Explain Swift's compilation process and LLVM integration.

Swift's compilation process involves multiple stages with LLVM integration for optimization.\n\n• **Compilation Stages:**\n1. **Parsing:** Source code → AST (Abstract Syntax Tree)\n2. **Semantic Analysis:** Type checking, name resolution\n3. **SIL Generation:** Swift Intermediate Language\n4. **SIL Optimization:** High-level optimizations\n5. **LLVM IR Generation:** Lower-level representation\n6. **LLVM Optimization:** Machine-independent optimizations\n7. **Code Generation:** Target-specific machine code\n\n• **Swift Intermediate Language (SIL):**\n\n// Swift source\nfunc add(_ a: Int, _ b: Int) -> Int {\n return a + b\n}\n\n// Simplified SIL representation\nsil @add : $@convention(thin) (Int, Int) -> Int {\nbb0(%0 : $Int, %1 : $Int):\n %2 = builtin "sadd_with_overflow_Int64"(%0 : $Int, %1 : $Int)\n %3 = tuple_extract %2 : $(Int, Builtin.Int1), 0\n return %3 : $Int\n}\n\n\n• **LLVM Integration:**\n• Swift compiler generates LLVM IR\n• LLVM performs optimizations\n• Target-specific code generation\n• Link-time optimization (LTO)\n\n• **Optimization Levels:**\nbash\n# Debug build (no optimization)\nswiftc -Onone source.swift\n\n# Release build (full optimization)\nswiftc -O source.swift\n\n# Size optimization\nswiftc -Osize source.swift\n\n\n• **Whole Module Optimization:**\n• Analyzes entire module together\n• Better inlining decisions\n• Dead code elimination\n• Devirtualization opportunities