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

move promote_type call inside the generated function body #208

Merged
merged 4 commits into from
Jul 10, 2018
Merged

move promote_type call inside the generated function body #208

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
09bcc55
move promote_type call inside the generated function body
rdeits Jun 20, 2018
1fcc722
also update hessian
rdeits Jun 20, 2018
c77760c
test coverage
rdeits Jun 20, 2018
92d5338
fix travis typo
rdeits Jul 3, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions .travis.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,3 +6,8 @@ julia:
matrix:
allow_failures:
- julia: nightly
after_success:
# push coverage results to Coveralls
- julia -e 'cd(Pkg.dir("WebIO")); Pkg.add("Coverage"); using Coverage; Coveralls.submit(Coveralls.process_folder())'
# push coverage results to Codecov
- julia -e 'cd(Pkg.dir("WebIO")); Pkg.add("Coverage"); using Coverage; Codecov.submit(Codecov.process_folder())'
12 changes: 8 additions & 4 deletions src/b-splines/indexing.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -108,9 +108,11 @@ end
for R in [:Real, :Any]
@eval @generated function gradient(itp::AbstractInterpolation{T,N}, xs::$R...) where {T,N}
n = count_interp_dims(itp, N)
Tg = promote_type(T, [x <: AbstractArray ? eltype(x) : x for x in xs]...)
xargs = [:(xs[$d]) for d in 1:length(xs)]
:(gradient!(Array{$Tg, 1}($n), itp, $(xargs...)))
quote
Tg = $(Expr(:call, :promote_type, T, [x <: AbstractArray ? eltype(x) : x for x in xs]...))
gradient!(Array{Tg, 1}($n), itp, $(xargs...))
end
end
end

Expand Down Expand Up @@ -158,9 +160,11 @@ end

@generated function hessian(itp::AbstractInterpolation{T,N}, xs...) where {T,N}
n = count_interp_dims(itp,N)
TH = promote_type(T, [x <: AbstractArray ? eltype(x) : x for x in xs]...)
xargs = [:(xs[$d]) for d in 1:length(xs)]
:(hessian!(Array{TH, 2}($n,$n), itp, $(xargs...)))
quote
TH = $(Expr(:call, :promote_type, T, [x <: AbstractArray ? eltype(x) : x for x in xs]...))
Copy link
Member

@stevengj stevengj Jul 3, 2018 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The problem is that this constructs an array and does the type computations at runtime. It may not be type-stable, either. It seems like you'd be better off doing something like TH = hessian_type(T, xs...) with

hessian_type(T) = T
hessian_type(T, x::AbstractArray, xs...) = promote_type(eltype(x), hessian_type(T, xs...))
hessian_type(T, x, xs...) = promote_type(typeof(x), hessian_type(T, xs...))

that can get evaluated at compile time.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Are you sure? That array is constructed inside the generator, not in the run-time code, and all the branching likewise happens inside the generator. I'm a fan of the lisp-y recursion trick anyway, but I'm pretty sure this case will work as-is.

Copy link
Member

@stevengj stevengj Jul 5, 2018 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Oh, right, I misread your code — you're right, there should continue to be no runtime cost here.

hessian!(Array{TH, 2}($n,$n), itp, $(xargs...))
end
end

hessian1(itp::AbstractInterpolation{T,1}, x) where {T} = hessian(itp, x)[1,1]