[hmgaudecker]

A finite initial could be ignored when an action axis contained one element, returning the action value even when it was below the floor. The paired maximum now reduces with the dtype identity and compares the result with the requested floor afterward.

Floating values stay encoded as integers during the paired reduction and floor comparison. This preserves subnormal ordering, the winning action ID, signed-zero ties, masking, and the existing maximum derivative convention.

Validation: exact-bit value and action-ID regressions, analytic gradient checks, full argmax and hard-maximum tests, zero-mass continuation checks, and combined continuation/probability tests passed at both precisions on Marvin. Normal lint/type/lock/source-certificate checks and certificate mutation controls passed.

The broader oracle module retains separate baseline failures when explicit precision contexts are mixed; this change does not claim to repair that behavior.

Based on #629. Closes #570.

[read-the-docs-community]

Documentation build overview

📚 pylcm | 🛠️ Build #35652921 | 📁 Comparing f9c0c16 against latest (f57269f)

  🔍 Preview build  

38 files changed · + 1 added · ± 37 modified

+ Added

• shocks/index.html

± Modified

• age-specialized/index.html
• architecture/index.html
• architecture-transition-ledger/index.html
• benchmarking/index.html
• beta-delta/index.html
• categories-with-characteristics/index.html
• certification/index.html
• choosing-a-solver/index.html
• collective-regimes/index.html
• collective-regimes-1/index.html
• and 27 more...

[codecov]

Codecov Report

❌ Patch coverage is 98.15498% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 94.11%. Comparing base (fc033b7) to head (f9c0c16).

Files with missing lines	Patch %	Lines
src/_lcm/regime_building/argmax.py	91.30%	4 Missing ⚠️
tests/test_distributed.py	96.96%	1 Missing ⚠️

Additional details and impacted files

@@                      Coverage Diff                       @@
##           codex/583-subnormal-liveness     #630    +/-   ##
==============================================================
  Coverage                         94.10%   94.11%            
==============================================================
  Files                               660      660            
  Lines                            110606   110856   +250     
==============================================================
+ Hits                             104089   104334   +245     
- Misses                             6517     6522     +5     

Flag	Coverage Δ
cpu-python	94.11% <98.15%> (+<0.01%)	⬆️

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.

[github-actions]

Benchmark comparison (main → HEAD)

Comparing c8430c80 (main) → f9c0c164 (HEAD)

Benchmark	Statistic	before	after	Ratio	Alert
ACA (reduced)	warm solve + simulate (reuses compiled code)	43.816 s	259.917 s	5.93	❌
	cold solve + simulate (first run, includes compilation)	791.91 s	1900.01 s	2.40	❌
	peak CPU mem	7.54 GB	11.75 GB	1.56	❌
Mahler-Yum	warm solve + simulate (reuses compiled code)	5.853 s	7.249 s	1.24	❌
	cold solve + simulate (first run, includes compilation)	18.37 s	28.04 s	1.53	❌
	peak CPU mem	1.69 GB	1.78 GB	1.06
	peak GPU mem: automatic solve + simulate	1.61 GB	496 MB	0.31
	peak GPU mem: load saved solution + simulate	1.89 GB	837 MB	0.44
	peak GPU mem: solve + save all persistable artifacts	571 MB	571 MB	1.00
Precautionary Savings - Solve	execution time	259.9 ms	83.5 ms	0.32
	first call (including compilation)	1.55 s	1.70 s	1.10
	peak CPU mem	1.44 GB	1.45 GB	1.01
Precautionary Savings - Simulate	execution time	900.0 ms	899.5 ms	1.00
	peak GPU mem	324 MB	640 MB	1.98	❌
	first call (including compilation)	2.91 s	3.97 s	1.36	❌
	peak CPU mem	1.59 GB	1.58 GB	0.99
Precautionary Savings - Solve & Simulate	execution time	866.3 ms	741.6 ms	0.86
	peak GPU mem	1.50 GB	3.74 GB	2.49	❌
	first call (including compilation)	4.07 s	5.27 s	1.29	❌
	peak CPU mem	1.54 GB	1.54 GB	1.00
Precautionary Savings - Solve & Simulate (irreg)	execution time	997.5 ms	886.9 ms	0.89
	peak GPU mem	2.20 GB	3.74 GB	1.70	❌
	first call (including compilation)	4.20 s	5.59 s	1.33	❌
	peak CPU mem	1.54 GB	1.55 GB	1.00
Iskhakov et al. (2017) - GridSearch Solve	execution time	363.4 ms	187.4 ms	0.52
	peak GPU mem	120 MB	120 MB	1.00
	first call (including compilation)	2.81 s	2.66 s	0.95
	peak CPU mem	1.57 GB	1.63 GB	1.04
Iskhakov et al. (2017) - DC-EGM Solve	execution time	651.1 ms	814.6 ms	1.25	❌
	first call (including compilation)	6.13 s	6.83 s	1.11	❌
	peak CPU mem	1.85 GB	1.98 GB	1.07
Iskhakov et al. (2017) - GridSearch Simulate	execution time	1.011 s	918.8 ms	0.91
	peak GPU mem	892 MB	890 MB	1.00
	first call (including compilation)	6.54 s	7.13 s	1.09
	peak CPU mem	1.70 GB	1.77 GB	1.04
Iskhakov et al. (2017) - DC-EGM Simulate	execution time	1.115 s	1.270 s	1.14	❌
	peak GPU mem	891 MB	891 MB	1.00
	first call (including compilation)	6.84 s	9.57 s	1.40	❌
	peak CPU mem	1.92 GB	2.04 GB	1.07
Collective Household - Construct	execution time	735.8 ms	1.497 s	2.03	❌
	peak CPU mem	1.40 GB	1.43 GB	1.02
Collective Household - Solve	execution time	481.2 ms	5.347 s	11.11	❌
	first call (including compilation)	2.88 s	44.87 s	15.57	❌
	peak CPU mem	1.46 GB	2.77 GB	1.90	❌
Collective Household - Simulate (1000)	execution time	1.189 s	37.917 s	31.89	❌
	first call (including compilation)	14.59 s	591.79 s	40.56	❌
	peak CPU mem	1.48 GB	31.50 GB	21.28	❌
Collective Household - Simulate (100000)	execution time	1.220 s	nan µs	nan
	peak GPU mem	189 MB	nan MB	nan
	first call (including compilation)	15.33 s	nan s	nan
	peak CPU mem	1.56 GB	nan MB	nan
Reference Chain - Solve (1)	execution time	217.7 ms	39.9 ms	0.18
	first call (including compilation)	1.76 s	1.75 s	0.99
	peak CPU mem	1.12 GB	1.19 GB	1.06
Reference Chain - Solve (8)	execution time	492.7 ms	259.5 ms	0.53
	first call (including compilation)	5.80 s	6.51 s	1.12	❌
	peak CPU mem	1.23 GB	1.31 GB	1.07
SimulationDispatch ('dissolution', 'off')	warm simulation time per period and regime (ms)	45.77 ms	74.67 ms	1.63	❌
	backend compile requests on second simulation call		0
SimulationDispatch ('dissolution', 'progress')	warm simulation time per period and regime (ms)	44.76 ms	78.89 ms	1.76	❌
	backend compile requests on second simulation call		0
SimulationDispatch ('multi_regime', 'off')	warm simulation time per period and regime (ms)	49.92 ms	51.83 ms	1.04
	backend compile requests on second simulation call		0
SimulationDispatch ('multi_regime', 'progress')	warm simulation time per period and regime (ms)	48.43 ms	51.53 ms	1.06
	backend compile requests on second simulation call		0