# The compilation time of max with dot distribution is kinda insane. workarounds?
**URL:** https://discourse.julialang.org/t/the-compilation-time-of-max-with-dot-distribution-is-kinda-insane-workarounds/138440
**Category:** General Usage
**Tags:** compilation
**Created:** [July 24, 2026, 4:55am UTC](https://discourse.julialang.org/t/the-compilation-time-of-max-with-dot-distribution-is-kinda-insane-workarounds/138440 "2026-07-24T04:55:02Z")
**Posts on this page:** 8
**Page:** 1
### Author:  [@rokke](https://discourse.julialang.org/u/rokke)
#### Post date: [July 24, 2026, 4:55am UTC](https://discourse.julialang.org/t/the-compilation-time-of-max-with-dot-distribution-is-kinda-insane-workarounds/138440/1 "2026-07-24T04:55:02Z")
I have a function that returns a tuple of two ints, and I want to find the largest int that shows up in each index. we’ve got this function in base called `max` which will do this for you, but the compilation time is kinda crazy:
```julia-auto
julia> @time max.((0,0), [(1,1) for i=1:100]...)
47.772725 seconds (14.32 M allocations: 549.282 MiB, 2.91% gc time, 99.41% compilation time)
(1, 1)
julia> # this is logically equivalent, but >500x faster
julia> @time (max(0, [1 for i=1:100]...), max(0, [1 for i=1:100]...))
0.078314 seconds (68.24 k allocations: 3.324 MiB, 99.28% compilation time)
(1, 1)
```
it also recompiles for each new set of arguments, so `max.((0,0), [(1,1) for i=1:101]...)` will take a long time again, and then so will `max.((0,0), [(1,1) for i=1:102]...)` after those two.
is there some way I can get max to not compile that long, other than the obvious
```julia-auto
julia> arr1 = arr2 = zeros(101)
julia> for (ind, i) in enumerate([(0, 0); [(1,1) for i=1:100]])
arr1[ind], arr2[ind] = i
end
julia> max(arr1...), max(arr2...)
```
I don’t want to run the tuple function (what I’ve been representing as `[(0, 0); [(1,1) for i=1:100]]`) more than I need to
* * *
* * *
```julia-auto
```
* * *
* * *
note `maximum([(0,0); [(1,1) for i=1:100]])` has different semantics. observe the difference here:
```julia-auto
julia> max.((0,0), [([1:100;][i],[100:-1:1;][i]) for i=1:100]...)
(100, 100)
julia> maximum([(0,0); [([1:100;][i],[100:-1:1;][i]) for i=1:100]])
(100, 1)
```
---
### Author:  [@rokke](https://discourse.julialang.org/u/rokke)
#### Post date: [July 24, 2026, 5:32am UTC](https://discourse.julialang.org/t/the-compilation-time-of-max-with-dot-distribution-is-kinda-insane-workarounds/138440/2 "2026-07-24T05:32:54Z")
aww, and here I was, thinking I was so smart. zip has the same issue:
```julia-auto
julia> @time collect(zip((0,0), [(1,1) for i=1:100]...))
203.997187 seconds (53.47 M allocations: 2.030 GiB, 2.30% gc time, 100.00% compilation time)
2-element Vector{NTuple{101, Int64}}:
(0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1)
(0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1)
```
---
### Author:  [@Sukera](https://discourse.julialang.org/u/Sukera)
#### Post date: [July 24, 2026, 5:34am UTC](https://discourse.julialang.org/t/the-compilation-time-of-max-with-dot-distribution-is-kinda-insane-workarounds/138440/3 "2026-07-24T05:34:38Z")
You’re splatting a 100 element long vector, which means you’re asking the compiler to create & optimize a function with 100 individual arguments. That’s likely the cause of the slowdown, not any individual function like `zip` or `max`.
---
### Author:  [@rokke](https://discourse.julialang.org/u/rokke)
#### Post date: [July 24, 2026, 5:55am UTC](https://discourse.julialang.org/t/the-compilation-time-of-max-with-dot-distribution-is-kinda-insane-workarounds/138440/4 "2026-07-24T05:55:49Z")
then yes, that’s what I’m asking how to not do. max works just fine on thousands of arguments if you don’t distribute, so I feel like there has to be a way to make this work…
```julia-auto
julia> @time max(1, [1 for i=1:10000]...)
0.059525 seconds (34.08 k allocations: 2.040 MiB, 94.97% compilation time)
1
```
---
### Author:  [@jules](https://discourse.julialang.org/u/jules)
#### Post date: [July 24, 2026, 6:08am UTC](https://discourse.julialang.org/t/the-compilation-time-of-max-with-dot-distribution-is-kinda-insane-workarounds/138440/5 "2026-07-24T06:08:24Z")
Maybe `maximum(first, x)` and `maximum(last, x)`
---
### Author:  [@rokke](https://discourse.julialang.org/u/rokke)
#### Post date: [July 24, 2026, 6:15am UTC](https://discourse.julialang.org/t/the-compilation-time-of-max-with-dot-distribution-is-kinda-insane-workarounds/138440/6 "2026-07-24T06:15:35Z")
nice! that’s much closer to being inline, gets rid of the big compile time
```julia-auto
julia> @time (let x=(([1:1000;][i],[1000:-1:1;][i]) for i=1:1000); maximum(first, x; init=0), maximum(last, x; init=0) end)
0.225317 seconds (220.40 k allocations: 40.799 MiB, 14.21% gc time, 82.82% compilation time)
(1000, 1000)
```
got it!
```julia-auto
julia> @time maximum.((first, last), Ref(([1:1000;][i],[1000:-1:1;][i]) for i=1:1000); init=0)
0.157316 seconds (197.25 k allocations: 39.709 MiB, 12.08% gc time, 84.11% compilation time)
(1000, 1000)
```
---
### Author:  [@cstjean](https://discourse.julialang.org/u/cstjean)
#### Post date: [July 24, 2026, 6:28am UTC](https://discourse.julialang.org/t/the-compilation-time-of-max-with-dot-distribution-is-kinda-insane-workarounds/138440/7 "2026-07-24T06:28:48Z")
`maximum` is the answer, but in general, this kind of problem is well solved by `reduce` and `mapreduce`. For instance: `mapreduce(first, max, [(1,1) for i=1:100], init=0)`
---
### Author:  [@wheeheee](https://discourse.julialang.org/u/wheeheee)
#### Post date: [July 25, 2026, 1:50am UTC](https://discourse.julialang.org/t/the-compilation-time-of-max-with-dot-distribution-is-kinda-insane-workarounds/138440/8 "2026-07-25T01:50:51Z")
yeah, `reduce(Base.BroadcastFunction(max), v; init=(0, 0))` works too and is somewhat “purer”
```julia-repl
julia> v = [(0,0); [(1, 1) for _ in 1:999]];
julia> @time reduce(Base.BroadcastFunction(max), v; init=0);
0.055134 seconds (167.51 k allocations: 8.329 MiB, 99.85% compilation time)
```