Very slow time to first plot, 2022

Is it possible to see which sysimage is loaded in an active REPL session? I am unsure if VSCode is really loading that image which took 15 minutes to build…

1 Like

You could follow the example from the docs here:

❯ julia -q

julia> Base.loaded_modules
Dict{Base.PkgId,Module} with 34 entries:
  Example [7876af07-990d-54b4-ab0e-23690620f79a] => Example

(where in the example the Example package was included in the sysimage)

You do need to go into the vs code plug-in settings and check the checkbox that tells it to look for the sys image and load it there’s a discussion of this in the docs somewhere

1 Like

Ah that makes sense

According to the NEWS file in Julia 1.9, loading times are going to improve a lot:

  • Time to first execution (TTFX, sometimes called time to first plot) is greatly reduced. Package precompilation now saves native code into a “pkgimage”, meaning that code generated during the precompilation process will not require compilation after package load. Use of pkgimages can be disabled via --pkgimages=no (#44527) (#47184).
    julia/ at v1.9.0-rc3 · JuliaLang/julia · GitHub

Is Makie going to benefit from it?
Does anyone know when Julia 1.9 will be released?



You can start using now, though (I think many people already did). In particular, if you install Julia with juliaup, it is very easy to keep the latest release candidate of 1.9 and any other julia version you are using installed simultaneously.


I’ve been using all the 1.9 rcs. I’ve seen nothing in my normal day to day use that makes the latest 1.9rcs feel less robust than a normal release, and have already switched to using them full time.

I can confirm that for the first time, I am using the release candidates in daily usage to get the improved TTFX.

I installed 1.9 rc3 but still got slow TTFP: Around 6 seconds to load CairoMakie and around 5 seconds for the first plot.
Anyway, I noticed a big improvement in the package loading time which was more than 20 seconds.

I know that these times are long compared to many other programming languages, so it is honestly fair to call it slow. It is hower a quite MASSIVE improvement of previous timings for Makie in particular, so it feels fast to me. Funny how relative “fast” is…

1 Like

The TTFX issue is pretty well discussed. When X=“simple interactive activity, like plotting some data”, Julia will probably never be as fast as some alternatives. When X=“meaningful, moderately complex numerical computations”, Julia is already competitive with best of the best.

Regarding 1.9, I bet it’s out “real soon now”.


You might be right :wink: