Reorganize some hvcat functions #16741
Conversation
|
Thanks. However I think this might cause the definitions in sparse/sparsematrix.jl to be called, since they are more specific. I think that's why these were added. |
|
Yes, I was thinking about this. I'll see what I can do about it. |
pabloferz
changed the title
pabloferz
force-pushed
the
pz/cat
branch
2 times, most recently
from
d19994a to
72c5219
Compare
|
Yes they were added specifically because the specialized methods would have been called. Maybe I used too specific types? Would it have sufficed with |
|
I don't think the building should give an ambiguous definition error. Is it a bug in inference or I am missing something obvious? |
pabloferz
force-pushed
the
pz/cat
branch
2 times, most recently
from
fa17dcf to
d1843c7
Compare
|
Ok. @JeffBezanson I think this is ready and addresses the issue with |
| hvcat(rows::Tuple{Vararg{Int}}, xs::AbstractMatrix...) = typed_hvcat(promote_eltype(xs...), rows, xs...) | ||
| hvcat{T}(rows::Tuple{Vararg{Int}}, xs::AbstractMatrix{T}...) = typed_hvcat(T, rows, xs...) | ||
|
|
||
| function typed_hvcat{T}(TT::Type{T}, rows::Tuple{Vararg{Int}}, as::AbstractMatrix{T}...) |
There was a problem hiding this comment.
This requires all the argument matrices to have the output type T as their element type already.
Also could you keep the code in place above to get a smaller diff?
There was a problem hiding this comment.
Fixed. There was an apparently unrelated AV error.
|
Thank you! |
This PR reorganize some
hvcatfunctions andremoves some duplicates of somefixes some problems that were introduced in commit e280a27. These cause some cat function to fallback to a generic (slow) version.catoperations that exist in bothabstractarray.jlandarray.jlThis should help with at least the
rand_mat_mulin #16128.