Investigate EIgen Performance Improvements for Tensorflow (understanding Neon intrinsics)

The resources used in understanding Neon intrinsics were the following:

1. ARM Neoverse N1 software optimisation guide

2. Neon instrinsics reference https://developer.arm.com/architectures/instruction-sets/intrinsics/?page=4&search=vadd

3. ARM SIMD vector instructions set https://developer.arm.com/documentation/100076/0100/a64-instruction-set-reference/a64-simd-vector-instructions?lang=en

4. Neon programmer’s guide https://developer.arm.com/documentation/den0018/a

5. Optimising C code with Neon intrinics https://developer.arm.com/documentation/102467/0100

6. Read the Neon introduction document https://developer.arm.com/documentation/102474/latest

7. Details of auto-vectorisation for Neon https://developer.arm.com/documentation/102525/0100/Why-rely-on-the-compiler-for-auto-vectorization-

8. Arm NEON programming quick reference https://community.arm.com/developer/tools-software/oss-platforms/b/android-blog/posts/arm-neon-programming-quick-reference

The Neon intrinsics examples on basic signal processing were tried on d05 colo machine on Linaro’s work centre.

The main goal here was to understand NEON vectorisation instructions. The two options that are available to deploy are the actual SIMD instructions using assembly language instructions or using a functional C-API called NEON intrinsics, which allows calling the same instructions from within C++/C source code, and take advantage of its vectorisation capabilities.