[jogibear9988]

Explicit conversion methods taking a ref parameter can receive a temporary instead of the original storage. Normalize the formal parameter type and distinguish direct ref passing from lifted conversions, which must use an unwrapped value copy.

Incoming references, value parameters, locals, fields, constants and method results; lifted/null cases, mutation, nullable struct results, discarded results and non-ref controls. These new custom-ref conversion regressions use the Standard expression API; the full LightExpression suite also runs. Regressions require strict FEC compilation. Native suites cover Standard and LightExpression.

Validation

Local validation uses master baseline 094d067 plus this change. Release solution build: zero warnings/errors.

Runtime	TestX groups	Legacy tests
.NET 10	493 passed	1,756 passed
.NET 8	493 passed	1,755 passed
.NET Framework 4.7.2	472 passed	1,639 passed

Measurements

Master baseline 094d0678 → this change, Release, Windows ARM64, .NET 8.0.28 / 10.0.9. Compilation uses prebuilt trees; runtime measures invocation through the same wrapper. Two fresh process rounds with reversed load order, seven samples per round, about 30 ms calibration, one logical CPU and tiered compilation disabled. Tables use the median of the two per-round medians.

The controls below produce correct results on master for their measured inputs. Newly supported trees have no valid before/after execution comparison because master rejects them, produces incorrect results or emits invalid IL. Timings include machine and JIT/code-placement variation; the unchanged add control helps show that variation. These observations do not prove universal performance neutrality.

net8.0 results (master → change)

API / workload	Compile ns	Invoke ns	Compile B/op	Invoke B/op
Standard / add_control	1783.3 → 1810.2	2.8 → 2.9	872.0 → 872.0	0.0 → 0.0
Standard / numeric_conversion	1711.3 → 1578.9	3.1 → 3.2	872.1 → 872.1	0.0 → 0.0
Standard / nullable_conversion	3071.8 → 3193.4	2.8 → 2.9	1648.1 → 1648.1	0.0 → 0.0
Standard / method_conversion	1514.6 → 1645.5	3.0 → 3.0	928.0 → 928.0	0.0 → 0.0
Standard / method_nullable_conversion	3072.0 → 3032.7	2.8 → 3.0	1704.0 → 1704.0	0.0 → 0.0
Standard / lifted_ref_conversion	3431.2 → 2959.2	2.8 → 2.7	1712.0 → 1712.0	0.0 → 0.0
Light / add_control	1315.2 → 1369.1	2.5 → 2.5	872.0 → 872.0	0.0 → 0.0
Light / numeric_conversion	1379.8 → 1427.8	3.1 → 3.1	872.0 → 872.0	0.0 → 0.0
Light / nullable_conversion	2771.3 → 3051.6	2.8 → 2.7	1648.1 → 1648.1	0.0 → 0.0

net10.0 results (master → change)

API / workload	Compile ns	Invoke ns	Compile B/op	Invoke B/op
Standard / add_control	4447.5 → 4593.4	2.6 → 2.6	872.1 → 872.1	0.0 → 0.0
Standard / numeric_conversion	4077.3 → 4068.4	2.5 → 2.5	872.1 → 872.1	0.0 → 0.0
Standard / nullable_conversion	7927.6 → 7699.5	2.4 → 2.4	1648.2 → 1648.2	0.0 → 0.0
Standard / method_conversion	4404.5 → 4366.8	2.5 → 2.5	928.0 → 928.0	0.0 → 0.0
Standard / method_nullable_conversion	7565.3 → 7762.5	2.3 → 2.3	1704.0 → 1704.0	0.0 → 0.0
Standard / lifted_ref_conversion	8139.0 → 7858.3	2.4 → 2.3	1712.0 → 1712.0	0.0 → 0.0
Light / add_control	3450.1 → 3418.9	2.3 → 2.2	872.0 → 872.0	0.0 → 0.0
Light / numeric_conversion	3032.9 → 3103.5	2.5 → 2.4	872.1 → 872.1	0.0 → 0.0
Light / nullable_conversion	7113.9 → 6738.5	2.5 → 2.4	1648.1 → 1648.1	0.0 → 0.0

Emitted IL differs in these measured workloads: none.
No measured allocation increase above 0.1 B/op in these controls.