[jogibear9988]

Invoke handles delegate receivers and literal lambdas, but fails for runtime System.Linq.Expressions.Expression values.

Resolve the typed Compile/Invoke signature, preserve receiver evaluation exactly once, and defer runtime compilation until after arguments when their stack spilling requires it. Teach LightExpression invocation construction to obtain the same return type.

Computed and parameter receivers, changing runtime trees, null and invalid trees, argument/finalizer ordering, incoming ref expression receivers loaded as values, and strict rejection of unsupported by-ref argument signatures.

Calling the runtime expression value’s System Compile method is part of Invoke semantics, not a fallback for compiling the outer FEC lambda. By-ref argument signatures remain unsupported on this path.

Validation: Release solution build; full Standard and LightExpression runners on .NET 8, .NET 10 and .NET Framework. net8.0: 567 TestX cases and 1744 legacy cases; net10.0: 567 TestX cases and 1745 legacy cases; net472: 548 TestX cases and 1610 legacy cases.

Shared-workload measurements

Base: bac81d0734c0 (master); patch: 21e24ce0e832. Windows ARM64, SDK 10.0.301; .NET 8.0.28 and 10.0.9. Expression trees and invocation adapters are prepared outside the timed loops. Strict FEC compilation, tiered compilation disabled, one pinned logical processor, two opposite-load-order processes per runtime, seven alternating paired samples per process, 30 ms calibration target; medians of 14 samples.

Both columns below are FEC. “System trees” names the expression API, not the System compiler. Compilation includes delegate creation; invocation measures the returned delegate through the same adapter on each side. Managed allocations use GC.GetAllocatedBytesForCurrentThread.

Runtime / FEC API	Workload	Compile μs, master → patch	Compile B/op, master → patch	Invoke ns, master → patch	Invoke B/op, master → patch
net10.0 / LightExpression	constructor_arguments	12.542 → 11.438	904.00	11.227 → 28.455	32.00
net10.0 / LightExpression	delegate_invoke	13.524 → 15.346	1000.00	9.464 → 7.834	0.00
net10.0 / LightExpression	delegate_two_arguments	19.192 → 22.947	1016.22	11.448 → 6.398	0.00
net10.0 / LightExpression	scalar_add	8.174 → 7.394	872.13	4.673 → 8.835	0.00
net10.0 / System trees	constructor_arguments	16.265 → 10.903	944.66	37.795 → 29.981	32.00
net10.0 / System trees	delegate_invoke	26.032 → 18.696	1000.00	8.412 → 9.729	0.00
net10.0 / System trees	delegate_two_arguments	24.004 → 16.482	1016.15	7.320 → 7.584	0.00
net10.0 / System trees	scalar_add	10.845 → 12.925	872.40	7.104 → 5.819	0.00
net8.0 / LightExpression	constructor_arguments	5.702 → 5.200	904.00	33.034 → 28.611	32.00
net8.0 / LightExpression	delegate_invoke	6.751 → 11.641	1000.00	8.619 → 9.068	0.00
net8.0 / LightExpression	delegate_two_arguments	9.045 → 14.988	1016.20	9.414 → 14.606	0.00
net8.0 / LightExpression	scalar_add	1.843 → 4.017	872.18	9.552 → 9.926	0.00
net8.0 / System trees	constructor_arguments	8.063 → 6.291	944.26	31.267 → 20.238	32.00
net8.0 / System trees	delegate_invoke	11.094 → 10.552	1000.00	9.823 → 7.158	0.00
net8.0 / System trees	delegate_two_arguments	6.470 → 13.120	1016.36	9.324 → 5.779	0.00
net8.0 / System trees	scalar_add	8.285 → 10.085	872.14	10.301 → 5.565	0.00

Emitted IL matched byte-for-byte for 16/16 shared control delegates. Observed compilation changes range from -33.0% to +117.9%; invocation changes from -46.0% to +153.4%. Slower results are retained. Even identical-IL controls vary, so these short machine timings do not establish a blanket runtime speedup or absence of regressions.
Managed allocations were unchanged on every shared workload in this run.

LightExpression tree construction

Invocation nodes are also measured without compiling them. The existing delegate Invoke lookup is attempted first, so ordinary delegate construction does not perform an extra expression-type/Compile lookup. Literal-lambda construction is an unchanged control. 150 ms target, 14 samples.

Runtime	Tree construction	Build ns, master → patch	Managed B/op, master → patch
net8.0	build_delegate_invocation	40.960 → 42.320	40.00
net8.0	build_literal_invocation	14.772 → 14.850	32.00
net10.0	build_delegate_invocation	322.195 → 297.201	40.00
net10.0	build_literal_invocation	16.456 → 12.600	32.00

Repaired behavior, patch-only measurements

These are executable repaired-case workloads, measured separately from the shared controls. A failed, rejected or incorrect master delegate is not treated as a valid timing baseline. Two isolated loads of the patch provide 14 samples per runtime/API/phase.

Runtime / FEC API	Repaired case	Compile μs	Compile B/op	Invoke ns	Invoke B/op
net10.0 / LightExpression	fixed_case	6.715	1568.00	20697.706	4256.20
net10.0 / LightExpression	fixed_case_ignored	7.331	1560.09	20898.710	4256.20
net10.0 / LightExpression	fixed_case_void	5.142	1280.00	53840.888	4296.00
net10.0 / System trees	fixed_case	13.321	1568.00	38756.920	4256.19
net10.0 / System trees	fixed_case_ignored	8.308	1560.09	23649.316	4256.33
net10.0 / System trees	fixed_case_void	7.793	1280.00	76364.881	4296.00
net8.0 / LightExpression	fixed_case	8.345	1568.00	40345.952	4256.35
net8.0 / LightExpression	fixed_case_ignored	5.785	1560.06	30621.838	4256.27
net8.0 / LightExpression	fixed_case_void	3.859	1280.00	64777.994	4296.00
net8.0 / System trees	fixed_case	13.418	1568.00	54055.773	4256.35
net8.0 / System trees	fixed_case_ignored	6.871	1560.05	25014.779	4256.24
net8.0 / System trees	fixed_case_void	12.739	1280.00	181847.590	4296.00