AMD Optimizing C/C++ Compiler

Last updated
Developer(s) Advanced Micro Devices
Initial release2017;8 years ago (2017)
Stable release
5.0.0 / October 10, 2024;9 months ago (2024-10-10)
Operating system Linux
Type Compiler
License Freeware
Website developer.amd.com/amd-aocc/

The AMD Optimizing C/C++ Compiler (AOCC) is an optimizing C/C++ and Fortran compiler suite from AMD targeting 32-bit and 64-bit Linux platforms. [1] [2] It is a proprietary fork of LLVM + Clang with various additional patches to improve performance for AMD's Zen microarchitecture in Epyc, and Ryzen microprocessors. [1] [3]

Contents

Release history

AOCC version history [4]
AOCC versionDateLLVM versionNotes
1.0.0May 20176.0Uses DragonEgg with GCC 4.8.2 for Fortran. [1]
1.1.0December 20176.0
1.2.0April 20186.0Introduction of the Flang Fortran frontend.
1.2.1July 20186.0Bugfix release.
1.3.0December 20187.0Last version to use DragonEgg.
2.0.0August 20198.0Support for 2nd generation Zen CPUs. [5]
2.1.0November 20199.0 [6]
2.2.0June 202010.0
2.3.0December 202011.0Introduction of OpenMP debugger
3.0.0March 202112.0Support for 3rd generation Zen CPUs
3.1.0July 202112.0
3.2.0December 202113.0
4.0.0November 202214.0Support for 4th generation Zen CPUs; Fortran 2008 and Fortran OMP 4.5 compliance improvements; Fortran debuggability improvements; improved variants of scalar, vector, and loop transformations [7]
4.1.0August 202316.0Compliance to C++17 standards; default optimization level changed to -O2; -fPIC and -fPIE options made default; bit reproducibility support improved for C/C++ and Fortran; general improvements targeting compile time reduction; new option to display the AOCC Optimization Report; tuned for use with the AMD Math Library v4.1 [8]
4.2.0February 202416.0Bug fixes with Fortran pre-processing and compilation; bug fixes impacting performance of C, C++, and Fortran applications; tuned for use with the AMD Math Library v4.2 [9]
5.0.0October 202417.0.6Support for 5th generation Zen CPUs; OpenMP 5.0 for C/C++; improved optimizations [10]

In a May 2017 benchmark comparing AOCC v1.0 to Clang 4 and 5, and GCC 6 through 8, Phoronix found AOCC provided marginal improvement over Clang 4.0 in several benchmarks and no difference in others. [3] Compilation time generally increased relative to Clang 4.0. Some benchmarks found some versions of GCC had better performance than some versions of Clang (AOCC included), and vice versa. In August 2019 Phoronix again benchmarked AOCC, now AOCC 2.0 against Clang 9 and GCC 9.1 and 10.0. [11]

Along with the compiler, AMD offers the AMD Optimizing CPU Libraries (AOCL), a set of numerical libraries that is roughly similar to Intel's Math Kernel Library and includes AMD Math Library (LibM), AMD Random Number Generator Library, AMD Secure RNG Library, AOCL-Sparse, BLIS, FFTW, libFLAME, and ScaLAPACK that are tuned for the Zen processors.

AMD also offers AMD μProf, a performance analysis tool similar to Intel VTune.

See also

References

  1. 1 2 3 "AOCC User guide" (PDF). AMD.
  2. ↑ "AMD Releases Optimizing C/C++ Compiler For Ryzen - Phoronix". www.phoronix.com.
  3. 1 2 "Benchmarking AMD's New AOCC Compiler For Ryzen - Phoronix". www.phoronix.com.
  4. ↑ "AOCC Archive Section". AMD.
  5. ↑ "AOCC 2.0 Install Guide" (PDF). developer.amd.com. Retrieved 2020-02-19.
  6. ↑ "AOCC 2.1 Install Guide" (PDF). developer.amd.com. Retrieved 2020-02-19.
  7. ↑ "AOCC 4.0 Release Notes" (PDF). www.amd.com. Retrieved 2024-04-08.
  8. ↑ "AOCC 4.1 Release Notes" (PDF). www.amd.com. Retrieved 2024-04-08.
  9. ↑ "AOCC 4.2 Release Notes" (PDF). www.amd.com. Retrieved 2024-04-08.
  10. ↑ "AOCC 5.0 Release Notes (New Features/Enhancements)". www.amd.com. Retrieved 2024-10-24.
  11. ↑ "GCC vs. LLVM Clang vs. AOCC Compiler Benchmarks On The AMD EPYC 7742 2P Linux Server - Phoronix". www.phoronix.com. Retrieved 2020-02-19.


This page is based on this Wikipedia article
Text is available under the CC BY-SA 4.0 license; additional terms may apply.
Images, videos and audio are available under their respective licenses.