One of the exciting additions on the way for the C++26 programming language is a standardized library around Single Instruction Multiple Data (SIMD) operations. This portable SIMD implementation makes it easier to leverage SIMD and data parallelism in C++ for better performance and to work across SIMD architectures like AVX-512.
Posted to the GCC patches mailing list on Wednesday was work on the C++26 SIMD implementation. The patches implement a variety of the C++26 SIMD features but simd.loadstore, simd.permute.dynamic, and other features are missing while simd.math is only partially implemented at this time.
Dr. Matthias Kretz of the GSI Helmholtz Center for Heavy Ion Research was the one to…
One of the exciting additions on the way for the C++26 programming language is a standardized library around Single Instruction Multiple Data (SIMD) operations. This portable SIMD implementation makes it easier to leverage SIMD and data parallelism in C++ for better performance and to work across SIMD architectures like AVX-512.
Posted to the GCC patches mailing list on Wednesday was work on the C++26 SIMD implementation. The patches implement a variety of the C++26 SIMD features but simd.loadstore, simd.permute.dynamic, and other features are missing while simd.math is only partially implemented at this time.
Dr. Matthias Kretz of the GSI Helmholtz Center for Heavy Ion Research was the one to post the ten thousand lines of new libstdc++ code for working on this C++26 SIMD support.
Those interested in this C++26 SIMD work for GCC can find the initial patches on the gcc-patches mailing list.