TCC has a number of features that differentiate it from other current C compilers:
Its small file size and memory footprint allow it to be used directly from a single 1.44 M floppy disk, such as a rescue disk.
TCC is intended to produce native x86, x86-64 and ARM code very quickly; according to Bellard, it compiles, assembles and links about nine times faster than GCC does.
TCC has a number of compiler-specific language features intended to improve its practicality, such as an optional memory and bound checker, for improved code stability.
In general, TCC's implementation emphasizes smallness instead of optimally-performing results. TCC generates code in a single pass, and does not perform most of the optimizations performed by other compilers. TCC compiles every statement on its own, and at the end of each statement register values are written back to the stack and must be re-read even if the next line uses the values in registers. TCC uses only some of the available registers. TCC performs a few optimizations, such as constant propagation for all operations, multiplications and divisions are optimized to shifts when appropriate, and comparison operators are specially optimized. It also does some simple register allocation, which prevents many extraneous save/load pairs inside a single statement. Here are two benchmark examples:
A recursive Fibonacci algorithm on a 1.8 GHz Intel Centrino laptop with 512 MB RAM yields a noticeable difference in results between Microsoft Visual C++ compiler 13.10.3052 and TCC. To calculate the 49th Fibonacci number, it took a MS Visual C++ program approximately 18% longer than the TCC compiled program.
With a tcc modified to compile GCC, running cc1 on itself required 518 seconds when compiled using GCC 3.4.2, 558 seconds using GCC 2.95.3, 545 using Microsoft C compiler, and 1145 seconds using tcc. The level of optimization in each compiler was -O1 or similar.
Uses
TCCBOOT, a hack where TCC loads and boots a Linux kernel from source in about 10 seconds. That is to say, it is a "boot loader" that reads Linux kernel source code from disk, writes executable instructions to memory, and begins running it. This did require changes to the Linux build process.
TCC was used to demonstrate a defense against the trust attack.
TCC has been used to compile GCC, though various patches were required to make this work.
Cinpy is a Python library that allows you to implement functions with C in Python modules. The functions are compiled with TCC at runtime. The results are made callable in Python through the ctypes library.
Comes installed on JavaScript Linux.
Has been used as a reference for the compiled version of the super micro-max chess program source.
History
TCC has its origins in the Obfuscated Tiny C Compiler, a program Bellard wrote to win the International Obfuscated C Code Contest in 2001. After that time, Bellard expanded and deobfuscated the program to produce tcc. At some time prior to 4 February 2012 Fabrice Bellard updated the project's official web page to report that he was no longer working on TCC. Since Bellard's departure from the project, various people and groups have distributed patches or maintained forks of TCC to build upon or fix issues with TCC. This includes Dave Dodge's collection of unofficial tcc patches, Debian and kfreebsd downstream patches, and grischka's gcc patches. Grischka also set up a public Git repository for the project that contains a mob branch where numerous contributions, including a shared build, cross-compilers, and SELinux compatibility were added. Grischka's GIT repository later became the official TCC repository.
Current status
As of December 2017 both the official TCC mailing list and the official Git repository show active discussion and development by many developers and interested users. In December 2017, grischka announced on the mailing list that TCC version 0.9.27 was released.
