How Clang compiles a function (2018)

The article discusses how Clang compiles a simple C++ function, `is_sorted`, into LLVM Intermediate Representation (IR). It emphasizes a high-level overview rather than delving into Clang's internal workings or complex C++ features. The function checks if an array is sorted and is compiled using the command `clang++ is_sorted.cpp -O0 -S -emit-llvm`, which instructs Clang to produce unoptimized LLVM IR. The output includes comments and metadata about the module, such as the target architecture and data layout.

The function's structure in LLVM IR is explained, highlighting the allocation of stack memory for parameters and local variables. The translation of the for loop into basic blocks is outlined, demonstrating how Clang handles control flow. Each basic block contains instructions that must adhere to specific entry and exit rules, with the first block initializing variables and branching to the loop condition.

The article details how the loop condition and body are represented in IR, including memory operations and comparisons. It also explains the handling of the return value and the final return instruction. The author notes that Clang generates what they term "degenerate SSA code," which meets SSA requirements but relies on memory rather than direct value manipulation. The post concludes with a promise to explore LLVM IR-level optimizations in future writings.