Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

adding a method to Base.Int can result in (seemingly) unrelated recompilation #29166

Closed
rfourquet opened this issue Sep 13, 2018 · 1 comment
Closed

adding a method to Base.Int can result in (seemingly) unrelated recompilation #29166

rfourquet opened this issue Sep 13, 2018 · 1 comment
Assignees
@JeffBezanson
Labels
compiler:latency Compiler latency regression Regression in behavior compared to a previous version

Comments

@rfourquet
Copy link
Member

This is a MWE:

struct X <: Integer end
Base.Int(::X) = 0

After entering this in the REPL, it becomes annoyingly slow (even for simple things as entering 1+2), while Julia apparently recompiles a bunch of functions. This problem apparently disappears for other Base builtin integers, or when using e.g. Real instead of Integer as the supertype.
Also, this doesn't happen on 0.7, so I will label as a regression.
@rfourquet rfourquet added the regression Regression in behavior compared to a previous version label Sep 13, 2018
@JeffBezanson JeffBezanson added the compiler:latency Compiler latency label Sep 13, 2018
@JeffBezanson JeffBezanson self-assigned this Jan 24, 2019
JeffBezanson added a commit that referenced this issue Jan 24, 2019
@JeffBezanson
improve some type information to reduce backedges; fixes #29166
23ef758 
reuse `Base._array_for` in lowering instead of manually inlining it
@JeffBezanson JeffBezanson mentioned this issue Jan 24, 2019
Merged
@JeffBezanson
Copy link
Member

I'll note that this is not really a bug --- AFAICT we were not invalidating any methods incorrectly --- but it did point to some places we could improve the typing of Base code. This new definition invalidates anything that calls Int(x::Integer). In general it is not easy to guess how many inferred methods might contain that, or any other function/type pair.

JeffBezanson added a commit that referenced this issue Jan 25, 2019
@JeffBezanson
improve some type information to reduce backedges; fixes #29166
7a488dd 
reuse `Base._array_for` in lowering instead of manually inlining it
JeffBezanson added a commit that referenced this issue Jan 28, 2019
@JeffBezanson
improve some type information to reduce backedges; fixes #29166
055c0cc 
reuse `Base._array_for` in lowering instead of manually inlining it

[ci skip]
JeffBezanson added a commit that referenced this issue Feb 7, 2019
@JeffBezanson
improve some type information to reduce backedges; fixes #29166
ed6e3fc 
reuse `Base._array_for` in lowering instead of manually inlining it
JeffBezanson added a commit that referenced this issue Feb 7, 2019
@JeffBezanson
improve some type information to reduce backedges; fixes #29166
560bf4a 
reuse `Base._array_for` in lowering instead of manually inlining it
@timholy timholy mentioned this issue Jun 30, 2020
Merged
Keno pushed a commit that referenced this issue Jun 5, 2024
@JeffBezanson
improve some type information to reduce backedges; fixes #29166
1ef2e75 
reuse `Base._array_for` in lowering instead of manually inlining it
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@JeffBezanson JeffBezanson

Labels
compiler:latency Compiler latency regression Regression in behavior compared to a previous version
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@JeffBezanson @rfourquet