This project is a benchmarking suite for prime number algorithms. It does not:

• Handle user authentication
• Process sensitive data
• Provide network services
• Execute untrusted code

The security considerations are minimal but documented here for completeness.

Prime sieve implementations must handle large values carefully:

// BAD: Overflows at n > ~46,340
int p_squared = p * p;

// GOOD: Use appropriate types
uint64_t p_squared = (uint64_t)p * p;

All Round 2 (1e9) implementations have been audited for overflow safety.

Bit manipulation macros must validate indices:

// Ensure idx < array_size * 64
#define SET_BIT(arr, idx) (arr[(idx) >> 6] |= (1ULL << ((idx) & 63)))

All implementations bounds-check against the declared sieve size.

The benchmarker limits execution time:

• Default timeout: 60 seconds per implementation
• Memory limit: Implicit via segmented sieve design

Users running with n > 10^12 should expect extended runtimes.

If you discover a bug that could cause:

• Incorrect prime counts
• Crashes on valid input
• Undefined behavior

Please open a GitHub issue with:

1. Input parameters (n value)
2. Implementation file
3. Compiler version and flags
4. Expected vs actual output

For security-sensitive issues (unlikely in this project), contact:

• Email: security@whispr.dev
• Response time: 72 hours

All implementations are verified against known values:

Run verification:

./c-primes-claude.exe | grep "Found"
# Expected: Found 41538 primes...

For reproducible builds, pin compiler versions:

# GCC
g++ --version  # Tested: 13.2.0

# Flags used
g++ -O3 -march=native -std=c++17 -flto

Different optimization levels may produce different timings but identical prime counts.

This project prioritizes correctness over security, as it processes no external input beyond command-line arguments.