@jk good arrays are typically arranged sequentially in standard memory and often large amounts of memory are read at a time

You will likely see slightly faster runtimes if each thread is reading the arrays in memory contiguous order- so it's better to provide 8 arrays to 8 processes, instead of trying to segment the arrays by an indexed space per array

Sign in to participate in the conversation
Awoo Space

Awoo.space is a Mastodon instance where members can rely on a team of moderators to help resolve conflict, and limits federation with other instances using a specific access list to minimize abuse.

While mature content is allowed here, we strongly believe in being able to choose to engage with content on your own terms, so please make sure to put mature and potentially sensitive content behind the CW feature with enough description that people know what it's about.

Before signing up, please read our community guidelines. While it's a very broad swath of topics it covers, please do your best! We believe that as long as you're putting forth genuine effort to limit harm you might cause – even if you haven't read the document – you'll be okay!