[shravanngoswamii]

The hint reserved room for a million entries on every evaluation of a model with more than 20 discrete latents.
That is about 50 MiB a call, and several hundred MiB a ForwardDiff gradient since the cached values are duals.
A ForwardDiff gradient of Eye Tracking drops from 94 to 39 ms, and of Cervix from 148 to 91 ms.

[github-actions]

Benchmark Results

Julia 1.13.4 on Intel(R) Xeon(R) 6975P-C (Linux)

Ratio = JuliaBUGS/Stan (lower is better for JuliaBUGS)

# Params shows Stan / JuliaBUGS when parameter counts differ.

Log Density uses generated evaluation. Gradient columns measure log density + gradient.

McRvs = Mooncake reverse mode with generated evaluation; FwdDiff = ForwardDiff with graph evaluation.

Model	# Params	Log Density Ratio	McRvs Ratio	FwdDiff Ratio
rats	65	0.47x	2.06x	86.95x
pumps	12	1.88x	2.11x	17.17x
bones	13	0.73x	3.55x	1.16x
oxford	244	1.58x	4.41x	575.23x
epil	303	0.47x	1.61x	988.53x
lsat	1006	0.65x	1.96x	1446.85x
schools	133	0.29x	1.84x	66.47x
beetles	2	2.47x	4.98x	11.70x
air	5	2.18x	1.94x	13.62x

Note: Stan benchmarks use hand-optimized Stan models, not direct BUGS translations. Comparison is illustrative only.

[github-actions]

JuliaBUGS.jl documentation for PR #571 is available at:
https://TuringLang.github.io/JuliaBUGS.jl/previews/PR571/

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 88.11%. Comparing base (5fd788c) to head (09de0a5).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #571      +/-   ##
==========================================
- Coverage   88.49%   88.11%   -0.38%     
==========================================
  Files          30       30              
  Lines        3851     3846       -5     
==========================================
- Hits         3408     3389      -19     
- Misses        443      457      +14     

Flag	Coverage Δ
callbacks	28.43% <ø> (-0.07%)	⬇️
compilation_model	50.06% <ø> (+0.06%)	⬆️
compilation_source_gen	44.91% <ø> (+0.05%)	⬆️
distributed_sampling	38.60% <ø> (+0.05%)	⬆️
elementary	27.93% <ø> (+0.03%)	⬆️
frontend	47.08% <ø> (+0.06%)	⬆️
gibbs	39.68% <ø> (+0.05%)	⬆️
graphs	21.17% <ø> (+0.02%)	⬆️
inference_chains	37.01% <ø> (-0.06%)	⬇️
inference_flexichains	?
inference_hmc	29.94% <ø> (+0.03%)	⬆️
inference_marginalization	31.01% <ø> (-0.10%)	⬇️
inference_mh	35.50% <ø> (+0.04%)	⬆️
inference_slice_sampling	26.31% <ø> (+0.03%)	⬆️
log_density	59.09% <ø> (-0.06%)	⬇️
model_operations	42.96% <ø> (-0.05%)	⬇️
parallel_sampling	?
params_with_stats	38.13% <ø> (-0.06%)	⬇️
serialization	34.32% <ø> (+0.04%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.