Avoid benchmark optimization on ARM by actually using the output array

alantech / alan

Autoscalable Programming Language

https://alan-lang.org

MIT License

305 stars 10 forks source link

Avoid benchmark optimization on ARM by actually using the output array #768

Closed dfellis closed 2 months ago

dfellis commented 2 months ago

This was discovered on the Raspberry Pi 5 when the single threaded performance was unbelievably fast. This change prevents the compiler from removing the actual calculation steps because now the output is actually used.