# Multi-threading + curve\_fit no acceleration but increase the memory consumption on my server

**URL:** https://discourse.julialang.org/t/multi-threading-curve-fit-no-acceleration-but-increase-the-memory-consumption-on-my-server/107429
**Category:** Performance
**Created:** [December 11, 2023, 11:40am UTC](https://discourse.julialang.org/t/multi-threading-curve-fit-no-acceleration-but-increase-the-memory-consumption-on-my-server/107429 "2023-12-11T11:40:11Z")
**Posts on this page:** 1
**Page:** 1

<div class="post-metadata">

### Author: ![aTrotier](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/atrotier/32/32165_2.png) [@aTrotier](https://discourse.julialang.org/u/aTrotier)
#### Post date: [December 11, 2023, 11:40am UTC](https://discourse.julialang.org/t/multi-threading-curve-fit-no-acceleration-but-increase-the-memory-consumption-on-my-server/107429/1 "2023-12-11T11:40:11Z")

</div>

Hi,

I am working with 3D + temporal dimension (128_128_96\*50) MRI datasets that I want to fit using curve\_fit with the following model.

The memory consumption is particularly important using multi-threading and does not improve the time to fit the whole dataset.

```julia
ima = Matrix(reshape(abs.(MRI_image, :, 50)')
T= eltype(ima)

function test_reco(ima,TE_vec,L)
   model_fit(t, p) = sqrt.((p[1] * exp.(-t / p[2])) .^ 2 .+ 2 * L * p[3]^2)
   Threads.@threads for i in axes(ima,2)
       y=view(ima,:,i)
       p0 = [maximum(y), T.(30),minimum(y)]
       fit = curve_fit(model_fit, TE_vec, y, p0,autodiff=:forwarddiff).param
   end
end

@time test_reco(ima,TE_vec,4)

```

which gives

> 232.342639 seconds (584.71 M allocations: 143.190 GiB, 9.08% gc time)

and without Threads.@threads

> 203.611481 seconds (584.71 M allocations: 143.190 GiB, 8.70% gc time)

Any advice ?