on the usage of LUTs inside the ALM, set up cell-bel pin map for the combinational cells in the ALM
    // so that each physical bel pin is only used for one net; and the logical functions can be implemented correctly.
    // This function should also insert route-through LUTs to legalise flipflop inputs as needed.
    auto &alm_data = labs.at(lab).alms.at(alm);
    alm_data.l6_mode = false;
    std::array<CellInfo *, 2> luts{getBoundBelCell(alm_data.lut_bels[0]), getBoundBelCell(alm_data.lut_bels[1])};
    std::array<CellInfo *, 4> ffs{getBoundBelCell(alm_data.ff_bels[0]), getBoundBelCell(alm_data.ff_bels[1]),
                                  getBoundBelCell(alm_data.ff_bels[2]), getBoundBelCell(alm_data.ff_bels[3])};

    for (int i = 0; i < 2; i++) {
        // Currently we treat LUT6s as a special case, as they never share inputs
        if (luts[i] != nullptr && luts[i]->type == id_MISTRAL_ALUT6) {
            alm_data.l6_mode = true;
            NPNR_ASSERT(luts[1 - i] == nullptr); // only allow one LUT6 per ALM and no other LUTs
            assign_lut6_inputs(luts[i], i);
        }
    }

    if (!alm_data.l6_mode) {
        // In L5 mode; which is what we use in this case
        //  - A and B are shared
        //  - C, E0, and F0 are exclusive to the top LUT5 secion
        //  - D, E1, and F1 are exclusive to the bottom LUT5 section
        // First find up to two shared inputs
        std::unordered_map<IdString, int> shared_nets;
        if (luts[0] && luts[1]) {
            for (int i = 0; i < luts[0]->combInfo.lut_input_count; i++) {
                for (int j = 0; j < luts[1]->combInfo.lut_input_count; j++) {
                    if (luts[0]->combInfo.lut_in[i] == nullptr)
                        continue;
                    if (luts[0]->combInfo.lut_in[i] != luts[1]->combInfo.lut_in[j])
                        continue;
                    IdString net = luts[0]->combInfo.lut_in[i]->name;
                    if (shared_nets.count(net))
                        continue;
                    int idx = int(shared_nets.size());
                    shared_nets[net] = idx;
                    if (shared_nets.size() >= 2)
                        goto shared_search_done;
                }
            }
        shared_search_done:;
        }
        // A and B can be used for half-specific nets if not assigned to shared nets
        bool a_avail = shared_nets.size() == 0, b_avail = shared_nets.size() <= 1;
        // Do the actual port assignment
        for (int i = 0; i < 2; i++) {
            if (!luts[i])
                continue;
            // Work out which physical ports are available
            std::vector<IdString> avail_phys_ports;
            avail_phys_ports.push_back((i == 1) ? id_D : id_C);
            if (b_avail)
                avail_phys_ports.push_back(id_B);
            if (a_avail)
                avail_phys_ports.push_back(id_A);
            // In arithmetic mode, Ei can only be used for D0 and Fi can only be used for D1
            if (!luts[i]->combInfo.is_carry) {
                avail_phys_ports.push_back((i == 1) ? id_E1 : id_E0);
                avail_phys_ports.push_back((i == 1) ? id_F1 : id_F0);
            }
            int phys_idx = 0;

            for (int j = 0; j < luts[i]->combInfo.lut_input_count; j++) {
                IdString log = get_lut_pin(luts[i], j);
                auto &bel_pins = luts[i]->pin_data[log].bel_pins;
                bel_pins.clear();

                NetInfo *net = get_net_or_empty(luts[i], log);
                if (net == nullptr) {
                    // Disconnected inputs don't need to be allocated a pin, because the router won't be routing these
                    continue;
                } else if (shared_nets.count(net->name)) {
                    // This pin is to be allocated one of the shared nets
                    bel_pins.push_back(shared_nets.at(net->name) ? id_B : id_A);
                } else if (log == id_D0) {
                    // Arithmetic
                    bel_pins.push_back((i == 1) ? id_E1 : id_E0); // reserved
                } else if (log == id_D1) {
                    bel_pins.push_back((i == 1) ? id_F1 : id_F0); // reserved
                } else {
                    // Allocate from the general pool of available physical pins
                    IdString phys = avail_phys_ports.at(phys_idx++);
                    bel_pins.push_back(phys);
                    // Mark A/B unavailable for the other LUT, if needed
                    if (phys == id_A)
                        a_avail = false;
                    else if (phys == id_B)
                        b_avail = false;
                }
            }
        }
    }

    // FF route-through insertion
    for (int i = 0; i < 2; i++) {
        // FF route-through will never be inserted if LUT is used
        if (luts[i])
            continue;
        for (int j = 0; j < 2; j++) {
            CellInfo *ff = ffs[i * 2 + j];
            if (!ff || !ff->ffInfo.datain)
                continue;
            CellInfo *rt_lut = createCell(id(stringf("%s$ROUTETHRU", nameOf(ff))), id_MISTRAL_BUF);
            rt_lut->addInput(id_A);
            rt_lut->addOutput(id_Q);
            // Disconnect the original data input to the FF, and connect it to the route-thru LUT instead
            NetInfo *datain = get_net_or_empty(ff, id_DATAIN);
            disconnect_port(getCtx(), ff, id_DATAIN);
            connect_port(getCtx(), datain, rt_lut, id_A);
            connect_ports(getCtx(), rt_lut, id_Q, ff, id_DATAIN);
            // Assign route-thru LUT physical ports, input goes to the first half-specific input
            NPNR_ASSERT(!alm_data.l6_mode);
            rt_lut->pin_data[id_A].bel_pins.push_back(i ? id_D : id_C);
            rt_lut->pin_data[id_Q].bel_pins.push_back(id_COMBOUT);
            assign_comb_info(rt_lut);
            // Place the route-thru LUT at the relevant combinational bel
            bindBel(alm_data.lut_bels[i], rt_lut, STRENGTH_STRONG);
            break;
        }
    }

    // TODO: in the future, as well as the reassignment here we will also have pseudo PIPs in front of the ALM so that
    // the router can permute LUTs for routeability; too. Here we will need to lock out some of those PIPs depending on
    // the usage of the ALM, as not all inputs are always interchangeable.
    // Get cells into an array for fast access
}

// This default cell-bel pin mapping is used to provide estimates during placement only. It will have errors and
// overlaps and a correct mapping will be resolved twixt placement and routing
const std::unordered_map<IdString, IdString> Arch::comb_pinmap = {
        {id_A, id_F0}, // fastest input first
        {id_B, id_E0}, {id_C, id_D}, {id_D, id_C},       {id_D0, id_C},       {id_D1, id_B},
        {id_E, id_B},  {id_F, id_A}, {id_Q, id_COMBOUT}, {id_SO, id_COMBOUT},
};

namespace {
// gets the value of the ith LUT init property of a given cell
uint64_t get_lut_init(const CellInfo *cell, int i)
{
    if (cell->type == id_MISTRAL_NOT) {
        return 1;
    } else if (cell->type == id_MISTRAL_BUF) {
        return 2;
    } else {
        IdString prop;
        if (cell->type == id_MISTRAL_ALUT_ARITH)
            prop = (i == 1) ? id_LUT1 : id_LUT0;
        else
            prop = id_LUT;
        auto fnd = cell->params.find(prop);
        if (fnd == cell->params.end())
            return 0;
        else
            return fnd->second.as_int64();
    }
}
// gets the state of a physical pin when evaluating the a given bit of LUT init for
bool get_phys_pin_val(bool l6_mode, bool arith_mode, int bit, IdString pin)
{
    switch (pin.index) {
    case ID_A:
        return (bit >> 0) & 0x1;
    case ID_B:
        return (bit >> 1) & 0x1;
    case ID_C:
        return (l6_mode && bit >= 32) ? ((bit >> 3) & 0x1) : ((bit >> 2) & 0x1);
    case ID_D:
        return (l6_mode && bit < 32) ? ((bit >> 3) & 0x1) : ((bit >> 2) & 0x1);
    case ID_E0:
    case ID_E1:
        return l6_mode ? ((bit >> 5) & 0x1) : ((bit >> 3) & 0x1);
    case ID_F0:
    case ID_F1:
        return arith_mode ? ((bit >> 3) & 0x1) : ((bit >> 4) & 0x1);
    default:
        NPNR_ASSERT_FALSE("unknown physical pin!");
    }
}
} // namespace

uint64_t Arch::compute_lut_mask(uint32_t lab, uint8_t alm)
{
    uint64_t mask = 0;
    auto &alm_data = labs.at(lab).alms.at(alm);
    std::array<CellInfo *, 2> luts{getBoundBelCell(alm_data.lut_bels[0]), getBoundBelCell(alm_data.lut_bels[1])};

    for (int i = 0; i < 2; i++) {
        CellInfo *lut = luts[i];
        if (!lut)
            continue;
        int offset = ((i == 1) && !alm_data.l6_mode) ? 32 : 0;
        bool arith = lut->combInfo.is_carry;
        for (int j = 0; j < (alm_data.l6_mode ? 64 : 32); j++) {
            // Evaluate LUT function at this point
            uint64_t init = get_lut_init(lut, (arith && j >= 16) ? 1 : 0);
            int index = 0;
            for (int k = 0; k < lut->combInfo.lut_input_count; k++) {
                IdString log_pin = get_lut_pin(lut, k);
                CellPinState state = lut->get_pin_state(log_pin);
                if (state == PIN_0)
                    continue;
                else if (state == PIN_1)
                    index |= (1 << k);
                // Ignore if no associated physical pin
                if (get_net_or_empty(lut, log_pin) == nullptr || lut->pin_data.at(log_pin).bel_pins.empty())
                    continue;
                // ALM inputs appear to be inverted by default (TODO: check!)
                // so only invert if an inverter has _not_ been folded into the pin
                bool inverted = (state != PIN_INV);
                // Depermute physical pin
                IdString phys_pin = lut->pin_data.at(log_pin).bel_pins.at(0);
                if (get_phys_pin_val(alm_data.l6_mode, arith, j, phys_pin) != inverted)
                    index |= (1 << k);
            }
            if ((init >> index) & 0x1) {
                mask |= (1ULL << uint64_t(j + offset));
            }
        }
    }

    // TODO: always inverted, or just certain paths?
    mask = ~mask;

#if 1
    if (getCtx()->debug) {
        auto pos = alm_data.lut_bels[0].pos;
        log("ALM %03d.%03d.%d\n", CycloneV::pos2x(pos), CycloneV::pos2y(pos), alm);
        for (int i = 0; i < 2; i++) {
            log("    LUT%d: ", i);
            if (luts[i]) {
                log("%s:%s", nameOf(luts[i]), nameOf(luts[i]->type));
                for (auto &pin : luts[i]->pin_data) {
                    if (!luts[i]->ports.count(pin.first) || luts[i]->ports.at(pin.first).type != PORT_IN)
                        continue;
                    log(" %s:", nameOf(pin.first));
                    if (pin.second.state == PIN_0)
                        log("0");
                    else if (pin.second.state == PIN_1)
                        log("1");
                    else if (pin.second.state == PIN_INV)
                        log("~");
                    for (auto bp : pin.second.bel_pins)
                        log("%s", nameOf(bp));
                }
            } else {
                log("<null>");
            }
            log("\n");
        }
        log("INIT: %016lx\n", mask);
        log("\n");
    }
#endif

    return mask;
}

NEXTPNR_NAMESPACE_END