1.10 enablement #330
1.10 enablement #330
Conversation
| LinearAlgebra.lmul!(A::$t{T,<:oneStridedVecOrMat}, | ||
| B::oneStridedVecOrMat{T}) where {T<:onemklFloat} = | ||
| trmm!('L', $uploc, 'N', $isunitc, one(T), parent(A), B, B) | ||
| LinearAlgebra.rmul!(A::oneStridedVecOrMat{T}, | ||
| B::$t{T,<:oneStridedVecOrMat}) where {T<:onemklFloat} = | ||
| trmm!('R', $uploc, 'N', $isunitc, one(T), parent(B), A, A) |
There was a problem hiding this comment.
By the signature of trmm!, only matrix arguments are allowed, so either it should be relaxed over there or, as is done here, restricted here.
| @@ -1100,7 +1100,7 @@ function trmm(side::Char, | |||
| alpha::Number, | |||
| A::oneStridedMatrix{T}, | |||
| B::oneStridedMatrix{T}) where T | |||
| trmm!(side, uplo, transa, diag, alpha, A, B) | |||
| trmm!(side, uplo, transa, diag, alpha, A, copy(B)) | |||
There was a problem hiding this comment.
I think this was oversight, and the corresponding test required an according change.
| LinearAlgebra.generic_mattridiv!(C::oneStridedMatrix{T}, uploc, isunitc, tfun::Function, A::oneStridedMatrix{T}, B::oneStridedMatrix{T}) where {T<:onemklFloat} = | ||
| trsm!('R', uploc, tfun === identity ? 'N' : tfun === transpose ? 'T' : 'C', isunitc, one(T), B, C === A ? C : copyto!(C, A)) |
There was a problem hiding this comment.
This adds rdiv!-functionality, which could be included in the "old" branch below.
|
Nightly fails, but I don't understand why. Something with StaticArrays.jl? |
|
Perhaps we should add testing on v1.10-something also to see whether things work as intended after the alpha2-release. |
|
Looks like StaticArraysCore.jl may need an upate, so that tests pass also on nightly? |
maleadt
changed the title
|
It's strange that symmetric/hermitian tests fail that have not been touched here. Also, the function calls before the tests call basic routines directly, not through |
|
This also fails outside of this PR, so let's just mark 1.10 as broken for now. |
Co-authored-by: Tim Besard <tim.besard@gmail.com>
This is to make use of JuliaLang/julia#50058. While that PR has not been merged, I'm putting this up to see whether this is working properly on older julia versions, and then we check nightly later.