add atomic operators for globals, memory, and setonce#52868
add atomic operators for globals, memory, and setonce#52868IanButterworth merged 2 commits intomasterfrom
Conversation
vtjnash
force-pushed
the
jn/moreatomics
branch
3 times, most recently
from
4456be2 to
cdd101e
Compare
vtjnash
force-pushed
the
jn/moreatomics
branch
from
cdd101e to
48bc10d
Compare
|
32-bit failing with but everything else looks ready for review |
| ## Conversions ## | ||
|
|
||
| convert(::Type{T}, a::AbstractArray) where {T<:GenericMemory} = a isa T ? a : T(a)::T | ||
| convert(::Type{T}, a::AbstractArray) where {T<:Memory} = a isa T ? a : T(a)::T |
There was a problem hiding this comment.
should the second argument here be AbstracVector?
There was a problem hiding this comment.
apparently not?
julia> @which convert(Vector{Int}, [1 2; 3 4])
convert(::Type{T}, a::AbstractArray) where T<:Array
@ Base array.jl:607
julia> convert(Vector{Int}, [1 2; 3 4])
ERROR: MethodError: no method matching Vector{Int64}(::Matrix{Int64})
48bc10d to
a872b19
Compare
|
I will merge tomorrow, and take any post-merge feedback instead then, if not gotten to before then. |
|
Do you have a list of all the new and deleted intrinsics/builtins? It's a bit hard to glean from the diff, since you also did some (appreciated) clean-up on things. |
| @atomiconce :sequentially_consistent :monotonic a.b.x = value | ||
|
|
||
| Perform the conditional assignment of value atomically if it was previously | ||
| unset, returning the value `success::Bool`. Where `success` indicates whether |
There was a problem hiding this comment.
So one thing is that unset isn't the nomenclature we seem to usually use, we mostly seem to use defined/undefined
There was a problem hiding this comment.
I guess that should say the field was not set or the field was undef?
There was a problem hiding this comment.
I think that might be a bit more inline to what we call it in other places?
There was a problem hiding this comment.
We use unset at lot for this also (e.g. unsetindex), since we use undefined to mean isbits types that are not initialized, while this is not undef->defined but rather unset->set
|
Here's the list of new functions, from the docs (though your comment made me realize I had missed adding the rest of the memoryref builtins there): Core.modifyglobal!
Core.swapglobal!
Core.replaceglobal!
Core.setglobalonce!
Core.memoryrefswap!
Core.memoryrefmodify!
Core.memoryrefreplace!
Core.memoryrefsetonce!
Core.setfieldonce! |
|
|
||
| """ | ||
| setfieldonce!(value, name::Union{Int,Symbol}, desired, | ||
| [success_order::Symbol, [fail_order::Symbol=success_order]) -> success::Bool |
There was a problem hiding this comment.
What is the default success_order? (Same for the other docstrings here)
There was a problem hiding this comment.
The default is not to have a success_order (in some places, equivalent to :not_atomic, but not identical)
| setfield!(value, name, desired, success_order) | ||
| end | ||
| return ok | ||
| """ |
There was a problem hiding this comment.
Probably should also get a compat note?
There was a problem hiding this comment.
Hopefully I got all those versions right
Adds an AtomicMemory{T} type (GenericMemory{:atomic,T,Core.CPU}) for
doing element-wise atomic operations on an array (with a lock
per-element if required).
Adds the full complement of atomic operations for global variables.
Also add a *setonce set of functions for atomically setting a value, but
only if it is unset. The complementary operation (Atomic `_unsetindex!`
for AtomicMemory) is not yet provided.
vtjnash
force-pushed
the
jn/moreatomics
branch
from
a872b19 to
d8d17b8
Compare
IanButterworth
removed
the
merge me
6 participants
This implements
AtomicMemory, atomic operations for globals, and the class of atomic function for setting a field or global or memory exactly once (setting undef => value). It is quite similar to an@atomicreplace, but where there was not a previous value. This is not needed for pointers, since it is simply calling C_NULL => pointer in that case, since there is no "undef" value that throws after load.plenty of future work still, for a later PR:
@atomic global x = yand@atomic closedover = yCore.AtomicBoxfor closure captureatomicsetindex& friends@atomic x = yand@atomic x[] = y, etc@atomiconce f()