AVX2 Optimization Support for Tensor Calculations in the Go Compiler

[alph4b3th]

Proposal Details

I would like to propose the implementation of a native optimization in the Go compiler that leverages AVX2 instructions for tensor calculations. This enhancement could offer significant performance improvements, especially for projects involving intensive data processing, artificial intelligence, machine learning, and scientific computations.

Integrating AVX2 would allow for more efficient execution of mathematical operations, contributing to better scalability and overall performance in Go-based systems. I believe this update would not only modernize the compiler but also meet the growing demands for high computational efficiency.

I am available to discuss technical details and provide practical examples that may assist in the implementation of this feature. Thank you for your attention and your ongoing efforts to enhance the Go language.

[Jorropo]

This lacks important details like are we talking about writing assembly (already possible and widely used in the ecosystem), intrinsics (see #67520) or autovectorization (does not require a proposal) ?