Sightglass Benchmark Results

Results use 99.00% confidence intervals (α = 0.01). CV is Coefficient of Variation (measurement consistency).
Benchmark CV% 📊 ./main-65abe37.so …branch-a84af50.so
cm-online-stats/cm-online-stats 4.46% 901.6M cycles (p25) 0.95% slower
dart-deltablue/dart-deltablue 4.77% 78.0M cycles (p25) 1.80% faster
dotnet-raytracer/dotnet-raytracer 3.40% 45.4M cycles (p25) 0.87% slower
gcc-loops 3.61% 6.4M cycles (p25) 3.01% faster
hashset 4.80% 10.2M cycles (p25) No significant difference
hex-simd 1.65% 7.5M cycles (p25) 0.43% slower
hoot-earley-boyer/hoot-earley-boyer 5.77% 212.6M cycles (p25) No significant difference
intgemm-simd 3.73% 7.4M cycles (p25) No significant difference
java-navier-stokes/java-navier-stokes 3.43% 16.8M cycles (p25) 1.78% slower
kotlin-richards/kotlin-richards 4.26% 38.6M cycles (p25) 1.19% slower
libsodium/libsodium-core5 4.34% 26.0M cycles (p25) No significant difference
libsodium/libsodium-hash3 5.48% 5.9M cycles (p25) 1.55% faster
libsodium/libsodium-keygen 3.73% 22.7M cycles (p25) No significant difference
libsodium/libsodium-siphashx24 5.10% 5.9M cycles (p25) No significant difference
ocaml-raytrace/ocaml-raytrace 4.91% 54.4M cycles (p25) No significant difference
richards 3.68% 20.4M cycles (p25) 1.03% faster
rust-json 12.61% 12.4M cycles (p25) No significant difference
rust-protobuf 3.72% 25.1M cycles (p25) No significant difference
rust-wasmprinter/rust-wasmprinter 3.39% 22.2M cycles (p25) 1.27% slower
shootout/shootout-heapsort 2.69% 25.1M cycles (p25) 1.18% slower
shootout/shootout-keccak 17.05% 19.5M cycles (p25) 5.38% faster
shootout/shootout-memmove 3.39% 343.5M cycles (p25) 0.69% slower
shootout/shootout-sieve 6.28% 7.6M cycles (p25) No significant difference
shootout/shootout-switch 9.99% 1.1M cycles (p25) 3.48% faster
shootout/shootout-xchacha20 4.94% 5.9M cycles (p25) 1.39% slower
splay/splay 1.92% 4.7B cycles (p25) No significant difference
starlingmonkey/starlingmonkey-markdown 3.03% 7.1M cycles (p25) 1.49% slower
tinygo/tinygo-regex 3.99% 13.8M cycles (p25) 0.76% faster
wat-bytecode-interp/wat-bytecode-interp 3.93% 6.8M cycles (p25) No significant difference

cm-online-stats/cm-online-stats

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 901.6M cycles (p25) — — 4.46%
…branch-a84af50.so 916.3M cycles 0.95% slower ±7.2M cycles 3.98%

dart-deltablue/dart-deltablue

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 78.0M cycles (p25) — — 4.77%
…branch-a84af50.so 77.0M cycles 1.80% faster ±691.4K cycles 4.62%

dotnet-raytracer/dotnet-raytracer

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 45.4M cycles (p25) — — 3.40%
…branch-a84af50.so 46.3M cycles 0.87% slower ±276.0K cycles 2.99%

gcc-loops

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 6.4M cycles (p25) — — 3.61%
…branch-a84af50.so 6.1M cycles 3.01% faster ±43.0K cycles 3.62%

hashset

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 10.2M cycles (p25) — — 4.80%
…branch-a84af50.so 10.3M cycles No significant difference ±84.9K cycles 3.91%

hex-simd

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 7.5M cycles (p25) — — 1.65%
…branch-a84af50.so 7.6M cycles 0.43% slower ±20.9K cycles 1.34%

hoot-earley-boyer/hoot-earley-boyer

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 212.6M cycles (p25) — — 5.77%
…branch-a84af50.so 212.8M cycles No significant difference ±2.2M cycles 5.06%

intgemm-simd

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 7.4M cycles (p25) — — 3.73%
…branch-a84af50.so 7.4M cycles No significant difference ±50.1K cycles 3.39%

java-navier-stokes/java-navier-stokes

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 16.8M cycles (p25) — — 3.43%
…branch-a84af50.so 16.9M cycles 1.78% slower ±128.6K cycles 4.46%

kotlin-richards/kotlin-richards

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 38.6M cycles (p25) — — 4.26%
…branch-a84af50.so 39.3M cycles 1.19% slower ±339.6K cycles 4.92%

libsodium/libsodium-core5

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 26.0M cycles (p25) — — 4.34%
…branch-a84af50.so 26.3M cycles No significant difference ±196.5K cycles 3.46%

libsodium/libsodium-hash3

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 5.9M cycles (p25) — — 5.48%
…branch-a84af50.so 5.7M cycles 1.55% faster ±60.1K cycles 5.31%

libsodium/libsodium-keygen

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 22.7M cycles (p25) — — 3.73%
…branch-a84af50.so 22.7M cycles No significant difference ±144.8K cycles 2.98%

libsodium/libsodium-siphashx24

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 5.9M cycles (p25) — — 5.10%
…branch-a84af50.so 5.8M cycles No significant difference ±58.2K cycles 5.40%

ocaml-raytrace/ocaml-raytrace

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 54.4M cycles (p25) — — 4.91%
…branch-a84af50.so 54.1M cycles No significant difference ±507.9K cycles 5.03%

richards

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 20.4M cycles (p25) — — 3.68%
…branch-a84af50.so 19.8M cycles 1.03% faster ±147.3K cycles 4.09%

rust-json

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 12.4M cycles (p25) — — 12.61%
…branch-a84af50.so 12.8M cycles No significant difference ±280.6K cycles 10.41%

rust-protobuf

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 25.1M cycles (p25) — — 3.72%
…branch-a84af50.so 25.1M cycles No significant difference ±173.1K cycles 3.62%

rust-wasmprinter/rust-wasmprinter

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 22.2M cycles (p25) — — 3.39%
…branch-a84af50.so 22.4M cycles 1.27% slower ±137.1K cycles 3.16%

shootout/shootout-heapsort

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 25.1M cycles (p25) — — 2.69%
…branch-a84af50.so 25.1M cycles 1.18% slower ±159.2K cycles 3.88%

shootout/shootout-keccak

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 19.5M cycles (p25) — — 17.05%
…branch-a84af50.so 18.7M cycles 5.38% faster ±543.1K cycles 10.91%

shootout/shootout-memmove

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 343.5M cycles (p25) — — 3.39%
…branch-a84af50.so 347.2M cycles 0.69% slower ±2.3M cycles 3.68%

shootout/shootout-sieve

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 7.6M cycles (p25) — — 6.28%
…branch-a84af50.so 7.6M cycles No significant difference ±94.7K cycles 6.72%

shootout/shootout-switch

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 1.1M cycles (p25) — — 9.99%
…branch-a84af50.so 1.0M cycles 3.48% faster ±22.0K cycles 11.90%

shootout/shootout-xchacha20

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 5.9M cycles (p25) — — 4.94%
…branch-a84af50.so 5.9M cycles 1.39% slower ±59.0K cycles 5.51%

splay/splay

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 4.7B cycles (p25) — — 1.92%
…branch-a84af50.so 4.7B cycles No significant difference ±15.1M cycles 1.52%

starlingmonkey/starlingmonkey-markdown

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 7.1M cycles (p25) — — 3.03%
…branch-a84af50.so 7.3M cycles 1.49% slower ±39.8K cycles 2.84%

tinygo/tinygo-regex

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 13.8M cycles (p25) — — 3.99%
…branch-a84af50.so 13.8M cycles 0.76% faster ±104.5K cycles 4.05%

wat-bytecode-interp/wat-bytecode-interp

Engine Cycles (p25) Performance Confidence Interval CV%
📊 ./main-65abe37.so 6.8M cycles (p25) — — 3.93%
…branch-a84af50.so 6.7M cycles No significant difference ±47.4K cycles 3.51%

Statistical Methods

Uses effect size analysis with confidence intervals based on Student's t-test. Statistical significance indicates the difference is likely real (not measurement noise). Tests with CV > 5% may have unreliable measurements.

Baseline Engine (📊): The baseline engine is the reference point for all comparisons. Performance differences are calculated relative to the baseline. The baseline engine appears with a chart icon (📊) and is highlighted with a light blue background in detailed tables.

P25 Metric: Cycle counts show the 25th percentile (p25) rather than mean or median. This metric is more robust to outliers, as noise in benchmarking typically appears as spikes on the higher end (system interrupts, background processes, etc.). P25 represents a "good run" that is consistently achievable—25% of runs are this fast or faster.

Color Coding: Green indicates significantly faster performance, red indicates significantly slower, and gray indicates high measurement noise (CV > 5%). No highlighting means no statistically significant difference from baseline.

References: Benchmark Suite RFC | Sightglass Documentation