# How to improve performance in a function that repeatedly defines and multiplies matrices **URL:** \ **Category:** Julia at Scale\ **Tags:** performance, parallel, linearalgebra, complex-numbers, matrix\ **Created:** [December 29, 2023, 9:47am UTC](https://discourse.julialang.org/t/how-to-improve-performance-in-a-function-that-repeatedly-defines-and-multiplies-matrices/108153 "2023-12-29T09:47:43Z")\ **Posts on this page:** 1\ **Showing post:** 33

**Author:** ![abraemer](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/abraemer/32/51403_2.png) [@abraemer](https://discourse.julialang.org/u/abraemer)\ **Post date:** [January 4, 2024, 10:03am UTC](https://discourse.julialang.org/t/how-to-improve-performance-in-a-function-that-repeatedly-defines-and-multiplies-matrices/108153/33 "2024-01-04T10:03:34Z")

Globals are only problematic if you access them repeatedly. Note in my code, that I only ever access them once in the moment where I call `calc`. `calc` itself does not access globals and `GHT` and `GFC` also shouldn’t. That is the very important difference! > [@Uranium238](#): > > The `@threads` seemed to provide inferior performance than the later. That is why I was discouraged from using that. `@threads` is not the only way to get thread-based parallelism and is really only a convenience tool. In your [other thread](https://discourse.julialang.org/t/incorrect-output-when-using-base-threads/108290/3) I explained how to use `Threads.@spawn` together with `ChunkSplitters.jl` to split the workload into chunks and preallocate workspaces for each chunk. Addendum: Also if you use things from `LinearAlgebra.jl` that use BLAS, then this opens another can of worms regarding threads… --- _[View the full topic](https://discourse.julialang.org/t/how-to-improve-performance-in-a-function-that-repeatedly-defines-and-multiplies-matrices/108153)._