-
-
Notifications
You must be signed in to change notification settings - Fork 5.5k
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.
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
abstract multiple inheritance #5
Comments
I'm not sure this blocks version 1.0. I agree we want it, but realistically it will take a bunch of time to settle and we know julia is perfectly usable without it. |
Is it that you think this will be very tricky to implement (understandably), or that you think the "no incest" rule will not be sufficient to make multiple inheritance sane? I would be willing to take a crack at this, but I'm afraid it may be beyond me. |
The first one. Everything in the type system is pretty fragile, hard to modify without unintended consequences. |
Ok, that's fair. I think I won't even attempt this then. Too hard and mucking about with the type system is definitely your expertise rather than mine. |
Maybe the no-incest rule in the initial suggestion can be weakened in a way similar to the following:
|
It seems to me that the no-incest rule presupposes that all descendents of
and then a user comes along and defines
It seems a bit harsh that this should invalidate the function
|
Another thought about multiple inheritance (which I am for) is that it seems like it could make things a lot trickier for type inference; the intersection of two abstract types would never be empty. If there were some kind of mechanism to specify that two types are disjoint (e.g., can not share a common descendent), that might help. |
The thinking here was that the |
@StefanKarpinski: That's very reasonable. My point was just that you cannot know at the time of the method definition for |
What about something like dominant and recessive alleles on biology http://en.wikipedia.org/wiki/Dominance_%28genetics%29 Indicate one abstract (parent) to have dominant alleles (methods) over the other (maybe the first on the list), and in case of redundancy choose the dominant. |
Hi there. I'm a new Julia user and multiple inheritance is important to me. Just adding my +1. |
Also, @diegozea's suggestion is basically what is done by Python's method resolution order (mro), which I have found very sensible. There is always the possibility of having multiple method name conflicts and wanting a different resolution for each one. In python, that situation is handled by the super method. |
In Gtk.jl (JuliaGraphics/Gtk.jl#20) there is also a need for inheriting from multiple interfaces. This might be faked by using type Unions but this has the drawback that the Union type cannot be extended. Therefore the question: Might it be possible to allow to extend Union types? Or is this almost the same as multiple inheritance and therefore the same effort to implement? @JeffBezanson, @StefanKarpinski |
This is part of the work to address #51352 by attempting to allow the compiler to perform SRAO on persistent data structures like `PersistentDict` as if they were regular immutable data structures. These sorts of data structures have very complicated internals (with lots of mutation, memory sharing, etc.), but a relatively simple interface. As such, it is unlikely that our compiler will have sufficient power to optimize this interface by analyzing the implementation. We thus need to come up with some other mechanism that gives the compiler license to perform the requisite optimization. One way would be to just hardcode `PersistentDict` into the compiler, optimizing it like any of the other builtin datatypes. However, this is of course very unsatisfying. At the other end of the spectrum would be something like a generic rewrite rule system (e-graphs anyone?) that would let the PersistentDict implementation declare its interface to the compiler and the compiler would use this for optimization (in a perfect world, the actual rewrite would then be checked using some sort of formal methods). I think that would be interesting, but we're very far from even being able to design something like that (at least in Base - experiments with external AbstractInterpreters in this direction are encouraged). This PR tries to come up with a reasonable middle ground, where the compiler gets some knowledge of the protocol hardcoded without having to know about the implementation details of the data structure. The basic ideas is that `Core` provides some magic generic functions that implementations can extend. Semantically, they are not special. They dispatch as usual, and implementations are expected to work properly even in the absence of any compiler optimizations. However, the compiler is semantically permitted to perform structural optimization using these magic generic functions. In the concrete case, this PR introduces the `KeyValue` interface which consists of two generic functions, `get` and `set`. The core optimization is that the compiler is allowed to rewrite any occurrence of `get(set(x, k, v), k)` into `v` without additional legality checks. In particular, the compiler performs no type checks, conversions, etc. The higher level implementation code is expected to do all that. This approach closely matches the general direction we've been taking in external AbstractInterpreters for embedding additional semantics and optimization opportunities into Julia code (although we generally use methods there, rather than full generic functions), so I think we have some evidence that this sort of approach works reasonably well. Nevertheless, this is certainly an experiment and the interface is explicitly declared unstable. This is fully working and implemented, but the optimization currently bails on anything but the simplest cases. Filling all those cases in is not particularly hard, but should be done along with a more invasive refactoring of SROA, so we should figure out the general direction here first and then we can finish all that up in a follow-up cleanup. Before: ``` julia> using BenchmarkTools julia> function foo() a = Base.PersistentDict(:a => 1) return a[:a] end foo (generic function with 1 method) julia> @benchmark foo() BenchmarkTools.Trial: 10000 samples with 993 evaluations. Range (min … max): 32.940 ns … 28.754 μs ┊ GC (min … max): 0.00% … 99.76% Time (median): 49.647 ns ┊ GC (median): 0.00% Time (mean ± σ): 57.519 ns ± 333.275 ns ┊ GC (mean ± σ): 10.81% ± 2.22% ▃█▅ ▁▃▅▅▃▁ ▁▃▂ ▂ ▁▂▄▃▅▇███▇▃▁▂▁▁▁▁▁▁▁▁▂▂▅██████▅▂▁▁▁▁▁▁▁▁▁▁▂▃▃▇███▇▆███▆▄▃▃▂▂ ▃ 32.9 ns Histogram: frequency by time 68.6 ns < Memory estimate: 128 bytes, allocs estimate: 4. julia> @code_typed foo() CodeInfo( 1 ─ %1 = invoke Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}}(Base.HashArrayMappedTries.undef::UndefInitializer, 1::Int64)::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %2 = %new(Base.HashArrayMappedTries.HAMT{Symbol, Int64}, %1, 0x00000000)::Base.HashArrayMappedTries.HAMT{Symbol, Int64} │ %3 = %new(Base.HashArrayMappedTries.Leaf{Symbol, Int64}, :a, 1)::Base.HashArrayMappedTries.Leaf{Symbol, Int64} │ %4 = Base.getfield(%2, :data)::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %5 = $(Expr(:boundscheck, true))::Bool └── goto #5 if not %5 2 ─ %7 = Base.sub_int(1, 1)::Int64 │ %8 = Base.bitcast(UInt64, %7)::UInt64 │ %9 = Base.getfield(%4, :size)::Tuple{Int64} │ %10 = $(Expr(:boundscheck, true))::Bool │ %11 = Base.getfield(%9, 1, %10)::Int64 │ %12 = Base.bitcast(UInt64, %11)::UInt64 │ %13 = Base.ult_int(%8, %12)::Bool └── goto #4 if not %13 3 ─ goto #5 4 ─ %16 = Core.tuple(1)::Tuple{Int64} │ invoke Base.throw_boundserror(%4::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}}, %16::Tuple{Int64})::Union{} └── unreachable 5 ┄ %19 = Base.getfield(%4, :ref)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %20 = Base.memoryref(%19, 1, false)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ Base.memoryrefset!(%20, %3, :not_atomic, false)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} └── goto #6 6 ─ %23 = Base.getfield(%2, :bitmap)::UInt32 │ %24 = Base.or_int(%23, 0x00010000)::UInt32 │ Base.setfield!(%2, :bitmap, %24)::UInt32 └── goto #7 7 ─ %27 = %new(Base.PersistentDict{Symbol, Int64}, %2)::Base.PersistentDict{Symbol, Int64} └── goto #8 8 ─ %29 = invoke Base.getindex(%27::Base.PersistentDict{Symbol, Int64},🅰️ :Symbol)::Int64 └── return %29 ``` After: ``` julia> using BenchmarkTools julia> function foo() a = Base.PersistentDict(:a => 1) return a[:a] end foo (generic function with 1 method) julia> @benchmark foo() BenchmarkTools.Trial: 10000 samples with 1000 evaluations. Range (min … max): 2.459 ns … 11.320 ns ┊ GC (min … max): 0.00% … 0.00% Time (median): 2.460 ns ┊ GC (median): 0.00% Time (mean ± σ): 2.469 ns ± 0.183 ns ┊ GC (mean ± σ): 0.00% ± 0.00% ▂ █ ▁ █ ▂ █▁▁▁▁█▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█▁▁▁▁█ █ 2.46 ns Histogram: log(frequency) by time 2.47 ns < Memory estimate: 0 bytes, allocs estimate: 0. julia> @code_typed foo() CodeInfo( 1 ─ return 1 ```
This is part of the work to address #51352 by attempting to allow the compiler to perform SRAO on persistent data structures like `PersistentDict` as if they were regular immutable data structures. These sorts of data structures have very complicated internals (with lots of mutation, memory sharing, etc.), but a relatively simple interface. As such, it is unlikely that our compiler will have sufficient power to optimize this interface by analyzing the implementation. We thus need to come up with some other mechanism that gives the compiler license to perform the requisite optimization. One way would be to just hardcode `PersistentDict` into the compiler, optimizing it like any of the other builtin datatypes. However, this is of course very unsatisfying. At the other end of the spectrum would be something like a generic rewrite rule system (e-graphs anyone?) that would let the PersistentDict implementation declare its interface to the compiler and the compiler would use this for optimization (in a perfect world, the actual rewrite would then be checked using some sort of formal methods). I think that would be interesting, but we're very far from even being able to design something like that (at least in Base - experiments with external AbstractInterpreters in this direction are encouraged). This PR tries to come up with a reasonable middle ground, where the compiler gets some knowledge of the protocol hardcoded without having to know about the implementation details of the data structure. The basic ideas is that `Core` provides some magic generic functions that implementations can extend. Semantically, they are not special. They dispatch as usual, and implementations are expected to work properly even in the absence of any compiler optimizations. However, the compiler is semantically permitted to perform structural optimization using these magic generic functions. In the concrete case, this PR introduces the `KeyValue` interface which consists of two generic functions, `get` and `set`. The core optimization is that the compiler is allowed to rewrite any occurrence of `get(set(x, k, v), k)` into `v` without additional legality checks. In particular, the compiler performs no type checks, conversions, etc. The higher level implementation code is expected to do all that. This approach closely matches the general direction we've been taking in external AbstractInterpreters for embedding additional semantics and optimization opportunities into Julia code (although we generally use methods there, rather than full generic functions), so I think we have some evidence that this sort of approach works reasonably well. Nevertheless, this is certainly an experiment and the interface is explicitly declared unstable. This is fully working and implemented, but the optimization currently bails on anything but the simplest cases. Filling all those cases in is not particularly hard, but should be done along with a more invasive refactoring of SROA, so we should figure out the general direction here first and then we can finish all that up in a follow-up cleanup. Before: ``` julia> using BenchmarkTools julia> function foo() a = Base.PersistentDict(:a => 1) return a[:a] end foo (generic function with 1 method) julia> @benchmark foo() BenchmarkTools.Trial: 10000 samples with 993 evaluations. Range (min … max): 32.940 ns … 28.754 μs ┊ GC (min … max): 0.00% … 99.76% Time (median): 49.647 ns ┊ GC (median): 0.00% Time (mean ± σ): 57.519 ns ± 333.275 ns ┊ GC (mean ± σ): 10.81% ± 2.22% ▃█▅ ▁▃▅▅▃▁ ▁▃▂ ▂ ▁▂▄▃▅▇███▇▃▁▂▁▁▁▁▁▁▁▁▂▂▅██████▅▂▁▁▁▁▁▁▁▁▁▁▂▃▃▇███▇▆███▆▄▃▃▂▂ ▃ 32.9 ns Histogram: frequency by time 68.6 ns < Memory estimate: 128 bytes, allocs estimate: 4. julia> @code_typed foo() CodeInfo( 1 ─ %1 = invoke Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}}(Base.HashArrayMappedTries.undef::UndefInitializer, 1::Int64)::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %2 = %new(Base.HashArrayMappedTries.HAMT{Symbol, Int64}, %1, 0x00000000)::Base.HashArrayMappedTries.HAMT{Symbol, Int64} │ %3 = %new(Base.HashArrayMappedTries.Leaf{Symbol, Int64}, :a, 1)::Base.HashArrayMappedTries.Leaf{Symbol, Int64} │ %4 = Base.getfield(%2, :data)::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %5 = $(Expr(:boundscheck, true))::Bool └── goto #5 if not %5 2 ─ %7 = Base.sub_int(1, 1)::Int64 │ %8 = Base.bitcast(UInt64, %7)::UInt64 │ %9 = Base.getfield(%4, :size)::Tuple{Int64} │ %10 = $(Expr(:boundscheck, true))::Bool │ %11 = Base.getfield(%9, 1, %10)::Int64 │ %12 = Base.bitcast(UInt64, %11)::UInt64 │ %13 = Base.ult_int(%8, %12)::Bool └── goto #4 if not %13 3 ─ goto #5 4 ─ %16 = Core.tuple(1)::Tuple{Int64} │ invoke Base.throw_boundserror(%4::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}}, %16::Tuple{Int64})::Union{} └── unreachable 5 ┄ %19 = Base.getfield(%4, :ref)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %20 = Base.memoryref(%19, 1, false)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ Base.memoryrefset!(%20, %3, :not_atomic, false)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} └── goto #6 6 ─ %23 = Base.getfield(%2, :bitmap)::UInt32 │ %24 = Base.or_int(%23, 0x00010000)::UInt32 │ Base.setfield!(%2, :bitmap, %24)::UInt32 └── goto #7 7 ─ %27 = %new(Base.PersistentDict{Symbol, Int64}, %2)::Base.PersistentDict{Symbol, Int64} └── goto #8 8 ─ %29 = invoke Base.getindex(%27::Base.PersistentDict{Symbol, Int64},🅰️ :Symbol)::Int64 └── return %29 ``` After: ``` julia> using BenchmarkTools julia> function foo() a = Base.PersistentDict(:a => 1) return a[:a] end foo (generic function with 1 method) julia> @benchmark foo() BenchmarkTools.Trial: 10000 samples with 1000 evaluations. Range (min … max): 2.459 ns … 11.320 ns ┊ GC (min … max): 0.00% … 0.00% Time (median): 2.460 ns ┊ GC (median): 0.00% Time (mean ± σ): 2.469 ns ± 0.183 ns ┊ GC (mean ± σ): 0.00% ± 0.00% ▂ █ ▁ █ ▂ █▁▁▁▁█▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█▁▁▁▁█ █ 2.46 ns Histogram: log(frequency) by time 2.47 ns < Memory estimate: 0 bytes, allocs estimate: 0. julia> @code_typed foo() CodeInfo( 1 ─ return 1 ```
This is part of the work to address #51352 by attempting to allow the compiler to perform SRAO on persistent data structures like `PersistentDict` as if they were regular immutable data structures. These sorts of data structures have very complicated internals (with lots of mutation, memory sharing, etc.), but a relatively simple interface. As such, it is unlikely that our compiler will have sufficient power to optimize this interface by analyzing the implementation. We thus need to come up with some other mechanism that gives the compiler license to perform the requisite optimization. One way would be to just hardcode `PersistentDict` into the compiler, optimizing it like any of the other builtin datatypes. However, this is of course very unsatisfying. At the other end of the spectrum would be something like a generic rewrite rule system (e-graphs anyone?) that would let the PersistentDict implementation declare its interface to the compiler and the compiler would use this for optimization (in a perfect world, the actual rewrite would then be checked using some sort of formal methods). I think that would be interesting, but we're very far from even being able to design something like that (at least in Base - experiments with external AbstractInterpreters in this direction are encouraged). This PR tries to come up with a reasonable middle ground, where the compiler gets some knowledge of the protocol hardcoded without having to know about the implementation details of the data structure. The basic ideas is that `Core` provides some magic generic functions that implementations can extend. Semantically, they are not special. They dispatch as usual, and implementations are expected to work properly even in the absence of any compiler optimizations. However, the compiler is semantically permitted to perform structural optimization using these magic generic functions. In the concrete case, this PR introduces the `KeyValue` interface which consists of two generic functions, `get` and `set`. The core optimization is that the compiler is allowed to rewrite any occurrence of `get(set(x, k, v), k)` into `v` without additional legality checks. In particular, the compiler performs no type checks, conversions, etc. The higher level implementation code is expected to do all that. This approach closely matches the general direction we've been taking in external AbstractInterpreters for embedding additional semantics and optimization opportunities into Julia code (although we generally use methods there, rather than full generic functions), so I think we have some evidence that this sort of approach works reasonably well. Nevertheless, this is certainly an experiment and the interface is explicitly declared unstable. ## Current Status This is fully working and implemented, but the optimization currently bails on anything but the simplest cases. Filling all those cases in is not particularly hard, but should be done along with a more invasive refactoring of SROA, so we should figure out the general direction here first and then we can finish all that up in a follow-up cleanup. ## Obligatory benchmark Before: ``` julia> using BenchmarkTools julia> function foo() a = Base.PersistentDict(:a => 1) return a[:a] end foo (generic function with 1 method) julia> @benchmark foo() BenchmarkTools.Trial: 10000 samples with 993 evaluations. Range (min … max): 32.940 ns … 28.754 μs ┊ GC (min … max): 0.00% … 99.76% Time (median): 49.647 ns ┊ GC (median): 0.00% Time (mean ± σ): 57.519 ns ± 333.275 ns ┊ GC (mean ± σ): 10.81% ± 2.22% ▃█▅ ▁▃▅▅▃▁ ▁▃▂ ▂ ▁▂▄▃▅▇███▇▃▁▂▁▁▁▁▁▁▁▁▂▂▅██████▅▂▁▁▁▁▁▁▁▁▁▁▂▃▃▇███▇▆███▆▄▃▃▂▂ ▃ 32.9 ns Histogram: frequency by time 68.6 ns < Memory estimate: 128 bytes, allocs estimate: 4. julia> @code_typed foo() CodeInfo( 1 ─ %1 = invoke Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}}(Base.HashArrayMappedTries.undef::UndefInitializer, 1::Int64)::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %2 = %new(Base.HashArrayMappedTries.HAMT{Symbol, Int64}, %1, 0x00000000)::Base.HashArrayMappedTries.HAMT{Symbol, Int64} │ %3 = %new(Base.HashArrayMappedTries.Leaf{Symbol, Int64}, :a, 1)::Base.HashArrayMappedTries.Leaf{Symbol, Int64} │ %4 = Base.getfield(%2, :data)::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %5 = $(Expr(:boundscheck, true))::Bool └── goto #5 if not %5 2 ─ %7 = Base.sub_int(1, 1)::Int64 │ %8 = Base.bitcast(UInt64, %7)::UInt64 │ %9 = Base.getfield(%4, :size)::Tuple{Int64} │ %10 = $(Expr(:boundscheck, true))::Bool │ %11 = Base.getfield(%9, 1, %10)::Int64 │ %12 = Base.bitcast(UInt64, %11)::UInt64 │ %13 = Base.ult_int(%8, %12)::Bool └── goto #4 if not %13 3 ─ goto #5 4 ─ %16 = Core.tuple(1)::Tuple{Int64} │ invoke Base.throw_boundserror(%4::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}}, %16::Tuple{Int64})::Union{} └── unreachable 5 ┄ %19 = Base.getfield(%4, :ref)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %20 = Base.memoryref(%19, 1, false)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ Base.memoryrefset!(%20, %3, :not_atomic, false)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} └── goto #6 6 ─ %23 = Base.getfield(%2, :bitmap)::UInt32 │ %24 = Base.or_int(%23, 0x00010000)::UInt32 │ Base.setfield!(%2, :bitmap, %24)::UInt32 └── goto #7 7 ─ %27 = %new(Base.PersistentDict{Symbol, Int64}, %2)::Base.PersistentDict{Symbol, Int64} └── goto #8 8 ─ %29 = invoke Base.getindex(%27::Base.PersistentDict{Symbol, Int64},🅰️ :Symbol)::Int64 └── return %29 ``` After: ``` julia> using BenchmarkTools julia> function foo() a = Base.PersistentDict(:a => 1) return a[:a] end foo (generic function with 1 method) julia> @benchmark foo() BenchmarkTools.Trial: 10000 samples with 1000 evaluations. Range (min … max): 2.459 ns … 11.320 ns ┊ GC (min … max): 0.00% … 0.00% Time (median): 2.460 ns ┊ GC (median): 0.00% Time (mean ± σ): 2.469 ns ± 0.183 ns ┊ GC (mean ± σ): 0.00% ± 0.00% ▂ █ ▁ █ ▂ █▁▁▁▁█▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█▁▁▁▁█ █ 2.46 ns Histogram: log(frequency) by time 2.47 ns < Memory estimate: 0 bytes, allocs estimate: 0. julia> @code_typed foo() CodeInfo( 1 ─ return 1 ```
This is part of the work to address JuliaLang#51352 by attempting to allow the compiler to perform SRAO on persistent data structures like `PersistentDict` as if they were regular immutable data structures. These sorts of data structures have very complicated internals (with lots of mutation, memory sharing, etc.), but a relatively simple interface. As such, it is unlikely that our compiler will have sufficient power to optimize this interface by analyzing the implementation. We thus need to come up with some other mechanism that gives the compiler license to perform the requisite optimization. One way would be to just hardcode `PersistentDict` into the compiler, optimizing it like any of the other builtin datatypes. However, this is of course very unsatisfying. At the other end of the spectrum would be something like a generic rewrite rule system (e-graphs anyone?) that would let the PersistentDict implementation declare its interface to the compiler and the compiler would use this for optimization (in a perfect world, the actual rewrite would then be checked using some sort of formal methods). I think that would be interesting, but we're very far from even being able to design something like that (at least in Base - experiments with external AbstractInterpreters in this direction are encouraged). This PR tries to come up with a reasonable middle ground, where the compiler gets some knowledge of the protocol hardcoded without having to know about the implementation details of the data structure. The basic ideas is that `Core` provides some magic generic functions that implementations can extend. Semantically, they are not special. They dispatch as usual, and implementations are expected to work properly even in the absence of any compiler optimizations. However, the compiler is semantically permitted to perform structural optimization using these magic generic functions. In the concrete case, this PR introduces the `KeyValue` interface which consists of two generic functions, `get` and `set`. The core optimization is that the compiler is allowed to rewrite any occurrence of `get(set(x, k, v), k)` into `v` without additional legality checks. In particular, the compiler performs no type checks, conversions, etc. The higher level implementation code is expected to do all that. This approach closely matches the general direction we've been taking in external AbstractInterpreters for embedding additional semantics and optimization opportunities into Julia code (although we generally use methods there, rather than full generic functions), so I think we have some evidence that this sort of approach works reasonably well. Nevertheless, this is certainly an experiment and the interface is explicitly declared unstable. ## Current Status This is fully working and implemented, but the optimization currently bails on anything but the simplest cases. Filling all those cases in is not particularly hard, but should be done along with a more invasive refactoring of SROA, so we should figure out the general direction here first and then we can finish all that up in a follow-up cleanup. ## Obligatory benchmark Before: ``` julia> using BenchmarkTools julia> function foo() a = Base.PersistentDict(:a => 1) return a[:a] end foo (generic function with 1 method) julia> @benchmark foo() BenchmarkTools.Trial: 10000 samples with 993 evaluations. Range (min … max): 32.940 ns … 28.754 μs ┊ GC (min … max): 0.00% … 99.76% Time (median): 49.647 ns ┊ GC (median): 0.00% Time (mean ± σ): 57.519 ns ± 333.275 ns ┊ GC (mean ± σ): 10.81% ± 2.22% ▃█▅ ▁▃▅▅▃▁ ▁▃▂ ▂ ▁▂▄▃▅▇███▇▃▁▂▁▁▁▁▁▁▁▁▂▂▅██████▅▂▁▁▁▁▁▁▁▁▁▁▂▃▃▇███▇▆███▆▄▃▃▂▂ ▃ 32.9 ns Histogram: frequency by time 68.6 ns < Memory estimate: 128 bytes, allocs estimate: 4. julia> @code_typed foo() CodeInfo( 1 ─ %1 = invoke Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}}(Base.HashArrayMappedTries.undef::UndefInitializer, 1::Int64)::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %2 = %new(Base.HashArrayMappedTries.HAMT{Symbol, Int64}, %1, 0x00000000)::Base.HashArrayMappedTries.HAMT{Symbol, Int64} │ %3 = %new(Base.HashArrayMappedTries.Leaf{Symbol, Int64}, :a, 1)::Base.HashArrayMappedTries.Leaf{Symbol, Int64} │ %4 = Base.getfield(%2, :data)::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %5 = $(Expr(:boundscheck, true))::Bool └── goto JuliaLang#5 if not %5 2 ─ %7 = Base.sub_int(1, 1)::Int64 │ %8 = Base.bitcast(UInt64, %7)::UInt64 │ %9 = Base.getfield(%4, :size)::Tuple{Int64} │ %10 = $(Expr(:boundscheck, true))::Bool │ %11 = Base.getfield(%9, 1, %10)::Int64 │ %12 = Base.bitcast(UInt64, %11)::UInt64 │ %13 = Base.ult_int(%8, %12)::Bool └── goto JuliaLang#4 if not %13 3 ─ goto JuliaLang#5 4 ─ %16 = Core.tuple(1)::Tuple{Int64} │ invoke Base.throw_boundserror(%4::Vector{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}}, %16::Tuple{Int64})::Union{} └── unreachable 5 ┄ %19 = Base.getfield(%4, :ref)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ %20 = Base.memoryref(%19, 1, false)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} │ Base.memoryrefset!(%20, %3, :not_atomic, false)::MemoryRef{Union{Base.HashArrayMappedTries.HAMT{Symbol, Int64}, Base.HashArrayMappedTries.Leaf{Symbol, Int64}}} └── goto JuliaLang#6 6 ─ %23 = Base.getfield(%2, :bitmap)::UInt32 │ %24 = Base.or_int(%23, 0x00010000)::UInt32 │ Base.setfield!(%2, :bitmap, %24)::UInt32 └── goto JuliaLang#7 7 ─ %27 = %new(Base.PersistentDict{Symbol, Int64}, %2)::Base.PersistentDict{Symbol, Int64} └── goto JuliaLang#8 8 ─ %29 = invoke Base.getindex(%27::Base.PersistentDict{Symbol, Int64},🅰️ :Symbol)::Int64 └── return %29 ``` After: ``` julia> using BenchmarkTools julia> function foo() a = Base.PersistentDict(:a => 1) return a[:a] end foo (generic function with 1 method) julia> @benchmark foo() BenchmarkTools.Trial: 10000 samples with 1000 evaluations. Range (min … max): 2.459 ns … 11.320 ns ┊ GC (min … max): 0.00% … 0.00% Time (median): 2.460 ns ┊ GC (median): 0.00% Time (mean ± σ): 2.469 ns ± 0.183 ns ┊ GC (mean ± σ): 0.00% ± 0.00% ▂ █ ▁ █ ▂ █▁▁▁▁█▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█▁▁▁▁█ █ 2.46 ns Histogram: log(frequency) by time 2.47 ns < Memory estimate: 0 bytes, allocs estimate: 0. julia> @code_typed foo() CodeInfo( 1 ─ return 1 ```
The C4 linearization algorithm I recently added to Gerbil Scheme combines multiple inheritance of classes/traits and single inheritance of structs/classes, so you can have the best of both worlds:
C4 extends the well-known C3 algorithm (for multiple inheritance of classes/traits) with an additional constraint to support structs/classes. https://github.com/mighty-gerbils/gerbil/blob/master/src/gerbil/runtime/c3.ss NB: Common Lisp, and the earlier tradition of multiple inheritance, calls "classes" the things with multiple inheritance and "structs" the things with single inheritance only. Smalltalk and after it Java, and the earlier and more prevalent tradition of single inheritance, calls "classes" the things with single inheritance and "traits" the things with multiple inheritance. Wonderful nomenclature, right? You're welcome! |
@fare well that's all kinds of groovy |
Followup to #53833 Fixes a failure seen in #53974 (below) I believe this is the more correct check to make? The heapsnapshot generated from this PR is viewable in vscode. ``` 2024-04-06 09:33:58 EDT From worker 7: ERROR: Base.InvalidCharError{Char}('\xc1\xae') 2024-04-06 09:33:58 EDT From worker 7: Stacktrace: 2024-04-06 09:33:58 EDT From worker 7: [1] throw_invalid_char(c::Char) 2024-04-06 09:33:58 EDT From worker 7: @ Base ./char.jl:86 2024-04-06 09:33:58 EDT From worker 7: [2] UInt32 2024-04-06 09:33:58 EDT From worker 7: @ ./char.jl:133 [inlined] 2024-04-06 09:33:58 EDT From worker 7: [3] category_code 2024-04-06 09:33:58 EDT From worker 7: @ ./strings/unicode.jl:339 [inlined] 2024-04-06 09:33:58 EDT From worker 7: [4] isassigned 2024-04-06 09:33:58 EDT From worker 7: @ ./strings/unicode.jl:355 [inlined] 2024-04-06 09:33:58 EDT From worker 7: [5] isassigned 2024-04-06 09:33:58 EDT From worker 7: @ /cache/build/tester-amdci5-14/julialang/julia-master/julia-41d026beaf/share/julia/stdlib/v1.12/Unicode/src/Unicode.jl:138 [inlined] 2024-04-06 09:33:58 EDT From worker 7: [6] print_str_escape_json(stream::IOStream, s::String) 2024-04-06 09:33:58 EDT From worker 7: @ Profile.HeapSnapshot /cache/build/tester-amdci5-14/julialang/julia-master/julia-41d026beaf/share/julia/stdlib/v1.12/Profile/src/heapsnapshot_reassemble.jl:239 2024-04-06 09:33:59 EDT From worker 7: [7] (::Profile.HeapSnapshot.var"#5#6"{IOStream})(strings_io::IOStream) 2024-04-06 09:33:59 EDT From worker 7: @ Profile.HeapSnapshot /cache/build/tester-amdci5-14/julialang/julia-master/julia-41d026beaf/share/julia/stdlib/v1.12/Profile/src/heapsnapshot_reassemble.jl:192 ```
Followup to #53833 Fixes a failure seen in #53974 (below) I believe this is the more correct check to make? The heapsnapshot generated from this PR is viewable in vscode. ``` 2024-04-06 09:33:58 EDT From worker 7: ERROR: Base.InvalidCharError{Char}('\xc1\xae') 2024-04-06 09:33:58 EDT From worker 7: Stacktrace: 2024-04-06 09:33:58 EDT From worker 7: [1] throw_invalid_char(c::Char) 2024-04-06 09:33:58 EDT From worker 7: @ Base ./char.jl:86 2024-04-06 09:33:58 EDT From worker 7: [2] UInt32 2024-04-06 09:33:58 EDT From worker 7: @ ./char.jl:133 [inlined] 2024-04-06 09:33:58 EDT From worker 7: [3] category_code 2024-04-06 09:33:58 EDT From worker 7: @ ./strings/unicode.jl:339 [inlined] 2024-04-06 09:33:58 EDT From worker 7: [4] isassigned 2024-04-06 09:33:58 EDT From worker 7: @ ./strings/unicode.jl:355 [inlined] 2024-04-06 09:33:58 EDT From worker 7: [5] isassigned 2024-04-06 09:33:58 EDT From worker 7: @ /cache/build/tester-amdci5-14/julialang/julia-master/julia-41d026beaf/share/julia/stdlib/v1.12/Unicode/src/Unicode.jl:138 [inlined] 2024-04-06 09:33:58 EDT From worker 7: [6] print_str_escape_json(stream::IOStream, s::String) 2024-04-06 09:33:58 EDT From worker 7: @ Profile.HeapSnapshot /cache/build/tester-amdci5-14/julialang/julia-master/julia-41d026beaf/share/julia/stdlib/v1.12/Profile/src/heapsnapshot_reassemble.jl:239 2024-04-06 09:33:59 EDT From worker 7: [7] (::Profile.HeapSnapshot.var"#5#6"{IOStream})(strings_io::IOStream) 2024-04-06 09:33:59 EDT From worker 7: @ Profile.HeapSnapshot /cache/build/tester-amdci5-14/julialang/julia-master/julia-41d026beaf/share/julia/stdlib/v1.12/Profile/src/heapsnapshot_reassemble.jl:192 ``` (cherry picked from commit c557636)
The former also handles vectors of pointers, which can occur after vectorization: ``` #5 0x00007f5bfe94de5e in llvm::cast<llvm::PointerType, llvm::Type> (Val=<optimized out>) at llvm/Support/Casting.h:578 578 assert(isa<To>(Val) && "cast<Ty>() argument of incompatible type!"); (rr) up #6 GCInvariantVerifier::visitAddrSpaceCastInst (this=this@entry=0x7ffd022fbf56, I=...) at julia/src/llvm-gc-invariant-verifier.cpp:66 66 unsigned ToAS = cast<PointerType>(I.getDestTy())->getAddressSpace(); (rr) call I.dump() %23 = addrspacecast <4 x ptr addrspace(10)> %wide.load to <4 x ptr addrspace(11)>, !dbg !43 ```
The former also handles vectors of pointers, which can occur after vectorization: ``` #5 0x00007f5bfe94de5e in llvm::cast<llvm::PointerType, llvm::Type> (Val=<optimized out>) at llvm/Support/Casting.h:578 578 assert(isa<To>(Val) && "cast<Ty>() argument of incompatible type!"); (rr) up #6 GCInvariantVerifier::visitAddrSpaceCastInst (this=this@entry=0x7ffd022fbf56, I=...) at julia/src/llvm-gc-invariant-verifier.cpp:66 66 unsigned ToAS = cast<PointerType>(I.getDestTy())->getAddressSpace(); (rr) call I.dump() %23 = addrspacecast <4 x ptr addrspace(10)> %wide.load to <4 x ptr addrspace(11)>, !dbg !43 ```
…ce. (#54113) The former also handles vectors of pointers, which can occur after vectorization: ``` #5 0x00007f5bfe94de5e in llvm::cast<llvm::PointerType, llvm::Type> (Val=<optimized out>) at llvm/Support/Casting.h:578 578 assert(isa<To>(Val) && "cast<Ty>() argument of incompatible type!"); (rr) up #6 GCInvariantVerifier::visitAddrSpaceCastInst (this=this@entry=0x7ffd022fbf56, I=...) at julia/src/llvm-gc-invariant-verifier.cpp:66 66 unsigned ToAS = cast<PointerType>(I.getDestTy())->getAddressSpace(); (rr) call I.dump() %23 = addrspacecast <4 x ptr addrspace(10)> %wide.load to <4 x ptr addrspace(11)>, !dbg !43 ``` Fixes aborts seen in #53070
As an application of #55545, this commit avoids the insertion of `:throw_undef_if_not` nodes when the defined-ness of a slot is guaranteed by abstract interpretation. ```julia julia> function isdefined_nothrow(c, x) local val if c val = x end if @isdefined val return val end return zero(Int) end; julia> @code_typed isdefined_nothrow(true, 42) ``` ```diff diff --git a/old b/new index c4980a5c9c..3d1d6d30f0 100644 --- a/old +++ b/new @@ -4,7 +4,6 @@ CodeInfo( 3 ┄ %3 = φ (#2 => x, #1 => #undef)::Int64 │ %4 = φ (#2 => true, #1 => false)::Bool └── goto #5 if not %4 -4 ─ $(Expr(:throw_undef_if_not, :val, :(%4)))::Any -└── return %3 +4 ─ return %3 5 ─ return 0 ) => Int64 ```
As an application of #55545, this commit avoids the insertion of `:throw_undef_if_not` nodes when the defined-ness of a slot is guaranteed by abstract interpretation. ```julia julia> function isdefined_nothrow(c, x) local val if c val = x end if @isdefined val return val end return zero(Int) end; julia> @code_typed isdefined_nothrow(true, 42) ``` ```diff diff --git a/old b/new index c4980a5c9c..3d1d6d30f0 100644 --- a/old +++ b/new @@ -4,7 +4,6 @@ CodeInfo( 3 ┄ %3 = φ (#2 => x, #1 => #undef)::Int64 │ %4 = φ (#2 => true, #1 => false)::Bool └── goto #5 if not %4 -4 ─ $(Expr(:throw_undef_if_not, :val, :(%4)))::Any -└── return %3 +4 ─ return %3 5 ─ return 0 ) => Int64 ```
As an application of #55545, this commit avoids the insertion of `:throw_undef_if_not` nodes when the defined-ness of a slot is guaranteed by abstract interpretation. ```julia julia> function isdefined_nothrow(c, x) local val if c val = x end if @isdefined val return val end return zero(Int) end; julia> @code_typed isdefined_nothrow(true, 42) ``` ```diff diff --git a/old b/new index c4980a5c9c..3d1d6d30f0 100644 --- a/old +++ b/new @@ -4,7 +4,6 @@ CodeInfo( 3 ┄ %3 = φ (#2 => x, #1 => #undef)::Int64 │ %4 = φ (#2 => true, #1 => false)::Bool └── goto #5 if not %4 -4 ─ $(Expr(:throw_undef_if_not, :val, :(%4)))::Any -└── return %3 +4 ─ return %3 5 ─ return 0 ) => Int64 ```
Prior to this, especially on macOS, the gc-safepoint here would cause the process to segfault as we had already freed the current_task state. Rearrange this code so that the GC interactions (except for the atomic store to current_task) are all handled before entering GC safe, and then signaling the thread is deleted (via setting current_task = NULL, published by jl_unlock_profile_wr to other threads) is last. ``` ERROR: Exception handler triggered on unmanaged thread. Process 53827 stopped * thread #5, stop reason = EXC_BAD_ACCESS (code=2, address=0x100018008) frame #0: 0x0000000100b74344 libjulia-internal.1.12.0.dylib`jl_delete_thread [inlined] jl_gc_state_set(ptls=0x000000011f8b3200, state='\x02', old_state=<unavailable>) at julia_threads.h:272:9 [opt] 269 assert(old_state != JL_GC_CONCURRENT_COLLECTOR_THREAD); 270 jl_atomic_store_release(&ptls->gc_state, state); 271 if (state == JL_GC_STATE_UNSAFE || old_state == JL_GC_STATE_UNSAFE) -> 272 jl_gc_safepoint_(ptls); 273 return old_state; 274 } 275 STATIC_INLINE int8_t jl_gc_state_save_and_set(jl_ptls_t ptls, Target 0: (julia) stopped. (lldb) up frame #1: 0x0000000100b74320 libjulia-internal.1.12.0.dylib`jl_delete_thread [inlined] jl_gc_state_save_and_set(ptls=0x000000011f8b3200, state='\x02') at julia_threads.h:278:12 [opt] 275 STATIC_INLINE int8_t jl_gc_state_save_and_set(jl_ptls_t ptls, 276 int8_t state) 277 { -> 278 return jl_gc_state_set(ptls, state, jl_atomic_load_relaxed(&ptls->gc_state)); 279 } 280 #ifdef __clang_gcanalyzer__ 281 // these might not be a safepoint (if they are no-op safe=>safe transitions), but we have to assume it could be (statically) (lldb) frame #2: 0x0000000100b7431c libjulia-internal.1.12.0.dylib`jl_delete_thread(value=0x000000011f8b3200) at threading.c:537:11 [opt] 534 ptls->root_task = NULL; 535 jl_free_thread_gc_state(ptls); 536 // then park in safe-region -> 537 (void)jl_gc_safe_enter(ptls); 538 } ```
Prior to this, especially on macOS, the gc-safepoint here would cause the process to segfault as we had already freed the current_task state. Rearrange this code so that the GC interactions (except for the atomic store to current_task) are all handled before entering GC safe, and then signaling the thread is deleted (via setting current_task = NULL, published by jl_unlock_profile_wr to other threads) is last. ``` ERROR: Exception handler triggered on unmanaged thread. Process 53827 stopped * thread #5, stop reason = EXC_BAD_ACCESS (code=2, address=0x100018008) frame #0: 0x0000000100b74344 libjulia-internal.1.12.0.dylib`jl_delete_thread [inlined] jl_gc_state_set(ptls=0x000000011f8b3200, state='\x02', old_state=<unavailable>) at julia_threads.h:272:9 [opt] 269 assert(old_state != JL_GC_CONCURRENT_COLLECTOR_THREAD); 270 jl_atomic_store_release(&ptls->gc_state, state); 271 if (state == JL_GC_STATE_UNSAFE || old_state == JL_GC_STATE_UNSAFE) -> 272 jl_gc_safepoint_(ptls); 273 return old_state; 274 } 275 STATIC_INLINE int8_t jl_gc_state_save_and_set(jl_ptls_t ptls, Target 0: (julia) stopped. (lldb) up frame #1: 0x0000000100b74320 libjulia-internal.1.12.0.dylib`jl_delete_thread [inlined] jl_gc_state_save_and_set(ptls=0x000000011f8b3200, state='\x02') at julia_threads.h:278:12 [opt] 275 STATIC_INLINE int8_t jl_gc_state_save_and_set(jl_ptls_t ptls, 276 int8_t state) 277 { -> 278 return jl_gc_state_set(ptls, state, jl_atomic_load_relaxed(&ptls->gc_state)); 279 } 280 #ifdef __clang_gcanalyzer__ 281 // these might not be a safepoint (if they are no-op safe=>safe transitions), but we have to assume it could be (statically) (lldb) frame #2: 0x0000000100b7431c libjulia-internal.1.12.0.dylib`jl_delete_thread(value=0x000000011f8b3200) at threading.c:537:11 [opt] 534 ptls->root_task = NULL; 535 jl_free_thread_gc_state(ptls); 536 // then park in safe-region -> 537 (void)jl_gc_safe_enter(ptls); 538 } ```
…#55803) This slightly improves our (LLVM) codegen for `Core.throw_methoderror` and `Core.current_scope` ```julia julia> foo() = Core.current_scope() julia> bar() = Core.throw_methoderror(+, nothing) ``` Before: ```llvm ; Function Signature: foo() define nonnull ptr @julia_foo_2488() #0 { top: %0 = call ptr @jl_get_builtin_fptr(ptr nonnull @"+Core.#current_scope#2491.jit") %Builtin_ret = call nonnull ptr %0(ptr nonnull @"jl_global#2492.jit", ptr null, i32 0) ret ptr %Builtin_ret } ; Function Signature: bar() define void @julia_bar_589() #0 { top: %jlcallframe1 = alloca [2 x ptr], align 8 %0 = call ptr @jl_get_builtin_fptr(ptr nonnull @"+Core.#throw_methoderror#591.jit") %jl_nothing = load ptr, ptr @jl_nothing, align 8 store ptr @"jl_global#593.jit", ptr %jlcallframe1, align 8 %1 = getelementptr inbounds ptr, ptr %jlcallframe1, i64 1 store ptr %jl_nothing, ptr %1, align 8 %Builtin_ret = call nonnull ptr %0(ptr nonnull @"jl_global#592.jit", ptr nonnull %jlcallframe1, i32 2) call void @llvm.trap() unreachable } ``` After: ```llvm ; Function Signature: foo() define nonnull ptr @julia_foo_713() #0 { top: %thread_ptr = call ptr asm "movq %fs:0, $0", "=r"() #5 %tls_ppgcstack = getelementptr inbounds i8, ptr %thread_ptr, i64 -8 %tls_pgcstack = load ptr, ptr %tls_ppgcstack, align 8 %current_scope = getelementptr inbounds i8, ptr %tls_pgcstack, i64 -72 %0 = load ptr, ptr %current_scope, align 8 ret ptr %0 } ; Function Signature: bar() define void @julia_bar_1581() #0 { top: %jlcallframe1 = alloca [2 x ptr], align 8 %jl_nothing = load ptr, ptr @jl_nothing, align 8 store ptr @"jl_global#1583.jit", ptr %jlcallframe1, align 8 %0 = getelementptr inbounds ptr, ptr %jlcallframe1, i64 1 store ptr %jl_nothing, ptr %0, align 8 %jl_f_throw_methoderror_ret = call nonnull ptr @jl_f_throw_methoderror(ptr null, ptr nonnull %jlcallframe1, i32 2) call void @llvm.trap() unreachable } ```
Prior to this, especially on macOS, the gc-safepoint here would cause the process to segfault as we had already freed the current_task state. Rearrange this code so that the GC interactions (except for the atomic store to current_task) are all handled before entering GC safe, and then signaling the thread is deleted (via setting current_task = NULL, published by jl_unlock_profile_wr to other threads) is last. ``` ERROR: Exception handler triggered on unmanaged thread. Process 53827 stopped * thread #5, stop reason = EXC_BAD_ACCESS (code=2, address=0x100018008) frame #0: 0x0000000100b74344 libjulia-internal.1.12.0.dylib`jl_delete_thread [inlined] jl_gc_state_set(ptls=0x000000011f8b3200, state='\x02', old_state=<unavailable>) at julia_threads.h:272:9 [opt] 269 assert(old_state != JL_GC_CONCURRENT_COLLECTOR_THREAD); 270 jl_atomic_store_release(&ptls->gc_state, state); 271 if (state == JL_GC_STATE_UNSAFE || old_state == JL_GC_STATE_UNSAFE) -> 272 jl_gc_safepoint_(ptls); 273 return old_state; 274 } 275 STATIC_INLINE int8_t jl_gc_state_save_and_set(jl_ptls_t ptls, Target 0: (julia) stopped. (lldb) up frame #1: 0x0000000100b74320 libjulia-internal.1.12.0.dylib`jl_delete_thread [inlined] jl_gc_state_save_and_set(ptls=0x000000011f8b3200, state='\x02') at julia_threads.h:278:12 [opt] 275 STATIC_INLINE int8_t jl_gc_state_save_and_set(jl_ptls_t ptls, 276 int8_t state) 277 { -> 278 return jl_gc_state_set(ptls, state, jl_atomic_load_relaxed(&ptls->gc_state)); 279 } 280 #ifdef __clang_gcanalyzer__ 281 // these might not be a safepoint (if they are no-op safe=>safe transitions), but we have to assume it could be (statically) (lldb) frame #2: 0x0000000100b7431c libjulia-internal.1.12.0.dylib`jl_delete_thread(value=0x000000011f8b3200) at threading.c:537:11 [opt] 534 ptls->root_task = NULL; 535 jl_free_thread_gc_state(ptls); 536 // then park in safe-region -> 537 (void)jl_gc_safe_enter(ptls); 538 } ```
E.g. this allows `finalizer` inlining in the following case: ```julia mutable struct ForeignBuffer{T} const ptr::Ptr{T} end const foreign_buffer_finalized = Ref(false) function foreign_alloc(::Type{T}, length) where T ptr = Libc.malloc(sizeof(T) * length) ptr = Base.unsafe_convert(Ptr{T}, ptr) obj = ForeignBuffer{T}(ptr) return finalizer(obj) do obj Base.@assume_effects :notaskstate :nothrow foreign_buffer_finalized[] = true Libc.free(obj.ptr) end end function f_EA_finalizer(N::Int) workspace = foreign_alloc(Float64, N) GC.@preserve workspace begin (;ptr) = workspace Base.@assume_effects :nothrow @noinline println(devnull, "ptr = ", ptr) end end ``` ```julia julia> @code_typed f_EA_finalizer(42) CodeInfo( 1 ── %1 = Base.mul_int(8, N)::Int64 │ %2 = Core.lshr_int(%1, 63)::Int64 │ %3 = Core.trunc_int(Core.UInt8, %2)::UInt8 │ %4 = Core.eq_int(%3, 0x01)::Bool └─── goto #3 if not %4 2 ── invoke Core.throw_inexacterror(:convert::Symbol, UInt64::Type, %1::Int64)::Union{} └─── unreachable 3 ── goto #4 4 ── %9 = Core.bitcast(Core.UInt64, %1)::UInt64 └─── goto #5 5 ── goto #6 6 ── goto #7 7 ── goto #8 8 ── %14 = $(Expr(:foreigncall, :(:malloc), Ptr{Nothing}, svec(UInt64), 0, :(:ccall), :(%9), :(%9)))::Ptr{Nothing} └─── goto #9 9 ── %16 = Base.bitcast(Ptr{Float64}, %14)::Ptr{Float64} │ %17 = %new(ForeignBuffer{Float64}, %16)::ForeignBuffer{Float64} └─── goto #10 10 ─ %19 = $(Expr(:gc_preserve_begin, :(%17))) │ %20 = Base.getfield(%17, :ptr)::Ptr{Float64} │ invoke Main.println(Main.devnull::Base.DevNull, "ptr = "::String, %20::Ptr{Float64})::Nothing │ $(Expr(:gc_preserve_end, :(%19))) │ %23 = Main.foreign_buffer_finalized::Base.RefValue{Bool} │ Base.setfield!(%23, :x, true)::Bool │ %25 = Base.getfield(%17, :ptr)::Ptr{Float64} │ %26 = Base.bitcast(Ptr{Nothing}, %25)::Ptr{Nothing} │ $(Expr(:foreigncall, :(:free), Nothing, svec(Ptr{Nothing}), 0, :(:ccall), :(%26), :(%25)))::Nothing └─── return nothing ) => Nothing ``` However, this is still a WIP. Before merging, I want to improve EA's precision a bit and at least fix the test case that is currently marked as `broken`. I also need to check its impact on compiler performance. Additionally, I believe this feature is not yet practical. In particular, there is still significant room for improvement in the following areas: - EA's interprocedural capabilities: currently EA is performed ad-hoc for limited frames because of latency reasons, which significantly reduces its precision in the presence of interprocedural calls. - Relaxing the `:nothrow` check for finalizer inlining: the current algorithm requires `:nothrow`-ness on all paths from the allocation of the mutable struct to its last use, which is not practical for real-world cases. Even when `:nothrow` cannot be guaranteed, auxiliary optimizations such as inserting a `finalize` call after the last use might still be possible.
E.g. this allows `finalizer` inlining in the following case: ```julia mutable struct ForeignBuffer{T} const ptr::Ptr{T} end const foreign_buffer_finalized = Ref(false) function foreign_alloc(::Type{T}, length) where T ptr = Libc.malloc(sizeof(T) * length) ptr = Base.unsafe_convert(Ptr{T}, ptr) obj = ForeignBuffer{T}(ptr) return finalizer(obj) do obj Base.@assume_effects :notaskstate :nothrow foreign_buffer_finalized[] = true Libc.free(obj.ptr) end end function f_EA_finalizer(N::Int) workspace = foreign_alloc(Float64, N) GC.@preserve workspace begin (;ptr) = workspace Base.@assume_effects :nothrow @noinline println(devnull, "ptr = ", ptr) end end ``` ```julia julia> @code_typed f_EA_finalizer(42) CodeInfo( 1 ── %1 = Base.mul_int(8, N)::Int64 │ %2 = Core.lshr_int(%1, 63)::Int64 │ %3 = Core.trunc_int(Core.UInt8, %2)::UInt8 │ %4 = Core.eq_int(%3, 0x01)::Bool └─── goto #3 if not %4 2 ── invoke Core.throw_inexacterror(:convert::Symbol, UInt64::Type, %1::Int64)::Union{} └─── unreachable 3 ── goto #4 4 ── %9 = Core.bitcast(Core.UInt64, %1)::UInt64 └─── goto #5 5 ── goto #6 6 ── goto #7 7 ── goto #8 8 ── %14 = $(Expr(:foreigncall, :(:malloc), Ptr{Nothing}, svec(UInt64), 0, :(:ccall), :(%9), :(%9)))::Ptr{Nothing} └─── goto #9 9 ── %16 = Base.bitcast(Ptr{Float64}, %14)::Ptr{Float64} │ %17 = %new(ForeignBuffer{Float64}, %16)::ForeignBuffer{Float64} └─── goto #10 10 ─ %19 = $(Expr(:gc_preserve_begin, :(%17))) │ %20 = Base.getfield(%17, :ptr)::Ptr{Float64} │ invoke Main.println(Main.devnull::Base.DevNull, "ptr = "::String, %20::Ptr{Float64})::Nothing │ $(Expr(:gc_preserve_end, :(%19))) │ %23 = Main.foreign_buffer_finalized::Base.RefValue{Bool} │ Base.setfield!(%23, :x, true)::Bool │ %25 = Base.getfield(%17, :ptr)::Ptr{Float64} │ %26 = Base.bitcast(Ptr{Nothing}, %25)::Ptr{Nothing} │ $(Expr(:foreigncall, :(:free), Nothing, svec(Ptr{Nothing}), 0, :(:ccall), :(%26), :(%25)))::Nothing └─── return nothing ) => Nothing ``` However, this is still a WIP. Before merging, I want to improve EA's precision a bit and at least fix the test case that is currently marked as `broken`. I also need to check its impact on compiler performance. Additionally, I believe this feature is not yet practical. In particular, there is still significant room for improvement in the following areas: - EA's interprocedural capabilities: currently EA is performed ad-hoc for limited frames because of latency reasons, which significantly reduces its precision in the presence of interprocedural calls. - Relaxing the `:nothrow` check for finalizer inlining: the current algorithm requires `:nothrow`-ness on all paths from the allocation of the mutable struct to its last use, which is not practical for real-world cases. Even when `:nothrow` cannot be guaranteed, auxiliary optimizations such as inserting a `finalize` call after the last use might still be possible.
E.g. this allows `finalizer` inlining in the following case: ```julia mutable struct ForeignBuffer{T} const ptr::Ptr{T} end const foreign_buffer_finalized = Ref(false) function foreign_alloc(::Type{T}, length) where T ptr = Libc.malloc(sizeof(T) * length) ptr = Base.unsafe_convert(Ptr{T}, ptr) obj = ForeignBuffer{T}(ptr) return finalizer(obj) do obj Base.@assume_effects :notaskstate :nothrow foreign_buffer_finalized[] = true Libc.free(obj.ptr) end end function f_EA_finalizer(N::Int) workspace = foreign_alloc(Float64, N) GC.@preserve workspace begin (;ptr) = workspace Base.@assume_effects :nothrow @noinline println(devnull, "ptr = ", ptr) end end ``` ```julia julia> @code_typed f_EA_finalizer(42) CodeInfo( 1 ── %1 = Base.mul_int(8, N)::Int64 │ %2 = Core.lshr_int(%1, 63)::Int64 │ %3 = Core.trunc_int(Core.UInt8, %2)::UInt8 │ %4 = Core.eq_int(%3, 0x01)::Bool └─── goto #3 if not %4 2 ── invoke Core.throw_inexacterror(:convert::Symbol, UInt64::Type, %1::Int64)::Union{} └─── unreachable 3 ── goto #4 4 ── %9 = Core.bitcast(Core.UInt64, %1)::UInt64 └─── goto #5 5 ── goto #6 6 ── goto #7 7 ── goto #8 8 ── %14 = $(Expr(:foreigncall, :(:malloc), Ptr{Nothing}, svec(UInt64), 0, :(:ccall), :(%9), :(%9)))::Ptr{Nothing} └─── goto #9 9 ── %16 = Base.bitcast(Ptr{Float64}, %14)::Ptr{Float64} │ %17 = %new(ForeignBuffer{Float64}, %16)::ForeignBuffer{Float64} └─── goto #10 10 ─ %19 = $(Expr(:gc_preserve_begin, :(%17))) │ %20 = Base.getfield(%17, :ptr)::Ptr{Float64} │ invoke Main.println(Main.devnull::Base.DevNull, "ptr = "::String, %20::Ptr{Float64})::Nothing │ $(Expr(:gc_preserve_end, :(%19))) │ %23 = Main.foreign_buffer_finalized::Base.RefValue{Bool} │ Base.setfield!(%23, :x, true)::Bool │ %25 = Base.getfield(%17, :ptr)::Ptr{Float64} │ %26 = Base.bitcast(Ptr{Nothing}, %25)::Ptr{Nothing} │ $(Expr(:foreigncall, :(:free), Nothing, svec(Ptr{Nothing}), 0, :(:ccall), :(%26), :(%25)))::Nothing └─── return nothing ) => Nothing ``` However, this is still a WIP. Before merging, I want to improve EA's precision a bit and at least fix the test case that is currently marked as `broken`. I also need to check its impact on compiler performance. Additionally, I believe this feature is not yet practical. In particular, there is still significant room for improvement in the following areas: - EA's interprocedural capabilities: currently EA is performed ad-hoc for limited frames because of latency reasons, which significantly reduces its precision in the presence of interprocedural calls. - Relaxing the `:nothrow` check for finalizer inlining: the current algorithm requires `:nothrow`-ness on all paths from the allocation of the mutable struct to its last use, which is not practical for real-world cases. Even when `:nothrow` cannot be guaranteed, auxiliary optimizations such as inserting a `finalize` call after the last use might still be possible.
E.g. this allows `finalizer` inlining in the following case: ```julia mutable struct ForeignBuffer{T} const ptr::Ptr{T} end const foreign_buffer_finalized = Ref(false) function foreign_alloc(::Type{T}, length) where T ptr = Libc.malloc(sizeof(T) * length) ptr = Base.unsafe_convert(Ptr{T}, ptr) obj = ForeignBuffer{T}(ptr) return finalizer(obj) do obj Base.@assume_effects :notaskstate :nothrow foreign_buffer_finalized[] = true Libc.free(obj.ptr) end end function f_EA_finalizer(N::Int) workspace = foreign_alloc(Float64, N) GC.@preserve workspace begin (;ptr) = workspace Base.@assume_effects :nothrow @noinline println(devnull, "ptr = ", ptr) end end ``` ```julia julia> @code_typed f_EA_finalizer(42) CodeInfo( 1 ── %1 = Base.mul_int(8, N)::Int64 │ %2 = Core.lshr_int(%1, 63)::Int64 │ %3 = Core.trunc_int(Core.UInt8, %2)::UInt8 │ %4 = Core.eq_int(%3, 0x01)::Bool └─── goto #3 if not %4 2 ── invoke Core.throw_inexacterror(:convert::Symbol, UInt64::Type, %1::Int64)::Union{} └─── unreachable 3 ── goto #4 4 ── %9 = Core.bitcast(Core.UInt64, %1)::UInt64 └─── goto #5 5 ── goto #6 6 ── goto #7 7 ── goto #8 8 ── %14 = $(Expr(:foreigncall, :(:malloc), Ptr{Nothing}, svec(UInt64), 0, :(:ccall), :(%9), :(%9)))::Ptr{Nothing} └─── goto #9 9 ── %16 = Base.bitcast(Ptr{Float64}, %14)::Ptr{Float64} │ %17 = %new(ForeignBuffer{Float64}, %16)::ForeignBuffer{Float64} └─── goto #10 10 ─ %19 = $(Expr(:gc_preserve_begin, :(%17))) │ %20 = Base.getfield(%17, :ptr)::Ptr{Float64} │ invoke Main.println(Main.devnull::Base.DevNull, "ptr = "::String, %20::Ptr{Float64})::Nothing │ $(Expr(:gc_preserve_end, :(%19))) │ %23 = Main.foreign_buffer_finalized::Base.RefValue{Bool} │ Base.setfield!(%23, :x, true)::Bool │ %25 = Base.getfield(%17, :ptr)::Ptr{Float64} │ %26 = Base.bitcast(Ptr{Nothing}, %25)::Ptr{Nothing} │ $(Expr(:foreigncall, :(:free), Nothing, svec(Ptr{Nothing}), 0, :(:ccall), :(%26), :(%25)))::Nothing └─── return nothing ) => Nothing ``` However, this is still a WIP. Before merging, I want to improve EA's precision a bit and at least fix the test case that is currently marked as `broken`. I also need to check its impact on compiler performance. Additionally, I believe this feature is not yet practical. In particular, there is still significant room for improvement in the following areas: - EA's interprocedural capabilities: currently EA is performed ad-hoc for limited frames because of latency reasons, which significantly reduces its precision in the presence of interprocedural calls. - Relaxing the `:nothrow` check for finalizer inlining: the current algorithm requires `:nothrow`-ness on all paths from the allocation of the mutable struct to its last use, which is not practical for real-world cases. Even when `:nothrow` cannot be guaranteed, auxiliary optimizations such as inserting a `finalize` call after the last use might still be possible.
E.g. this allows `finalizer` inlining in the following case: ```julia mutable struct ForeignBuffer{T} const ptr::Ptr{T} end const foreign_buffer_finalized = Ref(false) function foreign_alloc(::Type{T}, length) where T ptr = Libc.malloc(sizeof(T) * length) ptr = Base.unsafe_convert(Ptr{T}, ptr) obj = ForeignBuffer{T}(ptr) return finalizer(obj) do obj Base.@assume_effects :notaskstate :nothrow foreign_buffer_finalized[] = true Libc.free(obj.ptr) end end function f_EA_finalizer(N::Int) workspace = foreign_alloc(Float64, N) GC.@preserve workspace begin (;ptr) = workspace Base.@assume_effects :nothrow @noinline println(devnull, "ptr = ", ptr) end end ``` ```julia julia> @code_typed f_EA_finalizer(42) CodeInfo( 1 ── %1 = Base.mul_int(8, N)::Int64 │ %2 = Core.lshr_int(%1, 63)::Int64 │ %3 = Core.trunc_int(Core.UInt8, %2)::UInt8 │ %4 = Core.eq_int(%3, 0x01)::Bool └─── goto #3 if not %4 2 ── invoke Core.throw_inexacterror(:convert::Symbol, UInt64::Type, %1::Int64)::Union{} └─── unreachable 3 ── goto #4 4 ── %9 = Core.bitcast(Core.UInt64, %1)::UInt64 └─── goto #5 5 ── goto #6 6 ── goto #7 7 ── goto #8 8 ── %14 = $(Expr(:foreigncall, :(:malloc), Ptr{Nothing}, svec(UInt64), 0, :(:ccall), :(%9), :(%9)))::Ptr{Nothing} └─── goto #9 9 ── %16 = Base.bitcast(Ptr{Float64}, %14)::Ptr{Float64} │ %17 = %new(ForeignBuffer{Float64}, %16)::ForeignBuffer{Float64} └─── goto #10 10 ─ %19 = $(Expr(:gc_preserve_begin, :(%17))) │ %20 = Base.getfield(%17, :ptr)::Ptr{Float64} │ invoke Main.println(Main.devnull::Base.DevNull, "ptr = "::String, %20::Ptr{Float64})::Nothing │ $(Expr(:gc_preserve_end, :(%19))) │ %23 = Main.foreign_buffer_finalized::Base.RefValue{Bool} │ Base.setfield!(%23, :x, true)::Bool │ %25 = Base.getfield(%17, :ptr)::Ptr{Float64} │ %26 = Base.bitcast(Ptr{Nothing}, %25)::Ptr{Nothing} │ $(Expr(:foreigncall, :(:free), Nothing, svec(Ptr{Nothing}), 0, :(:ccall), :(%26), :(%25)))::Nothing └─── return nothing ) => Nothing ``` However, this is still a WIP. Before merging, I want to improve EA's precision a bit and at least fix the test case that is currently marked as `broken`. I also need to check its impact on compiler performance. Additionally, I believe this feature is not yet practical. In particular, there is still significant room for improvement in the following areas: - EA's interprocedural capabilities: currently EA is performed ad-hoc for limited frames because of latency reasons, which significantly reduces its precision in the presence of interprocedural calls. - Relaxing the `:nothrow` check for finalizer inlining: the current algorithm requires `:nothrow`-ness on all paths from the allocation of the mutable struct to its last use, which is not practical for real-world cases. Even when `:nothrow` cannot be guaranteed, auxiliary optimizations such as inserting a `finalize` call after the last use might still be possible.
E.g. this allows `finalizer` inlining in the following case: ```julia mutable struct ForeignBuffer{T} const ptr::Ptr{T} end const foreign_buffer_finalized = Ref(false) function foreign_alloc(::Type{T}, length) where T ptr = Libc.malloc(sizeof(T) * length) ptr = Base.unsafe_convert(Ptr{T}, ptr) obj = ForeignBuffer{T}(ptr) return finalizer(obj) do obj Base.@assume_effects :notaskstate :nothrow foreign_buffer_finalized[] = true Libc.free(obj.ptr) end end function f_EA_finalizer(N::Int) workspace = foreign_alloc(Float64, N) GC.@preserve workspace begin (;ptr) = workspace Base.@assume_effects :nothrow @noinline println(devnull, "ptr = ", ptr) end end ``` ```julia julia> @code_typed f_EA_finalizer(42) CodeInfo( 1 ── %1 = Base.mul_int(8, N)::Int64 │ %2 = Core.lshr_int(%1, 63)::Int64 │ %3 = Core.trunc_int(Core.UInt8, %2)::UInt8 │ %4 = Core.eq_int(%3, 0x01)::Bool └─── goto #3 if not %4 2 ── invoke Core.throw_inexacterror(:convert::Symbol, UInt64::Type, %1::Int64)::Union{} └─── unreachable 3 ── goto #4 4 ── %9 = Core.bitcast(Core.UInt64, %1)::UInt64 └─── goto #5 5 ── goto #6 6 ── goto #7 7 ── goto #8 8 ── %14 = $(Expr(:foreigncall, :(:malloc), Ptr{Nothing}, svec(UInt64), 0, :(:ccall), :(%9), :(%9)))::Ptr{Nothing} └─── goto #9 9 ── %16 = Base.bitcast(Ptr{Float64}, %14)::Ptr{Float64} │ %17 = %new(ForeignBuffer{Float64}, %16)::ForeignBuffer{Float64} └─── goto #10 10 ─ %19 = $(Expr(:gc_preserve_begin, :(%17))) │ %20 = Base.getfield(%17, :ptr)::Ptr{Float64} │ invoke Main.println(Main.devnull::Base.DevNull, "ptr = "::String, %20::Ptr{Float64})::Nothing │ $(Expr(:gc_preserve_end, :(%19))) │ %23 = Main.foreign_buffer_finalized::Base.RefValue{Bool} │ Base.setfield!(%23, :x, true)::Bool │ %25 = Base.getfield(%17, :ptr)::Ptr{Float64} │ %26 = Base.bitcast(Ptr{Nothing}, %25)::Ptr{Nothing} │ $(Expr(:foreigncall, :(:free), Nothing, svec(Ptr{Nothing}), 0, :(:ccall), :(%26), :(%25)))::Nothing └─── return nothing ) => Nothing ``` However, this is still a WIP. Before merging, I want to improve EA's precision a bit and at least fix the test case that is currently marked as `broken`. I also need to check its impact on compiler performance. Additionally, I believe this feature is not yet practical. In particular, there is still significant room for improvement in the following areas: - EA's interprocedural capabilities: currently EA is performed ad-hoc for limited frames because of latency reasons, which significantly reduces its precision in the presence of interprocedural calls. - Relaxing the `:nothrow` check for finalizer inlining: the current algorithm requires `:nothrow`-ness on all paths from the allocation of the mutable struct to its last use, which is not practical for real-world cases. Even when `:nothrow` cannot be guaranteed, auxiliary optimizations such as inserting a `finalize` call after the last use might still be possible.
In an email discussion we came to the conclusion that it made sense to have multiple inheritance in Julia with one fairly simple restriction:
This restriction, together with Julia not allowing inheritance from non-abstract types, seems to address all the practical issues one typically encounters with multiple inheritance. The following, for example, would be disallowed:
Note that a generic function is an object external to all types, not a name inside of a type as it would be in a traditional object-orientation language. Thus, one can have
f(a::A)
in one namespace andf(b::B)
in another namespace without problems, so long as thef
s in these two namespaces are distinct generic function objects.The text was updated successfully, but these errors were encountered: