Avoid goroutine fan-out in SumBigValues

[perfloop-agent]

Description

Avoids creating a goroutine for every bit slice in roaring64.BSI.SumBigValues.

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds new functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Performance improvement
• Code refactoring
• Documentation update
• Test improvements
• Build/CI changes

Changes Made

What was changed?

• Traverse bit slices synchronously and reuse one big.Int term.
• Added coverage for filtered, wide signed sums and a focused benchmark.

Why was it changed?

• Per-bit cardinality intersections are small, so goroutine, channel, and closure setup costs more than the work.

How was it changed?

• Shift each cardinality into the reusable term, add ordinary bit planes, and subtract the sign plane.

Testing

go test ./...

Checks: 3 passed.

Formatting

gofmt -s -l reports no files.

Fuzzing

Not run; this only changes aggregate arithmetic and scheduling.

Performance Impact

The benchmark uses a 63-bit BSI with 128 values. The traversal is intentionally serial, favoring small per-bit intersections over per-bit scheduling.

Workload: SumBigValues on a 63-bit BSI with 128 values

Metric	Before	After	Change
ns/op	34102	1872	94.5% lower
B/op	5297	144	97.3% lower
allocs/op	132	5	96.2% lower

Running Benchmarks

go test ./roaring64 -run '^$' -bench '^BenchmarkBSI64SumBigValues$' -benchmem

Performance Analysis

Reusing the term avoids building a result channel and a separate big.Int for each bit plane.

Breaking Changes

None.

Related Issues

None.

Additional Notes

None.

---

Generated by Perfloop. Measurements and checks.