Fix correctness issues in sizeof tfunc #36727
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Keno
added
compiler:codegen
base/compiler/tfuncs.jl
Outdated
| return false | ||
| end | ||
| if t !== Bottom | ||
| # The value corresponding to `x` at runtime could be a type. |
There was a problem hiding this comment.
I think this still gives the wrong answer for e.g. sizeof_nothrow(Type{Vector{Int}}) (or Type{String}). sizeof(Vector{Int}) throws.
The `Core.sizeof` function can take either a value or a type, which can be a bit confusing in the tfunc, because the tfunc operates on the types of the values passed to the builtin. In particular, we were incorrectly returning Const(16) (on 64bit) for sizeof_tfunc(UnionAll), because it was treating it the asme as `sizeof_tfunc(Const(UnionAll))`. Try to clean that up as well as fixing a similar confusion in the nothrow version of this function. Lastly, we had a similar fast path in codegen, which would try to inline the actual size for a constant datatype argument. For codegen, just rm that whole code path, since it should have no more information than inference and figuring it out in inference exposes strictly more optimization opportunities. Fixes #36710
|
@JeffBezanson I've updated this to more explicitly track the logic in the builtin. See what you think. |
JeffBezanson
approved these changes
Jul 20, 2020
JeffBezanson
pushed a commit
that referenced
this pull request
Jul 21, 2020
The `Core.sizeof` function can take either a value or a type, which can be a bit confusing in the tfunc, because the tfunc operates on the types of the values passed to the builtin. In particular, we were incorrectly returning Const(16) (on 64bit) for sizeof_tfunc(UnionAll), because it was treating it the asme as `sizeof_tfunc(Const(UnionAll))`. Try to clean that up as well as fixing a similar confusion in the nothrow version of this function. Lastly, we had a similar fast path in codegen, which would try to inline the actual size for a constant datatype argument. For codegen, just rm that whole code path, since it should have no more information than inference and figuring it out in inference exposes strictly more optimization opportunities. Fixes #36710 (cherry picked from commit 004cb25)
JeffBezanson
pushed a commit
that referenced
this pull request
Jul 22, 2020
The `Core.sizeof` function can take either a value or a type, which can be a bit confusing in the tfunc, because the tfunc operates on the types of the values passed to the builtin. In particular, we were incorrectly returning Const(16) (on 64bit) for sizeof_tfunc(UnionAll), because it was treating it the asme as `sizeof_tfunc(Const(UnionAll))`. Try to clean that up as well as fixing a similar confusion in the nothrow version of this function. Lastly, we had a similar fast path in codegen, which would try to inline the actual size for a constant datatype argument. For codegen, just rm that whole code path, since it should have no more information than inference and figuring it out in inference exposes strictly more optimization opportunities. Fixes #36710 (cherry picked from commit 004cb25)
simeonschaub
pushed a commit
to simeonschaub/julia
that referenced
this pull request
Aug 11, 2020
The `Core.sizeof` function can take either a value or a type, which can be a bit confusing in the tfunc, because the tfunc operates on the types of the values passed to the builtin. In particular, we were incorrectly returning Const(16) (on 64bit) for sizeof_tfunc(UnionAll), because it was treating it the asme as `sizeof_tfunc(Const(UnionAll))`. Try to clean that up as well as fixing a similar confusion in the nothrow version of this function. Lastly, we had a similar fast path in codegen, which would try to inline the actual size for a constant datatype argument. For codegen, just rm that whole code path, since it should have no more information than inference and figuring it out in inference exposes strictly more optimization opportunities. Fixes JuliaLang#36710
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The
Core.sizeoffunction can take either a value or a type,which can be a bit confusing in the tfunc, because the tfunc
operates on the types of the values passed to the builtin.
In particular, we were incorrectly returning Const(16) (on 64bit)
for sizeof_tfunc(UnionAll), because it was treating it the asme
as
sizeof_tfunc(Const(UnionAll)). Try to clean that up as wellas fixing a similar confusion in the nothrow version of this
function. Lastly, we had a similar fast path in codegen, which
would try to inline the actual size for a constant datatype argument.
For codegen, just rm that whole code path, since it should have no
more information than inference and figuring it out in inference
exposes strictly more optimization opportunities.
Fixes #36710