I am a student interested in benchmarking the performance of TFHE-rs in various environments.
I would like to know if TFHE-rs can be run on a Mac equipped with an ARM64-based M3 processor.
I executed the make bench_boolean command and noticed that the nightly-avx512 flag was included.
We know, ARM-based CPUs do not support AVX instructions, this seems to be the issue.
Are there any solutions or workarounds for this?
hello @kr3951
it should work, the nightly-avx512 has only an effect on x86 machines
do note that there are no specific optimizations for arm chips.
Cheers
Thank you for your quick response!
After rechecking the error message, it seems that the build command cargo nightly-2024-08-19 bench cannot work because nightly-2024-08-19 is not found.
When I remove nightly...., the error changes to error![feature] may not be used on the stable release channel and stop to compile the bench_boolean.
Anyway, what I was curious about is whether ARM SIMD instructions such as NEON are supported and how much acceleration can be expected from them.
But, do you mean that tfhe-rs does not supported yet, right?
Thank you!
you can install the required nightly toolchain with
make install_rs_check_toolchain
then you can relaunch the command
We don’t have manual ARM SIMD optimizations but LLVM can and does autovectorize stuff, so it will benefit from some ARM SIMD but we don’t have dedicated manual code for that in our code base @kr3951
It works! Thank you
Have a nice day!! 
Glad it works !
have a nice day too