Decimal128 by Miguel

Rust 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 Rust 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 Rust, d128 is measured against its in-language idiom peer rust_decimal on the compact bands it can represent, falling back to the libbid universal reference where it cannot (wide products, XD). 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/idiom peer on x86_64 (Intel i9-9880H): ratio = rust_decimal / Miguel or ratio = libbid / Miguel (> 1× ⇒ d128 faster), broken out by operation. rust_decimal has no wide-product multiply band (that falls back to libbid), and libbid isn’t used for add/subtract/divide, so those cells are blank.

  Add Subtract Multiply Divide
ratio = rust_decimal / Miguel 2.1× 1.9× — 0.6× – 5×
ratio = libbid / Miguel — — 2.0× – 13× —

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
rust add MIX P-fin arm64 thru 2.01 rust_decimal 3.36 1.67× Rrsw2 compact idiom peer
rust sub MIX P-fin arm64 thru 2.20 rust_decimal 3.55 1.61× Rrsw2 compact idiom peer
rust div CD P-fin arm64 thru 20.82 rust_decimal 13.90 0.67× Rrsw2 compact idiom peer
rust div WD P-fin arm64 thru 38.54 rust_decimal 20.09 0.52× Rrsw2 compact idiom peer
rust div ET P-fin arm64 thru 6.27 rust_decimal 3.84 0.61× Rrsw2 compact idiom peer
rust div PT P-fin arm64 thru 4.00 rust_decimal 15.34 3.83× Rrsw2 compact idiom peer
rust mul CP P-fin arm64 thru 1.13 libbid 23.57 20.86× Rrsw2  
rust mul WP P-fin arm64 thru 14.43 libbid 34.52 2.39× Rrsw2  

x86_64 (Intel i9-9880H).

port op cat profile arch mode ns alt alt ns ratio run notes
rust add MIX P-fin x86_64 thru 6.85 rust_decimal 14.88 2.17× xRrsw2 compact idiom peer
rust sub MIX P-fin x86_64 thru 8.30 rust_decimal 15.96 1.92× xRrsw2 compact idiom peer
rust div CD P-fin x86_64 thru 77.86 rust_decimal 59.10 0.76× xRrsw2 compact idiom peer
rust div WD P-fin x86_64 thru 93.58 rust_decimal 79.71 0.85× xRrsw2 compact idiom peer
rust div ET P-fin x86_64 thru 25.17 rust_decimal 15.46 0.61× xRrsw2 compact idiom peer
rust div PT P-fin x86_64 thru 10.78 rust_decimal 55.19 5.12× xRrsw2 compact idiom peer
rust mul CP P-fin x86_64 thru 3.85 libbid 50.91 13.22× xRrsw2  
rust mul WP P-fin x86_64 thru 32.72 libbid 66.63 2.04× xRrsw2  

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
rust add SQss P-gen arm64 thru 0.81 rust_decimal 3.82 4.72× Rrsw2 compact idiom peer
rust add SQos P-gen arm64 thru 2.00 rust_decimal 3.41 1.71× Rrsw2 compact idiom peer
rust add NQss P-gen arm64 thru 3.37 rust_decimal 5.67 1.68× Rrsw2 compact idiom peer
rust add NQos P-gen arm64 thru 3.76 rust_decimal 5.11 1.36× Rrsw2 compact idiom peer
rust add MQss P-gen arm64 thru 5.22 rust_decimal 6.11 1.17× Rrsw2 compact idiom peer
rust add MQos P-gen arm64 thru 11.25 rust_decimal 5.52 0.49× Rrsw2 compact idiom peer
rust add OQss P-gen arm64 thru 9.06 - - - Rrsw2  
rust add OQos P-gen arm64 thru 13.78 - - - Rrsw2  
rust add FQss P-gen arm64 thru 8.41 - - - Rrsw2  
rust add FQos P-gen arm64 thru 10.25 - - - Rrsw2  

x86_64 (Intel i9-9880H).

port op cat profile arch mode ours alt alt ns ratio run notes
rust add SQss P-gen x86_64 thru 2.53 rust_decimal 14.21 5.62× xRrsw2 compact idiom peer
rust add SQos P-gen x86_64 thru 6.59 rust_decimal 18.49 2.81× xRrsw2 compact idiom peer
rust add NQss P-gen x86_64 thru 11.20 rust_decimal 21.32 1.90× xRrsw2 compact idiom peer
rust add NQos P-gen x86_64 thru 11.45 rust_decimal 22.93 2.00× xRrsw2 compact idiom peer
rust add MQss P-gen x86_64 thru 14.54 rust_decimal 20.40 1.40× xRrsw2 compact idiom peer
rust add MQos P-gen x86_64 thru 26.48 rust_decimal 20.41 0.77× xRrsw2 compact idiom peer
rust add OQss P-gen x86_64 thru 32.30 - - - xRrsw2  
rust add OQos P-gen x86_64 thru 44.79 - - - xRrsw2  
rust add FQss P-gen x86_64 thru 20.10 - - - xRrsw2  
rust add FQos P-gen x86_64 thru 24.88 - - - xRrsw2  

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
rust sub SQss P-gen arm64 thru 1.36 rust_decimal 3.27 2.40× Rrsw2 compact idiom peer
rust sub SQos P-gen arm64 thru 1.06 rust_decimal 4.39 4.14× Rrsw2 compact idiom peer
rust sub NQss P-gen arm64 thru 3.74 rust_decimal 5.10 1.36× Rrsw2 compact idiom peer
rust sub NQos P-gen arm64 thru 3.45 rust_decimal 5.52 1.60× Rrsw2 compact idiom peer
rust sub MQss P-gen arm64 thru 11.32 rust_decimal 5.48 0.48× Rrsw2 compact idiom peer
rust sub MQos P-gen arm64 thru 5.08 rust_decimal 6.16 1.21× Rrsw2 compact idiom peer
rust sub OQss P-gen arm64 thru 13.44 - - - Rrsw2  
rust sub OQos P-gen arm64 thru 9.05 - - - Rrsw2  
rust sub FQss P-gen arm64 thru 9.95 - - - Rrsw2  
rust sub FQos P-gen arm64 thru 8.06 - - - Rrsw2  

x86_64 (Intel i9-9880H).

port op cat profile arch mode ours alt alt ns ratio run notes
rust sub SQss P-gen x86_64 thru 5.08 rust_decimal 15.58 3.07× xRrsw2 compact idiom peer
rust sub SQos P-gen x86_64 thru 4.03 rust_decimal 15.07 3.74× xRrsw2 compact idiom peer
rust sub NQss P-gen x86_64 thru 11.41 rust_decimal 19.36 1.70× xRrsw2 compact idiom peer
rust sub NQos P-gen x86_64 thru 10.86 rust_decimal 18.69 1.72× xRrsw2 compact idiom peer
rust sub MQss P-gen x86_64 thru 25.85 rust_decimal 19.03 0.74× xRrsw2 compact idiom peer
rust sub MQos P-gen x86_64 thru 14.91 rust_decimal 18.73 1.26× xRrsw2 compact idiom peer
rust sub OQss P-gen x86_64 thru 43.96 - - - xRrsw2  
rust sub OQos P-gen x86_64 thru 32.49 - - - xRrsw2  
rust sub FQss P-gen x86_64 thru 24.65 - - - xRrsw2  
rust sub FQos P-gen x86_64 thru 19.85 - - - xRrsw2  

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
rust mul CP P-gen arm64 thru 1.52 libbid 23.13 15.22× Rrsw2  
rust mul WP P-gen arm64 thru 13.81 libbid 35.06 2.54× Rrsw2  
rust mul XP P-gen arm64 thru 25.03 libbid 45.24 1.81× Rrsw2  

x86_64 (Intel i9-9880H).

port op cat profile arch mode ours alt alt ns ratio run notes
rust mul CP P-gen x86_64 thru 6.07 libbid 50.55 8.33× xRrsw2  
rust mul WP P-gen x86_64 thru 32.36 libbid 73.62 2.28× xRrsw2  
rust mul XP P-gen x86_64 thru 46.76 libbid 104.96 2.24× xRrsw2  

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
rust div CD P-gen arm64 thru 20.04 libbid 36.77 1.83× Rrsw2  
rust div WD P-gen arm64 thru 37.80 libbid 37.54 0.99× Rrsw2  
rust div XD P-gen arm64 thru 37.84 libbid 38.97 1.03× Rrsw2  
rust div ET P-gen arm64 thru 9.60 libbid 11.68 1.22× Rrsw2  
rust div PT P-gen arm64 thru 3.99 libbid 11.45 2.87× Rrsw2  

x86_64 (Intel i9-9880H).

port op cat profile arch mode ours alt alt ns ratio run notes
rust div CD P-gen x86_64 thru 82.14 libbid 86.65 1.05× xRrsw2  
rust div WD P-gen x86_64 thru 90.76 libbid 88.12 0.97× xRrsw2  
rust div XD P-gen x86_64 thru 90.46 libbid 88.89 0.98× xRrsw2  
rust div ET P-gen x86_64 thru 31.05 libbid 31.98 1.03× xRrsw2  
rust div PT P-gen x86_64 thru 10.49 libbid 32.28 3.08× xRrsw2  

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
rust fma FN FMA arm64 thru 22.84 - - - Rrsw2  
rust fma FF FMA arm64 thru 35.77 - - - Rrsw2  

x86_64 (Intel i9-9880H).

port op cat profile arch mode ours alt alt ns ratio run notes
rust fma FN FMA x86_64 thru 67.47 - - - xRrsw2  
rust fma FF FMA x86_64 thru 73.82 - - - xRrsw2