Decimal128 by Miguel

C Benchmark Results

IEEE 754-2019 compliant decimal128 high-performance software solution created by Miguel.

Living document — as-measured results. Category codes, profiles, modes & method: Benchmark Key.

This is the C view of decimal128 as-measured, band by band, with explicit ratios. It opens with the realistic financial-mix (P-fin) headline, then the per-operation band characterization (P-gen) and FMA. In C, d128 is measured against the Intel libbid universal reference plus IBM decQuad (DPD) and libmpdecimal. It is data only — the categories, magnitude profiles, units, and methodology are defined in the Benchmark Key (and, authoritatively, BenchmarkMatrix.md). The cross-port d128 band-shape matrices (all ports, no alternatives) live in Port-Comparison Benchmark Results; the full index of per-language pages is on the Benchmarks hub.

Summary — Ratio Range by Operation

Each row below is the ratio for that reference library on x86_64 (Intel i9-9880H): ratio = libbid / Miguel, ratio = decQuad / Miguel, or ratio = mpdecimal / Miguel (> 1× ⇒ d128 faster), broken out by operation.

  Add Subtract Multiply Divide
ratio = libbid / Miguel 6× 5× 1.7× – 18× 0.9× – 2.8×
ratio = decQuad / Miguel 8× 14× 2.0× – 21× 2.2× – 9×
ratio = mpdecimal / Miguel 7× 6× 1.2× – 12× 2.5× – 14×

FinMix — realistic financial mix (P-fin)

The headline: one realistic 64-bit financial operation mix — a MIX add/sub stream, mul CP/WP, div CD/WD/ET/PT — ratio = alt / Miguel (> 1 ⇒ d128 faster). This is the profile closest to real financial code.

arm64 (M3 Pro).

port op cat profile arch mode ns alt alt ns ratio run notes
c add MIX P-fin arm64 thru 1.58 libbid 10.72 6.78× Rc2  
c sub MIX P-fin arm64 thru 2.07 libbid 11.80 5.70× Rc2  
c mul CP P-fin arm64 thru 1.33 libbid 23.57 17.72× Rc2  
c mul WP P-fin arm64 thru 20.54 libbid 34.52 1.68× Rc2  
c div CD P-fin arm64 thru 30.56 libbid 35.07 1.15× Rc2  
c div WD P-fin arm64 thru 41.05 libbid 40.37 0.98× Rc2  
c div ET P-fin arm64 thru 6.34 libbid 6.09 0.96× Rc2  
c div PT P-fin arm64 thru 3.13 libbid 6.09 1.95× Rc2  
c add MIX P-fin arm64 thru 1.58 decQuad 13.90 8.80× Rc2  
c sub MIX P-fin arm64 thru 2.07 decQuad 30.55 14.76× Rc2  
c mul CP P-fin arm64 thru 1.33 decQuad 21.42 16.11× Rc2  
c mul WP P-fin arm64 thru 20.54 decQuad 26.06 1.27× Rc2  
c div CD P-fin arm64 thru 30.56 decQuad 73.26 2.40× Rc2  
c div WD P-fin arm64 thru 41.05 decQuad 122.46 2.98× Rc2  
c div ET P-fin arm64 thru 6.34 decQuad 41.42 6.53× Rc2  
c div PT P-fin arm64 thru 3.13 decQuad 39.32 12.56× Rc2  
c add MIX P-fin arm64 thru 1.58 mpdecimal 13.03 8.25× Rc2  
c sub MIX P-fin arm64 thru 2.07 mpdecimal 14.20 6.86× Rc2  
c mul CP P-fin arm64 thru 1.33 mpdecimal 10.35 7.78× Rc2  
c mul WP P-fin arm64 thru 20.54 mpdecimal 30.01 1.46× Rc2  
c div CD P-fin arm64 thru 30.56 mpdecimal 62.34 2.04× Rc2  
c div WD P-fin arm64 thru 41.05 mpdecimal 90.11 2.20× Rc2  
c div ET P-fin arm64 thru 6.34 mpdecimal 57.78 9.11× Rc2  
c div PT P-fin arm64 thru 3.13 mpdecimal 45.93 14.67× Rc2  

x86_64 (Intel i9-9880H).

port op cat profile arch mode ns alt alt ns ratio run notes
c add MIX P-fin x86_64 thru 4.96 libbid 30.56 6.16× xRc2  
c sub MIX P-fin x86_64 thru 5.92 libbid 31.87 5.38× xRc2  
c mul CP P-fin x86_64 thru 2.82 libbid 50.91 18.05× xRc2  
c mul WP P-fin x86_64 thru 38.05 libbid 66.63 1.75× xRc2  
c div CD P-fin x86_64 thru 66.61 libbid 82.11 1.23× xRc2  
c div WD P-fin x86_64 thru 94.04 libbid 86.90 0.92× xRc2  
c div ET P-fin x86_64 thru 17.24 libbid 21.25 1.23× xRc2  
c div PT P-fin x86_64 thru 7.51 libbid 21.42 2.85× xRc2  
c add MIX P-fin x86_64 thru 4.96 decQuad 39.88 8.04× xRc2  
c sub MIX P-fin x86_64 thru 5.92 decQuad 86.38 14.59× xRc2  
c mul CP P-fin x86_64 thru 2.82 decQuad 61.25 21.72× xRc2  
c mul WP P-fin x86_64 thru 38.05 decQuad 76.73 2.02× xRc2  
c div CD P-fin x86_64 thru 66.61 decQuad 147.02 2.21× xRc2  
c div WD P-fin x86_64 thru 94.04 decQuad 265.55 2.82× xRc2  
c div ET P-fin x86_64 thru 17.24 decQuad 81.75 4.74× xRc2  
c div PT P-fin x86_64 thru 7.51 decQuad 73.24 9.75× xRc2  
c add MIX P-fin x86_64 thru 4.96 mpdecimal 35.65 7.19× xRc2  
c sub MIX P-fin x86_64 thru 5.92 mpdecimal 40.06 6.77× xRc2  
c mul CP P-fin x86_64 thru 2.82 mpdecimal 34.58 12.26× xRc2  
c mul WP P-fin x86_64 thru 38.05 mpdecimal 47.63 1.25× xRc2  
c div CD P-fin x86_64 thru 66.61 mpdecimal 169.56 2.55× xRc2  
c div WD P-fin x86_64 thru 94.04 mpdecimal 297.68 3.17× xRc2  
c div ET P-fin x86_64 thru 17.24 mpdecimal 151.08 8.76× xRc2  
c div PT P-fin x86_64 thru 7.51 mpdecimal 105.39 14.03× xRc2  

Add — SQ · NQ · MQ · OQ · FQ

Swept 4096-input average per band (bare thru; ns/op = Time/4096 over the shared decimal128-resources/swept/ corpus, byte-identical operands every port). Relational peer table with explicit ratios; the cross-port d128 band-shape matrices are in Port-Comparison Benchmark Results.

arm64 (M3 Pro).

port op cat profile arch mode ours alt alt ns ratio run notes
c add SQss P-gen arm64 thru 0.84 libbid 7.96 9.48× Rc2  
c add SQos P-gen arm64 thru 2.05 libbid 8.69 4.24× Rc2  
c add NQss P-gen arm64 thru 3.54 libbid 9.36 2.64× Rc2  
c add NQos P-gen arm64 thru 4.07 libbid 9.78 2.40× Rc2  
c add MQss P-gen arm64 thru 5.68 libbid 9.75 1.72× Rc2  
c add MQos P-gen arm64 thru 9.34 libbid 9.71 1.04× Rc2  
c add OQss P-gen arm64 thru 11.45 libbid 13.66 1.19× Rc2  
c add OQos P-gen arm64 thru 16.60 libbid 15.31 0.92× Rc2  
c add FQss P-gen arm64 thru 8.03 libbid 9.32 1.16× Rc2  
c add FQos P-gen arm64 thru 10.10 libbid 10.38 1.03× Rc2  
c add SQss P-gen arm64 thru 0.84 decQuad 7.64 9.10× Rc2  
c add SQos P-gen arm64 thru 2.05 decQuad 33.13 16.16× Rc2  
c add NQss P-gen arm64 thru 3.54 decQuad 27.01 7.63× Rc2  
c add NQos P-gen arm64 thru 4.07 decQuad 35.59 8.74× Rc2  
c add MQss P-gen arm64 thru 5.68 decQuad 24.73 4.35× Rc2  
c add MQos P-gen arm64 thru 9.34 decQuad 31.37 3.36× Rc2  
c add OQss P-gen arm64 thru 11.45 decQuad 33.69 2.94× Rc2  
c add OQos P-gen arm64 thru 16.60 decQuad 37.34 2.25× Rc2  
c add FQss P-gen arm64 thru 8.03 decQuad 27.06 3.37× Rc2  
c add FQos P-gen arm64 thru 10.10 decQuad 26.86 2.66× Rc2  
c add SQss P-gen arm64 thru 0.84 mpdecimal 11.83 14.08× Rc2  
c add SQos P-gen arm64 thru 2.05 mpdecimal 13.58 6.62× Rc2  
c add NQss P-gen arm64 thru 3.54 mpdecimal 24.24 6.85× Rc2  
c add NQos P-gen arm64 thru 4.07 mpdecimal 28.27 6.95× Rc2  
c add MQss P-gen arm64 thru 5.68 mpdecimal 20.03 3.53× Rc2  
c add MQos P-gen arm64 thru 9.34 mpdecimal 25.96 2.78× Rc2  
c add OQss P-gen arm64 thru 11.45 mpdecimal 46.76 4.08× Rc2  
c add OQos P-gen arm64 thru 16.60 mpdecimal 48.72 2.93× Rc2  
c add FQss P-gen arm64 thru 8.03 mpdecimal 39.81 4.96× Rc2  
c add FQos P-gen arm64 thru 10.10 mpdecimal 42.48 4.21× Rc2  

x86_64 (Intel i9-9880H).

port op cat profile arch mode ours alt alt ns ratio run notes
c add SQss P-gen x86_64 thru 2.28 libbid 30.12 13.21× xRc2  
c add SQos P-gen x86_64 thru 5.09 libbid 27.81 5.46× xRc2  
c add NQss P-gen x86_64 thru 10.13 libbid 31.74 3.13× xRc2  
c add NQos P-gen x86_64 thru 10.32 libbid 30.17 2.92× xRc2  
c add MQss P-gen x86_64 thru 13.01 libbid 29.55 2.27× xRc2  
c add MQos P-gen x86_64 thru 24.27 libbid 28.78 1.19× xRc2  
c add OQss P-gen x86_64 thru 30.45 libbid 46.76 1.54× xRc2  
c add OQos P-gen x86_64 thru 38.89 libbid 47.44 1.22× xRc2  
c add FQss P-gen x86_64 thru 19.46 libbid 33.02 1.70× xRc2  
c add FQos P-gen x86_64 thru 26.44 libbid 32.54 1.23× xRc2  
c add SQss P-gen x86_64 thru 2.28 decQuad 21.23 9.31× xRc2  
c add SQos P-gen x86_64 thru 5.09 decQuad 87.00 17.09× xRc2  
c add NQss P-gen x86_64 thru 10.13 decQuad 75.37 7.44× xRc2  
c add NQos P-gen x86_64 thru 10.32 decQuad 91.17 8.83× xRc2  
c add MQss P-gen x86_64 thru 13.01 decQuad 71.37 5.49× xRc2  
c add MQos P-gen x86_64 thru 24.27 decQuad 85.48 3.52× xRc2  
c add OQss P-gen x86_64 thru 30.45 decQuad 87.09 2.86× xRc2  
c add OQos P-gen x86_64 thru 38.89 decQuad 100.65 2.59× xRc2  
c add FQss P-gen x86_64 thru 19.46 decQuad 71.68 3.68× xRc2  
c add FQos P-gen x86_64 thru 26.44 decQuad 76.01 2.87× xRc2  
c add SQss P-gen x86_64 thru 2.28 mpdecimal 31.16 13.67× xRc2  
c add SQos P-gen x86_64 thru 5.09 mpdecimal 35.74 7.02× xRc2  
c add NQss P-gen x86_64 thru 10.13 mpdecimal 52.39 5.17× xRc2  
c add NQos P-gen x86_64 thru 10.32 mpdecimal 54.55 5.29× xRc2  
c add MQss P-gen x86_64 thru 13.01 mpdecimal 52.10 4.00× xRc2  
c add MQos P-gen x86_64 thru 24.27 mpdecimal 53.88 2.22× xRc2  
c add OQss P-gen x86_64 thru 30.45 mpdecimal 136.07 4.47× xRc2  
c add OQos P-gen x86_64 thru 38.89 mpdecimal 144.17 3.71× xRc2  
c add FQss P-gen x86_64 thru 19.46 mpdecimal 92.14 4.73× xRc2  
c add FQos P-gen x86_64 thru 26.44 mpdecimal 96.21 3.64× xRc2  

Subtract — SQ · NQ · MQ · OQ · FQ

Swept 4096-input average per band (bare thru; ns/op = Time/4096 over the shared decimal128-resources/swept/ corpus, byte-identical operands every port). Relational peer table with explicit ratios; the cross-port d128 band-shape matrices are in Port-Comparison Benchmark Results.

arm64 (M3 Pro).

port op cat profile arch mode ours alt alt ns ratio run notes
c sub SQss P-gen arm64 thru 1.51 libbid 9.68 6.41× Rc2  
c sub SQos P-gen arm64 thru 0.93 libbid 10.04 10.80× Rc2  
c sub NQss P-gen arm64 thru 4.06 libbid 10.87 2.68× Rc2  
c sub NQos P-gen arm64 thru 3.46 libbid 11.42 3.30× Rc2  
c sub MQss P-gen arm64 thru 9.73 libbid 9.94 1.02× Rc2  
c sub MQos P-gen arm64 thru 5.59 libbid 9.78 1.75× Rc2  
c sub OQss P-gen arm64 thru 17.20 libbid 15.79 0.92× Rc2  
c sub OQos P-gen arm64 thru 11.85 libbid 14.84 1.25× Rc2  
c sub FQss P-gen arm64 thru 10.70 libbid 9.39 0.88× Rc2  
c sub FQos P-gen arm64 thru 8.95 libbid 9.48 1.06× Rc2  
c sub SQss P-gen arm64 thru 1.51 decQuad 34.54 22.87× Rc2  
c sub SQos P-gen arm64 thru 0.93 decQuad 8.17 8.78× Rc2  
c sub NQss P-gen arm64 thru 4.06 decQuad 36.03 8.87× Rc2  
c sub NQos P-gen arm64 thru 3.46 decQuad 28.26 8.17× Rc2  
c sub MQss P-gen arm64 thru 9.73 decQuad 33.32 3.42× Rc2  
c sub MQos P-gen arm64 thru 5.59 decQuad 26.46 4.73× Rc2  
c sub OQss P-gen arm64 thru 17.20 decQuad 39.30 2.28× Rc2  
c sub OQos P-gen arm64 thru 11.85 decQuad 34.31 2.90× Rc2  
c sub FQss P-gen arm64 thru 10.70 decQuad 27.74 2.59× Rc2  
c sub FQos P-gen arm64 thru 8.95 decQuad 28.82 3.22× Rc2  
c sub SQss P-gen arm64 thru 1.51 mpdecimal 13.58 8.99× Rc2  
c sub SQos P-gen arm64 thru 0.93 mpdecimal 11.71 12.59× Rc2  
c sub NQss P-gen arm64 thru 4.06 mpdecimal 24.31 5.99× Rc2  
c sub NQos P-gen arm64 thru 3.46 mpdecimal 21.78 6.29× Rc2  
c sub MQss P-gen arm64 thru 9.73 mpdecimal 25.56 2.63× Rc2  
c sub MQos P-gen arm64 thru 5.59 mpdecimal 21.03 3.76× Rc2  
c sub OQss P-gen arm64 thru 17.20 mpdecimal 46.84 2.72× Rc2  
c sub OQos P-gen arm64 thru 11.85 mpdecimal 45.29 3.82× Rc2  
c sub FQss P-gen arm64 thru 10.70 mpdecimal 41.97 3.92× Rc2  
c sub FQos P-gen arm64 thru 8.95 mpdecimal 40.85 4.56× Rc2  

x86_64 (Intel i9-9880H).

port op cat profile arch mode ours alt alt ns ratio run notes
c sub SQss P-gen x86_64 thru 4.57 libbid 33.73 7.38× xRc2  
c sub SQos P-gen x86_64 thru 2.32 libbid 32.99 14.22× xRc2  
c sub NQss P-gen x86_64 thru 10.94 libbid 34.85 3.19× xRc2  
c sub NQos P-gen x86_64 thru 10.72 libbid 35.61 3.32× xRc2  
c sub MQss P-gen x86_64 thru 24.68 libbid 34.98 1.42× xRc2  
c sub MQos P-gen x86_64 thru 13.60 libbid 34.66 2.55× xRc2  
c sub OQss P-gen x86_64 thru 38.90 libbid 51.72 1.33× xRc2  
c sub OQos P-gen x86_64 thru 30.00 libbid 50.81 1.69× xRc2  
c sub FQss P-gen x86_64 thru 26.63 libbid 37.25 1.40× xRc2  
c sub FQos P-gen x86_64 thru 20.62 libbid 36.53 1.77× xRc2  
c sub SQss P-gen x86_64 thru 4.57 decQuad 92.66 20.28× xRc2  
c sub SQos P-gen x86_64 thru 2.32 decQuad 23.06 9.94× xRc2  
c sub NQss P-gen x86_64 thru 10.94 decQuad 95.80 8.76× xRc2  
c sub NQos P-gen x86_64 thru 10.72 decQuad 81.60 7.61× xRc2  
c sub MQss P-gen x86_64 thru 24.68 decQuad 92.74 3.76× xRc2  
c sub MQos P-gen x86_64 thru 13.60 decQuad 78.14 5.75× xRc2  
c sub OQss P-gen x86_64 thru 38.90 decQuad 103.86 2.67× xRc2  
c sub OQos P-gen x86_64 thru 30.00 decQuad 92.53 3.08× xRc2  
c sub FQss P-gen x86_64 thru 26.63 decQuad 79.07 2.97× xRc2  
c sub FQos P-gen x86_64 thru 20.62 decQuad 76.85 3.73× xRc2  
c sub SQss P-gen x86_64 thru 4.57 mpdecimal 36.53 7.99× xRc2  
c sub SQos P-gen x86_64 thru 2.32 mpdecimal 31.62 13.63× xRc2  
c sub NQss P-gen x86_64 thru 10.94 mpdecimal 55.56 5.08× xRc2  
c sub NQos P-gen x86_64 thru 10.72 mpdecimal 52.92 4.94× xRc2  
c sub MQss P-gen x86_64 thru 24.68 mpdecimal 55.14 2.23× xRc2  
c sub MQos P-gen x86_64 thru 13.60 mpdecimal 53.58 3.94× xRc2  
c sub OQss P-gen x86_64 thru 38.90 mpdecimal 139.81 3.59× xRc2  
c sub OQos P-gen x86_64 thru 30.00 mpdecimal 138.05 4.60× xRc2  
c sub FQss P-gen x86_64 thru 26.63 mpdecimal 94.19 3.54× xRc2  
c sub FQos P-gen x86_64 thru 20.62 mpdecimal 90.62 4.39× xRc2  

Multiply — CP · WP · XP

Swept 4096-input average per band (bare thru; ns/op = Time/4096 over the shared decimal128-resources/swept/ corpus, byte-identical operands every port). Relational peer table with explicit ratios; the cross-port d128 band-shape matrices are in Port-Comparison Benchmark Results.

arm64 (M3 Pro).

port op cat profile arch mode ours alt alt ns ratio run notes
c mul CP P-gen arm64 thru 1.50 libbid 23.13 15.42× Rc2 no scaling — the cheap multiply
c mul WP P-gen arm64 thru 20.65 libbid 35.06 1.70× Rc2 128-bit recipMulPow10
c mul XP P-gen arm64 thru 25.47 libbid 45.24 1.78× Rc2 256-bit recipMulPow10; 1.19× ≈ the recipmul-256 work-order’s 1.18–1.54× band
c mul CP P-gen arm64 thru 1.50 decQuad 21.33 14.22× Rc2 vs DPD
c mul WP P-gen arm64 thru 20.65 decQuad 27.37 1.33× Rc2 vs DPD
c mul XP P-gen arm64 thru 25.47 decQuad 30.17 1.18× Rc2 decQuad edges d128 on the widest product (software DPD’s flat cost; libbid still slower)
c mul CP P-gen arm64 thru 1.50 mpdecimal 21.77 14.51× Rc2 no-scale multiply vs libmpdec
c mul WP P-gen arm64 thru 20.65 mpdecimal 53.31 2.58× Rc2 128-bit recipMulPow10
c mul XP P-gen arm64 thru 25.47 mpdecimal 73.49 2.89× Rc2 d128 wins the widest product vs libmpdec (unlike decQuad)

x86_64 (Intel i9-9880H).

port op cat profile arch mode ours alt alt ns ratio run notes
c mul CP P-gen x86_64 thru 3.86 libbid 50.55 13.10× xRc2  
c mul WP P-gen x86_64 thru 37.34 libbid 73.62 1.97× xRc2  
c mul XP P-gen x86_64 thru 47.91 libbid 104.96 2.19× xRc2  
c mul CP P-gen x86_64 thru 3.86 decQuad 61.15 15.84× xRc2  
c mul WP P-gen x86_64 thru 37.34 decQuad 80.41 2.15× xRc2  
c mul XP P-gen x86_64 thru 47.91 decQuad 96.83 2.02× xRc2  
c mul CP P-gen x86_64 thru 3.86 mpdecimal 68.97 17.87× xRc2  
c mul WP P-gen x86_64 thru 37.34 mpdecimal 201.04 5.38× xRc2  
c mul XP P-gen x86_64 thru 47.91 mpdecimal 244.70 5.11× xRc2  

Divide — CD · WD · XD (+ ET · PT)

Swept 4096-input average per band (bare thru; ns/op = Time/4096 over the shared decimal128-resources/swept/ corpus, byte-identical operands every port). Relational peer table with explicit ratios; the cross-port d128 band-shape matrices are in Port-Comparison Benchmark Results.

arm64 (M3 Pro).

port op cat profile arch mode ours alt alt ns ratio run notes
c div CD P-gen arm64 thru 28.50 libbid 36.77 1.29× Rc2 128÷64 quotient-first (§2.4.10); libbid wins — the compact-divide weakness persists
c div WD P-gen arm64 thru 37.47 libbid 37.54 1.00× Rc2 256÷64; ≈ parity / slight loss
c div XD P-gen arm64 thru 32.73 libbid 38.97 1.19× Rc2 256÷128 Möller–Granlund
c div ET P-gen arm64 thru 8.24 libbid 11.68 1.42× Rc2 quotient-first exact early-out — beats libbid’s exact fast path
c div PT P-gen arm64 thru 3.12 libbid 11.45 3.67× Rc2 divPow10Divisor (§2.4.9); d128’s fastest divide (coeff-1 form)
c div CD P-gen arm64 thru 28.50 decQuad 71.29 2.50× Rc2 vs DPD
c div WD P-gen arm64 thru 37.47 decQuad 116.34 3.10× Rc2 vs DPD
c div XD P-gen arm64 thru 32.73 decQuad 173.67 5.31× Rc2 vs DPD — decNumber divide is slow
c div ET P-gen arm64 thru 8.24 decQuad 48.44 5.88× Rc2 vs DPD
c div PT P-gen arm64 thru 3.12 decQuad 44.50 14.26× Rc2 vs DPD
c div CD P-gen arm64 thru 28.50 mpdecimal 61.20 2.15× Rc2 narrowest divide gap (libmpdec’s compact divide is its cheapest, like d128’s weakness)
c div WD P-gen arm64 thru 37.47 mpdecimal 90.34 2.41× Rc2 256÷64
c div XD P-gen arm64 thru 32.73 mpdecimal 141.26 4.32× Rc2 Cowlishaw signature (CD 59 < WD 87 < XD 144)
c div ET P-gen arm64 thru 8.24 mpdecimal 61.67 7.48× Rc2 libmpdec has no exact early-out
c div PT P-gen arm64 thru 3.12 mpdecimal 51.52 16.51× Rc2 d128’s biggest divide win vs libmpdec

x86_64 (Intel i9-9880H).

port op cat profile arch mode ours alt alt ns ratio run notes
c div CD P-gen x86_64 thru 75.71 libbid 86.65 1.14× xRc2  
c div WD P-gen x86_64 thru 92.95 libbid 88.12 0.95× xRc2  
c div XD P-gen x86_64 thru 81.39 libbid 88.89 1.09× xRc2  
c div ET P-gen x86_64 thru 27.78 libbid 31.98 1.15× xRc2  
c div PT P-gen x86_64 thru 7.41 libbid 32.28 4.36× xRc2  
c div CD P-gen x86_64 thru 75.71 decQuad 152.28 2.01× xRc2  
c div WD P-gen x86_64 thru 92.95 decQuad 268.83 2.89× xRc2  
c div XD P-gen x86_64 thru 81.39 decQuad 432.16 5.31× xRc2  
c div ET P-gen x86_64 thru 27.78 decQuad 107.02 3.85× xRc2  
c div PT P-gen x86_64 thru 7.41 decQuad 90.81 12.26× xRc2  
c div CD P-gen x86_64 thru 75.71 mpdecimal 173.43 2.29× xRc2  
c div WD P-gen x86_64 thru 92.95 mpdecimal 301.84 3.25× xRc2  
c div XD P-gen x86_64 thru 81.39 mpdecimal 384.11 4.72× xRc2  
c div ET P-gen x86_64 thru 27.78 mpdecimal 171.97 6.19× xRc2  
c div PT P-gen x86_64 thru 7.41 mpdecimal 113.34 15.30× xRc2  

FMA — FN (Barrett) · FF (fits-128)

Swept 4096-input average per band (bare thru; ns/op = Time/4096 over the shared decimal128-resources/swept/ corpus, byte-identical operands every port). Relational peer table with explicit ratios; the cross-port d128 band-shape matrices are in Port-Comparison Benchmark Results.

arm64 (M3 Pro).

port op cat profile arch mode ours alt alt ns ratio run notes
c fma FN FMA arm64 thru 76.24 libbid 82.83 1.09× Rc2  
c fma FF FMA arm64 thru 39.70 libbid 58.46 1.47× Rc2  
c fma FN FMA arm64 thru 76.24 decQuad 61.70 0.81× Rc2  
c fma FF FMA arm64 thru 39.70 decQuad 72.91 1.84× Rc2  
c fma FN FMA arm64 thru 76.24 mpdecimal 89.49 1.17× Rc2  
c fma FF FMA arm64 thru 39.70 mpdecimal 154.27 3.89× Rc2  

x86_64 (Intel i9-9880H).

port op cat profile arch mode ours alt alt ns ratio run notes
c fma FN FMA x86_64 thru 130.67 libbid 176.51 1.35× xRc2  
c fma FF FMA x86_64 thru 83.43 libbid 135.93 1.63× xRc2  
c fma FN FMA x86_64 thru 130.67 decQuad 159.73 1.22× xRc2  
c fma FF FMA x86_64 thru 83.43 decQuad 168.32 2.02× xRc2  
c fma FN FMA x86_64 thru 130.67 mpdecimal 285.81 2.19× xRc2  
c fma FF FMA x86_64 thru 83.43 mpdecimal 372.25 4.46× xRc2