Skip to content

Conversation

@JeffBezanson
Copy link
Contributor

On julia master this reduces the time to first plot for me by about 5 seconds. So far I have not detected any measurable slowdowns e.g. for plotting many points (the drawing itself inside gksterm is by far the limiting factor).

@codecov
Copy link

codecov bot commented Apr 6, 2020

Codecov Report

Merging #2544 into master will decrease coverage by 0.42%.
The diff coverage is n/a.

Impacted file tree graph

@@            Coverage Diff             @@
##           master    #2544      +/-   ##
==========================================
- Coverage   66.15%   65.72%   -0.43%     
==========================================
  Files          25       25              
  Lines        7395     7393       -2     
==========================================
- Hits         4892     4859      -33     
- Misses       2503     2534      +31
Impacted Files Coverage Δ
src/Plots.jl 32.25% <ø> (ø) ⬆️
src/examples.jl 49.09% <0%> (-3.64%) ⬇️
src/args.jl 70.94% <0%> (-2.32%) ⬇️
src/types.jl 53.84% <0%> (-1.93%) ⬇️
src/axes.jl 70.2% <0%> (-0.94%) ⬇️
src/backends/pgfplotsx.jl 64.43% <0%> (-0.93%) ⬇️
src/pipeline.jl 91.13% <0%> (-0.5%) ⬇️
src/recipes.jl 39.85% <0%> (-0.31%) ⬇️
src/series.jl 47.6% <0%> (-0.3%) ⬇️
src/components.jl 34.65% <0%> (-0.29%) ⬇️
... and 2 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ae396ef...6807f53. Read the comment docs.

@daschw
Copy link
Member

daschw commented Apr 6, 2020

Nice, thanks a lot!

@daschw daschw merged commit 45569c0 into JuliaPlots:master Apr 6, 2020
@ChrisRackauckas
Copy link
Member

Why not -O0? Does 1 offer some optimizations that helps Plots?

@mkborregaard
Copy link
Member

Another question: will this behaviour change if called by packagecompiler to form a system image?

@JeffBezanson
Copy link
Contributor Author

-O0 would be fine too. I made some changes to -O1 that make it very close to -O0 compiler performance while generating significantly better code, so I feel -O1 is overall a good default. It often pays for itself.

This option currently only affects the JIT; building a system image will only use whatever -O option is passed on the command line.

@mkborregaard
Copy link
Member

Perfect <3

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants