We will be testing portability for the new 64-bit ARM architecture coming out in the near future and trying to apply the new auto-vectorization feature in the GCC Compiler to see how much of an impact it will make to optimize the code. During the planning phase I was tasked with looking for an open-source software where this could potentially be useful, the criteria of the search being software that potentially handles a large amount of data. After a very good consult, I decided to go with the xz software which is a general compression tool with a high compression ratio. You can find the link to the website over here:
The plan is to implement the changes to the compiler options. Run several test to test the speed using a software ARM 64-bit emulator, and check the differences in performance and if the changes don't break anything. The results should help us determine whether or not the auto-vectorization feature is viable for this software and does not have any other strange side effects. I will try to merge the changes into the main stream if everything is successful using the hopefully positive results to advocate for it, though the results themselves will not represent the full effect of these changes due to the fact that 64-bit ARM is not yet out and we are using an emulator to emulate the effects. In the next blog I will be showing the changes I make and the various tests and results ran to show the benefits of the new auto-vectorization feature. Since this will be implemented on the ARM V9 using SIMD it could possibly be applied to every device that uses that architecture and ground breaking depending on application, so hopefully everything works out.