Fully drop _tuple_any and improve type-based offset axes check.
#45260
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
N5N3
force-pushed
the
remove_tuple_any
branch
3 times, most recently
from
8c42ee9
to
8dd18c3
Compare
aviatesk
reviewed
May 11, 2022
N5N3
force-pushed
the
remove_tuple_any
branch
2 times, most recently
from
9d37b62
to
8a02e52
Compare
vtjnash
reviewed
May 11, 2022
base/range.jl
Outdated
| @@ -687,6 +687,13 @@ step_hp(r::AbstractRange) = step(r) | |||
|
|
|||
| axes(r::AbstractRange) = (oneto(length(r)),) | |||
|
|
|||
| # Needed to ensure `has_offset_axes` can constant-fold. | |||
| has_offset_axes(::StepRange) = false | |||
| let baseints = Union{Int8,UInt8,Int16,UInt16,Int32,UInt32,Int64,UInt64,Int128,UInt128} | |||
There was a problem hiding this comment.
Suggested change
| let baseints = Union{Int8,UInt8,Int16,UInt16,Int32,UInt32,Int64,UInt64,Int128,UInt128} | |
| let baseints = BitInteger |
There was a problem hiding this comment.
BitInteger is defined later in int.jl. Maybe we can replace it with Union{bigints, smallints}
base/range.jl
Outdated
| has_offset_axes(::StepRange) = false | ||
| let baseints = Union{Int8,UInt8,Int16,UInt16,Int32,UInt32,Int64,UInt64,Int128,UInt128} | ||
| global firstindex | ||
| firstindex(::StepRange{T,<:baseints}) where {T<:baseints} = sizeof(T) < sizeof(Int) ? 1 : one(T) |
There was a problem hiding this comment.
T might be a Union and not have a size?
There was a problem hiding this comment.
Well, we should use sizeof(first(r)) instead.
The new length has similar problem.
To make sure the result is correct, we'd better split L694 into:
firstindex(r::StepRange{<:bigints,<:BitInteger}) = one(last(r)-first(r))
firstindex(::StepRange{<:smallints,<:BitInteger}) = 1| @@ -799,7 +806,7 @@ let smallints = (Int === Int64 ? | |||
| function length(r::OrdinalRange{<:smallints}) | |||
| s = step(r) | |||
| isempty(r) && return 0 | |||
| return div(Int(last(r)) - Int(first(r)), s) + 1 | |||
| return Int(div(Int(last(r)) - Int(first(r)), s)) + 1 | |||
There was a problem hiding this comment.
Are we sure it is legal to cast this to an Int? This seems highly unrelated to the commit message
There was a problem hiding this comment.
This is for length(::StepRange{Int8,Int128}). Without this, we'll return a Int128 for non-empty r.
I believe it's safe as we convert the div result rather than step s. Int should be enough to represent the length
vtjnash
reviewed
May 11, 2022
N5N3
force-pushed
the
remove_tuple_any
branch
3 times, most recently
from
987a9b0
to
ab36599
Compare
`_any_tuple` should be functionally consistent. (And has no bootstrap problem.)
And define some "cheap" `firstindex`
This avoid some unneeded `axes` call, thus more possible be folded by the compiler.
Co-Authored-By: Jameson Nash <[email protected]>
And define `firstindex` accordingly. Co-Authored-By: Jameson Nash <[email protected]>
ffucci
pushed a commit
to ffucci/julia
that referenced
this pull request
Aug 11, 2022
* Follow up to JuliaLang#45236 (make `length(::StepRange{Int8,Int128})` type-stable) * Fully drop `_tuple_any` (unneeded now) * Make sure `has_offset_axes(::StepRange)` could be const folded. And define some "cheap" `firstindex` * Do offset axes check on `A`'s parent rather than itself. This avoid some unneeded `axes` call, thus more possible be folded by the compiler. Co-authored-by: Jameson Nash <[email protected]>
pcjentsch
pushed a commit
to pcjentsch/julia
that referenced
this pull request
Aug 18, 2022
* Follow up to JuliaLang#45236 (make `length(::StepRange{Int8,Int128})` type-stable) * Fully drop `_tuple_any` (unneeded now) * Make sure `has_offset_axes(::StepRange)` could be const folded. And define some "cheap" `firstindex` * Do offset axes check on `A`'s parent rather than itself. This avoid some unneeded `axes` call, thus more possible be folded by the compiler. Co-authored-by: Jameson Nash <[email protected]>
This was referenced Jan 28, 2024
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.
Base._tuple_anyis only used inBase.has_offset_axes.After #44063, we have a similar method
_any_tuple, which should be functionally consistent for offset axes check.This PR fully removes
_tuple_any, and makeBase.has_offset_axescall_any_tuple.(Ideally, we'd better call public api
anyinstead. But our compiler fails to fold some check as a const.)This PR also omits some
axescall inhas_offset_axesby checkingindices/parent'saxesrather than self'saxes.This helps #44040 a little.
Test added.