Open PDBF Benchmark Suite

GT PDBF Logic Synthesis Benchmark Suite

Open benchmark specifications for logic synthesis of Partially Defined Boolean Functions (PDBFs), with published GT gate count, logic depth and synthesis runtime.

Open synthesis challenge: download any PLA specification, synthesize it with your preferred combinational logic synthesis tool, and compare your verified result with the published GT result.

Purpose of the benchmark suite

This page provides reproducible PDBF synthesis problems in standard PLA format. The benchmark suite collects experimental families used across the GT paper series: Contextual RTL and open-source RTL candidate studies associated with Paper II, real-RTL legal-completion validation associated with Paper III, memory-oriented PDBF experiments associated with Paper IV, and large-input scalability experiments associated with Paper VI. Additional semantic and AI-inspired PDBF benchmarks are provided as supporting open synthesis challenges.

GT synthesis results are reported as simple two-input logic-gate count, maximum logic depth, and synthesis time. The present benchmark release provides the original PLA specifications and synthesis statistics. GT synthesized netlists are not included in this release.

PDBF model. A PDBF defines required Boolean behavior on its care domain and leaves the remaining behavior free. Each PLA row associates an input cube xi ∈ {0,1,−}n with an output pattern yi ∈ {0,1,−}m. In an input cube, 0 and 1 constrain an input bit, while “−” means that either input value is accepted. In the output pattern, 0 and 1 specify required behavior, while “−” denotes an unconstrained output. If every output for an input cube is “−”, that cube imposes no behavioral constraint and is therefore omitted from the PDBF PLA. Input/output behavior not constrained by the retained PLA rows is free and may be assigned by any Legal Completion.

AI-Inspired Contextual PDBF Benchmarks

These benchmarks are synthetic, reproducible AI-like decision and classification functions expressed directly as contextual PDBFs. They are intended to study combinational synthesis over strongly constrained quantized feature domains; they are not results extracted from trained neural-network models and should not be interpreted as AI-accuracy benchmarks.

This table reports 21 benchmarks from the updated result listing. G / L denotes gate count / maximum logic levels. PI/PO/Cubes gives the input count, output count and number of specified care terms.

BenchmarkPI/PO/CubesABC
decpla
G / L
ABC
deepsyn
G / L
ABC
transtoch
G / L
GT
G / L
GT → ABC
deepsyn
G / L
GT → ABC
transtoch
G / L
PLA
AI_autonomous_braking21 / 4 / 2,18723 / 620 / 5—20 / 5——PLA
AI_battery_management22 / 5 / 2,30426 / 523 / 5—20 / 7
23 / 6
18 / 7
20 / 5
—PLA
AI_crop_health24 / 6 / 5,18423 / 5—16 / 618 / 5—16 / 6PLA
AI_data_center_cooling21 / 6 / 1,72826 / 6—20 / 521 / 7—20 / 5PLA
AI_drone_obstacle24 / 6 / 6,56134 / 530 / 527 / 7
26 / 10
28 / 9
28 / 10
31 / 7
27 / 6
28 / 5
24 / 7PLA
AI_ecg_arrhythmia18 / 4 / 7298 / 4—8 / 48 / 4——PLA
AI_financial_fraud21 / 5 / 2,18731 / 6—21 / 521 / 5——PLA
AI_food_quality22 / 6 / 2,30421 / 520 / 5—20 / 5——PLA
AI_hand_gesture24 / 4 / 6,56112 / 5—12 / 512 / 5
14 / 4
——PLA
AI_jet_engine_health23 / 6 / 3,07267 / 1047 / 744 / 1131 / 14—30 / 8PLA
AI_keyword_spotting20 / 4 / 1,0249 / 5——9 / 5——PLA
AI_malware_behavior24 / 6 / 5,18453 / 930 / 632 / 725 / 13
27 / 6
25 / 6—PLA
AI_maritime_navigation24 / 7 / 4,09652 / 936 / 1245 / 7
41 / 9
32 / 9
34 / 8
39 / 7
31 / 7
32 / 6
30 / 10PLA
AI_medical_triage24 / 6 / 6,56132 / 524 / 523 / 625 / 824 / 523 / 6PLA
AI_predictive_maintenance24 / 6 / 6,56159 / 942 / 6
31 / 13
34 / 831 / 10
32 / 9
35 / 8
—26 / 7
28 / 6
PLA
AI_railway_signal_anomaly24 / 6 / 5,18446 / 830 / 734 / 631 / 8
31 / 10
34 / 8
52 / 7
49 / 728 / 9
28 / 8
27 / 11
PLA
AI_road_scene23 / 6 / 3,07235 / 7—31 / 7
30 / 10
28 / 8
35 / 5
28 / 9
34 / 6
27 / 7
31 / 6
—PLA
AI_sepsis_risk22 / 6 / 2,30444 / 6—28 / 827 / 8
30 / 7
—26 / 6PLA
AI_sonar_target22 / 6 / 2,30436 / 6—23 / 5
22 / 6
19 / 12
20 / 10
21 / 7
26 / 5
—25 / 6PLA
AI_wafer_defect24 / 5 / 4,09626 / 619 / 518 / 618 / 6
26 / 5
25 / 5—PLA
AI_wildfire_risk22 / 6 / 2,30432 / 734 / 529 / 627 / 12
28 / 10
31 / 9
32 / 7
34 / 8
—26 / 7
26 / 5
PLA

ABC decpla starts from the PDBF PLA; ABC deepsyn and transtoch report optimized decpla implementations. GT reports the original GT implementations; GT → ABC reports ABC optimization of GT netlists. Multiple entries retain all supplied implementations and their gate/depth tradeoffs; entries in different columns need not correspond by line. Hover over a GT → ABC result to see its source GT gate/level pair and filename. Dashes mean no result was supplied. Times were not supplied. Values are transcribed from filenames, not independently remeasured. Display names omit _contextual.

Medium-input semantic PDBF benchmarks

In this benchmark suite, examples with hundreds primary inputs are classified as medium-input PDBFs. The cases below model networking, security, storage, telecommunications and policy/control hardware.

BenchmarkPI/PO/CubesABC
decpla
G / L
ABC
deepsyn
G / L
ABC
Time, s
GT
G / L
GT
Time, s
PLA
ipv6_5g_upf_gtpu_policy299 / 18 / 16,3842,623 / 221,677 / 172,862151 / 23
236 / 21
210
337
PLA
ipv6_bgp_evpn_route_policy398 / 18 / 16,38416 / 316 / 327214 / 31PLA
ipv6_dns_doh_security_policy317 / 18 / 16,3841,887 / 18930 / 161,509118 / 23238PLA
ipv6_firewall_policy_classifier96 / 14 / 8,19223,139 / 2819,010 / 196,8762,383 / 209
2,542 / 102
8,286
12,056
PLA
ipv6_geneve_nsh_service_chain_policy257 / 17 / 16,3842,052 / 18728 / 161,959641 / 5017,070PLA
ipv6_ids_ips_signature_policy300 / 18 / 16,3843,734 / 222,223 / 175,102231 / 25601PLA
ipv6_ipsec_sa_security_policy216 / 18 / 16,3845,749 / 223,304 / 177,061627 / 62
1,009 / 39
9,085
16,082
PLA
ipv6_mpls_srv6_interworking_policy294 / 18 / 16,38410,365 / 277,751 / 217,103393 / 261,240PLA
ipv6_nvmeof_storage_policy286 / 18 / 16,3841,404 / 17767 / 131,312234 / 22596PLA
ipv6_ptp_tsn_policy219 / 18 / 16,3841,437 / 18771 / 151,348216 / 36415PLA
ipv6_quic_ddos_mitigation_policy313 / 18 / 16,3842,378 / 241,355 / 274,702224 / 18
224 / 26
243
123
PLA
ipv6_security_telemetry_policy_medium246 / 16 / 8,192198 / 10132 / 8112105 / 21
106 / 18
118 / 16
19
13
21
PLA
ipv6_service_chain_security_policy_narrow221 / 16 / 8,1922,364 / 221,286 / 172,749445 / 46
459 / 26
1,421
326
PLA
ipv6_service_mesh_load_balancer_policy263 / 18 / 16,3849,677 / 256,439 / 189,6861,133 / 6615,048PLA
ipv6_storage_replication_erasure_policy253 / 17 / 16,3842,148 / 19780 / 162,258326 / 313,076PLA
ipv6_vxlan_microsegmentation_ct_policy212 / 18 / 12,2881,677 / 20963 / 151,647328 / 25178PLA

Results are reported as AIG two-input AND gates / maximum AIG levels. Multiple GT entries are alternative implementations, with times in matching order. ABC adjusted time is the supplied total decpla + deepsyn runtime divided by 8 and rounded to the nearest whole second (half seconds round up); it is not the unadjusted measured runtime. GT times are unchanged. Separate decpla runtimes were not supplied. Results are transcribed from the supplied filenames. The ABC deepsyn column reports the result after decpla followed by deepsyn. Display names omit the _semantic suffix. Logic depth is a structural metric, not physical delay.

Randomly generated PDBFs: cube quantity to netlist parameters comparison

The first eight benchmarks keep the size fixed at 250 primary inputs and 1 primary output while increasing the number of PDBF cubes from 16 to 2,000. An additional benchmark has 250 primary inputs, 100 primary outputs and 2,000 cubes. G / L denotes gate count / maximum logic levels.

BenchmarkPI/PO/CubesABC
decpla
G / L
ABC
deepsyn
G / L
ABC
transtoch
G / L
GT
G / L
GT
Time, s
PLA
E250_1_16250 / 1 / 161 / 11 / 11 / 11 / 1—PLA
E250_1_32250 / 1 / 324 / 34 / 34 / 34 / 3—PLA
E250_1_64250 / 1 / 6414 / 614 / 6—8 / 40.010PLA
E250_1_128250 / 1 / 12869 / 949 / 1047 / 1719 / 60.004PLA
E250_1_256250 / 1 / 256171 / 13102 / 10—40 / 80.005PLA
E250_1_500250 / 1 / 500319 / 16227 / 12
223 / 14
—92 / 18
93 / 11
1
0.670
PLA
E250_1_1000250 / 1 / 1,000636 / 19477 / 14—194 / 22
220 / 12
17
0.871
PLA
E250_1_2000250 / 1 / 2,0001,266 / 22
1,290 / 23
978 / 16
980 / 15
—392 / 35
460 / 14
72
4
PLA
Example_250_100_2000250 / 100 / 2,000128,569 / 24109,108 / 18—25,508 / 366
38,143 / 84
38,701 / 91
50,866 / 16
6,543
1,131
1,191
2,108
PLA

ABC decpla reports synthesis from the PLA; deepsyn and transtoch report subsequent ABC optimization. Multiple pairs preserve the supplied alternatives. Entries in different columns need not correspond by line or branch. A dash indicates no result supplied. GT times are in seconds and correspond line by line to the GT gate/level pairs; millisecond runtimes have been converted to seconds.

Paper VI — PDBF Input-Scalability Challenge

This Paper VI benchmark series studies PDBF synthesis as the primary-input dimension grows from ten thousand to hundreds of thousands and one million inputs.

Benchmark PI PO Cubes GT Gates GT Levels GT Time, s PLA
Example_10000_1000_5000 10,000 1,000 5,000 769,799 47 26,875 PLA
Ex100000_1_1000 100,000 1 1,000 89 15 63 PLA
Ex_1000000_1_100 1,000,000 1 100 6 4 4 PLA

The results illustrate that primary-input count alone does not determine synthesis difficulty. The care structure, number of outputs, cube structure and interactions among conditions can be equally important.

Reported synthesis metrics

PI — number of primary inputs. PO — number of primary outputs. Cubes — number of PLA product terms.

For the GT results, Gates denotes the number of internal two-input simple/AIG logic nodes. Complemented AIG edges do not count as additional gates. Levels is the maximum logic depth from a primary input to a primary output.

GT Time is the reported synthesis runtime in seconds for the corresponding experiment.

Verification rules and acceptance criteria

A synthesized network is a valid implementation of a benchmark only if it satisfies every specified PDBF output value over the complete care domain of the PLA.

  1. For every PLA cube and every output specified as 0, the synthesized output must be 0 for every concrete binary input assignment covered by that cube.
  2. For every PLA cube and every output specified as 1, the synthesized output must be 1 for every concrete binary input assignment covered by that cube.
  3. An output value - is unrestricted. A term for which all outputs are - imposes no PDBF constraint and should not be included in the effective truth table.
  4. Verification must cover the complete input region represented by a cube, not only one representative minterm. If a PLA input is - and belongs to the effective support of the synthesized output, all relevant assignments of that free support variable must be checked or proven equivalent.
  5. If all primary inputs in the effective support of an output are fixed within a PLA cube, evaluation of that support assignment proves the output over the complete cube; primary inputs outside the output support cannot affect the result.
  6. Two overlapping cubes may not impose contradictory values on the same output. If two terms require opposite values, 0/1 or 1/0, for an output, their input cubes must be disjoint. Therefore at least one input position must contain an explicit complementary pair 0/1 or 1/0.
  7. Behavior outside the PDBF care domain is unrestricted. Two valid synthesized networks do not have to be globally equivalent if they make different choices in don't-care regions.
Verification status: a result should be described as VERIFIED only when the complete PDBF care domain has been checked or proven against the synthesized network. Simulation of selected test vectors alone should be described as sampled or unverified.

Results obtained with other synthesis tools

We welcome independent experiments with these benchmarks. If you obtain a synthesis result using another combinational logic synthesis tool, please inform us of the result and the synthesis conditions.

For meaningful comparison, please identify the tool and version, synthesis commands/options, hardware platform, runtime, peak memory if available, gate/node count and logic depth. The resulting implementation can be validated against the original PLA using the verification criteria above.

A valid result with fewer gates, fewer levels, or competitive synthesis time is welcome.

Input-size terminology used in this suite

The terms below are benchmark-suite classifications rather than claims of universal EDA terminology:

ClassPrimary inputs
Small< 200
Medium200 – 9,999
Large10,000 – 99,999
Very large100,000 – 999,999
Extreme≥ 1,000,000