Clang

Starting in 2005, Apple Inc. made extensive use of LLVM in a number of commercial products,[12] including the iOS SDK and Xcode 3.1.

One of the first uses of LLVM was an OpenGL code compiler for OS X that converts OpenGL calls into more fundamental calls for graphics processing units (GPU) that do not support certain features. This allowed Apple to support OpenGL on computers using Intel Graphics Media Accelerator (GMA) chipsets, increasing performance on those machines.[13] For GPUs that support it, the code is compiled to fully exploit the underlying hardware, but on GMA machines, LLVM compiles the same OpenGL code into subroutines to ensure continued proper function.

LLVM was intended originally to use GCC's front end, but GCC turned out to cause some problems for developers of LLVM and at Apple. The GCC source code is a large and somewhat cumbersome system for developers to work with; as one long-time GCC developer put it referring to LLVM, "Trying to make the hippo dance is not really a lot of fun".[14]

Apple software makes heavy use of Objective-C, but the Objective-C front-end in GCC is a low priority for GCC developers. Also, GCC does not integrate smoothly into Apple's integrated development environment (IDE).[15] Finally, GCC is licensed under the terms of GNU General Public License (GPL) version 3, which requires developers who distribute extensions for, or modified versions of, GCC to make their source code available, whereas LLVM has a BSD-like license[16] which does not have such a requirement.

Apple chose to develop a new compiler front end from scratch, supporting C, Objective-C and C++.[15] This "clang" project was open-sourced in July 2007.[17]

Clang is intended to work on top of LLVM.[16] The combination of Clang and LLVM provides most of the toolchain, to allow replacing the full GCC stack. Because it is built with a library-based design, like the rest of LLVM, Clang is easy to embed into other applications. This is one reason why most OpenCL implementations are built with Clang and LLVM.

One of Clang's main goals is to provide a library-based architecture,[18] to allow the compiler to be more tightly tied to tools that interact with source code, such as an integrated development environment (IDE) graphical user interface (GUI). In contrast, GCC is designed to work in a compile-link-debug workflow, and integrating it with other tools is not always easy. For instance, GCC uses a step called fold that is key to the overall compile process, which has the side effect of translating the code tree into a form that looks unlike the original source code. If an error is found during or after the fold step, it can be difficult to translate that back into one location in the original source. Also, vendors using the GCC stack within IDEs use separate tools to index the code, to provide features like syntax highlighting and autocomplete.

Clang is designed to retain more information during the compiling process than GCC, and to preserve the overall form of the original code. The goal of this is to make it easier to map errors back into the original source. The error reports offered by Clang are also aimed to be more detailed and specific, as well as machine-readable, so IDEs can index the output of the compiler during compiling. Modular design of the compiler can offer source code indexing, syntax checking, and other features normally associated with rapid application development systems. The parse tree is also more suitable for supporting automated code refactoring, as it directly represents the original source code.

Clang compiles only C-like languages, such as C, C++, Objective-C, Objective-C++, OpenCL, CUDA, and HIP. For other languages, like Ada, LLVM remains dependent on GCC or another compiler frontend. In many cases, Clang can be used or swapped out for GCC as needed, with no other effects on the toolchain as a whole. It supports most of the commonly used GCC options. A sub-project Flang by Nvidia and The Portland Group added Fortran support.[19]

Clang is designed to be highly compatible with GCC.[9] Clang's command-line interface is similar to and shares many flags and options with GCC. Clang implements many GNU language extensions and enables them by default. Clang implements many GCC compiler intrinsics purely for compatibility. For example, even though Clang implements atomic intrinsics which correspond exactly with C11 atomics, it also implements GCC's __sync_* intrinsics for compatibility with GCC and libstdc++. Clang also maintains ABI compatibility with GCC-generated object code. In practice Clang can often be used as a drop-in replacement for GCC.[20]

Clang's developers aim to reduce memory footprint and increase compilation speed compared to competing compilers, such as GCC. In October 2007, they report that Clang compiled the Carbon libraries more than twice as fast as GCC, while using about one-sixth GCC's memory and disk space.[21] However, as of 2011 this was not a typical result.[22][23] As of mid-2014, Clang won more than a third of the benchmarks, with GCC winning most.[24] As of 2014, performance of Clang-compiled programs lagged behind performance of the GCC-compiled program, sometimes by large factors (up to 5.5x),[24] replicating earlier reports of slower performance.[22]

More recent comparisons in November 2016 indicate that both compilers have evolved to increase their performance. As of GCC 4.8.2 versus clang 3.4, on a large harness of test files, GCC outperforms clang by approximately 17% on well-optimized source code. Test results are code-specific, and unoptimized C source code can reverse such differences. The two compilers now seem broadly comparable.[25]

This table presents only significant steps and releases in Clang history.

• AMD Optimizing C/C++ Compiler
• LLDB
• Portable C Compiler

• Official website
• LLVMdev: New LLVM C front-end: "clang", announcement (11 July 2007)
• Presentation: Ted Kremenek - Finding Bugs with the Clang Static Analyzer, Slides
• Presentation: Steve Naroff - Clang Internals, Slides
• 2009 DevMtg Clang presentation