Why Apple Silicon runs so fast in ODE bench?

I just run the bench on my Macbook pro(m1) in Matlab bata which has full support for Apple Silicon. I found the ODE test result so surprising that it runs more than two times faster than any other computers. I have been wondering the reasons for a long time. Is there any special optimization on Apple Silicon or arm?

Answer 1

[SOUMNATH PAUL]

Hi,

The performance gains you're observing on your MacBook Pro with Apple Silicon (M1) during the ODE (Ordinary Differential Equations) test in MATLAB beta could be attributed to several factors:

1. The Apple M1 chip is built on a different architecture (ARM64) compared to the Intel x86 architecture. ARM processors, in general, have shown efficiency gains in certain workloads, including mathematical computations like solving ODEs.
2. MATLAB has likely been optimized to take advantage of the specific features and architecture of Apple Silicon. Apple provides tools and resources to developers to optimize their software for M1, and developers often release updates to make their applications run more efficiently on Apple's hardware.
3. The M1 chip features a unified memory architecture, which means that the CPU and GPU share the same memory. This can lead to faster data transfer between the CPU and GPU, potentially improving performance for certain computations.
4. ARM architectures, including Apple Silicon, support Single Instruction, Multiple Data (SIMD) instructions. These instructions allow parallel processing of multiple data elements in a single instruction. If MATLAB has been optimized to leverage SIMD instructions on Apple Silicon, it could result in improved performance for certain tasks. Below link provides more information on this https://en.wikipedia.org/wiki/Single_instruction,_multiple_data
5. The beta version you're using might be taking full advantage of the native support for Apple Silicon, which means that the software is compiled directly for the M1 architecture rather than being translated through Rosetta 2.
6. The M1 chip comes with a set of optimized math libraries. If MATLAB is utilizing these libraries efficiently, it could contribute to better performance in mathematical computations.

Kindly refer to this link for more information on using MATLAB from apple computers https://blogs.mathworks.com/matlab/2023/06/22/native-apple-silicon-support-in-the-matlab-simulink-r2023b-pre-release/

It's worth noting that performance gains can differ based on the specific task or workload. The observations you made during the ODE test may not necessarily be universally applicable to all types of computations. The overall enhancement in performance on Apple Silicon is likely attributed to a combination of factors, including hardware architecture, software optimization, and specific task characteristics.

Hope it helps!

Regards,

Soumnath