[jogibear9988]

Integer switch-table normalization adds negative origins instead of subtracting them, and can feed 64-bit operands to the 32-bit IL switch instruction. Wide values may alias their low bits; negative or unsigned table origins can select the wrong/default branch.

Keep Int64/UInt64 operands (including enums) on the existing ordered-comparison path. Emit Int32 origins/outlier comparisons and subtract the actual table origin. Preserve upstream duplicate-test handling. The production diff is eight added/five removed lines.

Directly based on upstream master 750e1e8ea602a7e72a90072ee80b0c1376f5b72d; one commit (9d727f3b), independent of #558 and the other standalone PRs.

Correctness and validation

Four shared methods cover first-case duplicate selection, signed/unsigned 64-bit operands and enums, low-bit aliases, 32-bit origins at negative and unsigned boundaries, defaults, and single evaluation of the switch operand.

The unfixed .NET 10 regression fails the integer switch matrix in both APIs; the same matrix passes after the patch.

• Focused Standard and LightExpression regressions pass on .NET 8, .NET 10 and .NET Framework 4.7.2 (4 registered methods per API).
• Full Release solution build passes with zero warnings/errors.
• Legacy native suites pass: 1,756 cases on .NET 10 and 1,755 on .NET 8.
• Framework runner passes 198 TestX + 1,639 legacy cases.
• The full .NET 8/10 runner still reports the existing nullable MultiplyChecked parity failure for (1,1) and (2,1) in both APIs. Unmodified master was rebuilt and rerun alongside this branch; the complete failure logs are byte-for-byte identical. No regression assertions were skipped or weakened. Legacy suites were invoked separately because the existing TestX failure gates the normal runner.

Performance checks

Compared only with the master revision above, using already-working expressions. The broken expressions establish correctness and are not treated as a valid before/after performance baseline. Both Standard and LightExpression are measured. Framework is correctness-tested; the timings below cover .NET 8.0.28 and .NET 10.0.9 on Windows ARM64, SDK 10.0.301, Release.

Prebuilt expression trees; Light conversion outside the timed section; strict FEC compilation, no System fallback. Tiering disabled, fixed CPU affinity, warmed compilation and invocation. Two fresh processes with opposite load order, nine alternating samples per side, GC before samples and a 256 MiB no-GC region where available. Compilation below uses the longer approximately 100 ms pair-calibration run; invocation uses the approximately 24 ms run. Times are geometric means of the two process medians.

All measured, already-working controls retain identical decoded emitted IL and the same managed allocation per compilation/invocation. The ref-box invocation control intentionally allocates its boxed int (24 B); other invocation controls allocate zero. No persistent compiler metadata, extra collection pass, or closure allocation is added by this fix.

Timing remains noisy, including the slower rows below and variation when comparing master with itself. These measurements do not prove zero CPU regression across inputs/machines; no speedup or universal performance-equivalence claim is made.

Runtime	API	Control	Compile ns, master → fix	Change	Invoke ns, master → fix	Compile B/op, both	Invoke B/op, both
net10.0	Standard	add_control	1162.6 → 1166.0	+0.3%	1.94 → 1.87	872	0
net10.0	Standard	reference_constant	933.7 → 978.7	+4.8%	1.85 → 1.85	984	0
net10.0	Standard	switch_int32	1808.3 → 1799.5	-0.5%	2.18 → 2.13	1176	0
net10.0	Standard	switch_int64_two_cases	1614.1 → 1588.1	-1.6%	2.10 → 2.12	1160	0
net10.0	Light	add_control	992.1 → 1103.6	+11.2%	1.94 → 1.88	872	0
net10.0	Light	reference_constant	983.2 → 932.6	-5.2%	1.94 → 1.91	984	0
net10.0	Light	switch_int32	1572.4 → 1559.3	-0.8%	2.27 → 2.23	1176	0
net10.0	Light	switch_int64_two_cases	1472.2 → 1498.3	+1.8%	1.96 → 2.00	1160	0
net8.0	Standard	add_control	1099.5 → 1065.5	-3.1%	1.86 → 1.88	872	0
net8.0	Standard	reference_constant	1007.5 → 949.0	-5.8%	1.88 → 1.92	984	0
net8.0	Standard	switch_int32	1474.6 → 1575.9	+6.9%	2.16 → 2.14	1176	0
net8.0	Standard	switch_int64_two_cases	1540.6 → 1554.5	+0.9%	2.08 → 2.08	1160	0
net8.0	Light	add_control	1074.2 → 1303.7	+21.4%	1.88 → 1.87	872	0
net8.0	Light	reference_constant	903.7 → 854.5	-5.4%	1.89 → 1.90	984	0
net8.0	Light	switch_int32	1471.1 → 1524.0	+3.6%	2.15 → 2.14	1176	0
net8.0	Light	switch_int64_two_cases	1474.9 → 1424.8	-3.4%	2.02 → 2.06	1160	0

The initial short run included all 16 Standard/Light controls per runtime: add, reference constant, Int32→Int64 conversion, root ref boxing, built-in Not, method-based parameter Not, a positive-origin Int32 table switch, and a two-case Int64 comparison switch. IL and allocation checks passed across all of them. The longer run focuses on the affected operation family plus add/constant controls; all short-run timing ranges, including unselected controls, are summarized below.

Runtime	Initial compile range	Initial compile geometric mean change	Longer master/master control range	Longer master/master geometric mean change
net10.0	-12.4% to +12.7%	+0.2%	-5.3% to +5.8%	+1.4%
net8.0	-9.4% to +9.2%	-0.3%	-5.4% to +9.7%	+1.3%