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.

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

UnitRange broacasting with literals gives a Vector #24586

Closed
c42f opened this issue Nov 12, 2017 · 9 comments · Fixed by #26891
Closed

UnitRange broacasting with literals gives a Vector #24586

c42f opened this issue Nov 12, 2017 · 9 comments · Fixed by #26891
Labels
domain:broadcast Applying a function over a collection

Comments

@c42f
Copy link
Member

c42f commented Nov 12, 2017

The lowering of dotted operators can have surprising effects when literals are mixed with ranges. In particular:

julia> a = 1:4
1:4

julia> a .+ 1    # Not a range!
4-element Array{Int64,1}:
 2
 3
 4
 5

julia> b = 1
1

julia> a .+ b  # A range, as is probably desired
2:5

julia> broadcast(+, a, 1)  # Also a range
2:5

The reason is that a .+ 1 is lowered to broadcast(x->x+1, a), so this doesn't match the special broadcast rule broadcast(::typeof(+), r::AbstractRange, x::Number).

In 0.6 this perhaps wasn't a big problem as you could use +(r::Range, x::Number), but this has been deprecated. However, the problem isn't specific to ranges - it's a general problem for any situation where you'd like the output type to depend on the function being broadcasted.

@c42f
Copy link
Member Author

c42f commented Nov 12, 2017

@Keno this was your gripe from today.

@fredrikekre
Copy link
Member

See #23445 and the linked issues/PRs. In particular #23445 (comment)

@StefanKarpinski
Copy link
Sponsor Member

StefanKarpinski commented Nov 12, 2017

I really think we should revisit the special casing of literals in broadcast and see if we can’t get the same effect with constant propagation. @vtjnash has been working on that, so I wonder if it may soon be good enough for this.

@c42f
Copy link
Member Author

c42f commented Nov 13, 2017

Improved constant propagation sounds like a great plan.

If that doesn't pan out in the short term, perhaps we could change the lowering to allow access to the rawer form in a way which doesn't break things. It'd be ugly internally, but we could give access to both the original function and the closure capturing the literals if a .+ 1 lowered to broadcast_opt(broadcast_raw(+, a, 1), x->x+1, a)

And in turn we had the following default implementations:

struct NoRawBroadcast; end
broadcast_raw(args...) = NoRawBroadcast()
broadcast_opt(::NoRawBroadcast, f, args...) = broadcast(f, args...)
broadcast_opt(a, args...) = a

Users would then overload braodcast_raw(::typeof(+), ::AbstractRange, ::Number) to fix the literal problem.

I know, it's a pretty ugly solution. But it does show that lowering gives us some options to solve this with the current state of the compiler technology.

@timholy
Copy link
Sponsor Member

timholy commented Nov 13, 2017

As part of #23939 we could give AbstractRange its own BroadcastStyle and appropriate combination rules.

@c42f
Copy link
Member Author

c42f commented Nov 13, 2017

Would that help? It seems to me we need something more like Jameson's work in #23692 to solve this issue.

[Edit] which is essentially attacking the same problem as the hack I proposed above, but in a more general and non-disgusting way :-)

@andyferris
Copy link
Member

@timholy we need to specialize the output container based on the input containers and the operator, unitrange .+ integer --> unitrange, potentially unitrange .+ unitrange --> range, etc. Here it's that the dot fusion at lowering that masks the use of +.

Constant propagation may solve the literal problem but it's a symptom of fusion in general. I agree with @c42f that it seems to me we either need something like #23692, or perhaps to make broadcast lazy-and-not-fused-at-lowering.

@timholy
Copy link
Sponsor Member

timholy commented Nov 13, 2017

Hmm, yes, here we do need to pay attention to the operator.

When it's operator-specific then we could even consider just short-circuiting the whole call. But that would presumably run into trouble with 1 .+ r .- 1. So either the combination rules need to take the operator into account, and/or we need #23692.

@JeffBezanson JeffBezanson added the domain:broadcast Applying a function over a collection label Nov 14, 2017
@JeffBezanson
Copy link
Sponsor Member

Ah, savor the sour fruits of breaking referential transparency.

mbauman added a commit that referenced this issue Apr 23, 2018
This patch represents the combined efforts of four individuals, over 60
commits, and an iterated design over (at least) three pull requests that
spanned nearly an entire year (closes #22063, #23692, #25377 by superceding
them).

This introduces a pure Julia data structure that represents a fused broadcast
expression.  For example, the expression `2 .* (x .+ 1)` lowers to:

```julia
julia> Meta.@lower 2 .* (x .+ 1)
:($(Expr(:thunk, CodeInfo(:(begin
      Core.SSAValue(0) = (Base.getproperty)(Base.Broadcast, :materialize)
      Core.SSAValue(1) = (Base.getproperty)(Base.Broadcast, :make)
      Core.SSAValue(2) = (Base.getproperty)(Base.Broadcast, :make)
      Core.SSAValue(3) = (Core.SSAValue(2))(+, x, 1)
      Core.SSAValue(4) = (Core.SSAValue(1))(*, 2, Core.SSAValue(3))
      Core.SSAValue(5) = (Core.SSAValue(0))(Core.SSAValue(4))
      return Core.SSAValue(5)
  end)))))
```

Or, slightly more readably as:

```julia
using .Broadcast: materialize, make
materialize(make(*, 2, make(+, x, 1)))
```

The `Broadcast.make` function serves two purposes. Its primary purpose is to
construct the `Broadcast.Broadcasted` objects that hold onto the function, the
tuple of arguments (potentially including nested `Broadcasted` arguments), and
sometimes a set of `axes` to include knowledge of the outer shape. The
secondary purpose, however, is to allow an "out" for objects that _don't_ want
to participate in fusion. For example, if `x` is a range in the above `2 .* (x
.+ 1)` expression, it needn't allocate an array and operate elementwise — it
can just compute and return a new range. Thus custom structures are able to
specialize `Broadcast.make(f, args...)` just as they'd specialize on `f`
normally to return an immediate result.

`Broadcast.materialize` is identity for everything _except_ `Broadcasted`
objects for which it allocates an appropriate result and computes the
broadcast. It does two things: it `initialize`s the outermost `Broadcasted`
object to compute its axes and then `copy`s it.

Similarly, an in-place fused broadcast like `y .= 2 .* (x .+ 1)` uses the exact
same expression tree to compute the right-hand side of the expression as above,
and then uses `materialize!(y, make(*, 2, make(+, x, 1)))` to `instantiate` the
`Broadcasted` expression tree and then `copyto!` it into the given destination.

All-together, this forms a complete API for custom types to extend and
customize the behavior of broadcast (fixes #22060). It uses the existing
`BroadcastStyle`s throughout to simplify dispatch on many arguments:

* Custom types can opt-out of broadcast fusion by specializing
  `Broadcast.make(f, args...)` or `Broadcast.make(::BroadcastStyle, f, args...)`.

* The `Broadcasted` object computes and stores the type of the combined
  `BroadcastStyle` of its arguments as its first type parameter, allowing for
  easy dispatch and specialization.

* Custom Broadcast storage is still allocated via `broadcast_similar`, however
  instead of passing just a function as a first argument, the entire
  `Broadcasted` object is passed as a final argument. This potentially allows
  for much more runtime specialization dependent upon the exact expression
  given.

* Custom broadcast implmentations for a `CustomStyle` are defined by
  specializing `copy(bc::Broadcasted{CustomStyle})` or
  `copyto!(dest::AbstractArray, bc::Broadcasted{CustomStyle})`.

* Fallback broadcast specializations for a given output object of type `Dest`
  (for the `DefaultArrayStyle` or another such style that hasn't implemented
  assignments into such an object) are defined by specializing
  `copyto(dest::Dest, bc::Broadcasted{Nothing})`.

As it fully supports range broadcasting, this now deprecates `(1:5) + 2` to
`.+`, just as had been done for all `AbstractArray`s in general.

As a first-mover proof of concept, LinearAlgebra uses this new system to
improve broadcasting over structured arrays. Before, broadcasting over a
structured matrix would result in a sparse array. Now, broadcasting over a
structured matrix will _either_ return an appropriately structured matrix _or_
a dense array. This does incur a type instability (in the form of a
discriminated union) in some situations, but thanks to type-based introspection
of the `Broadcasted` wrapper commonly used functions can be special cased to be
type stable.  For example:

```julia
julia> f(d) = round.(Int, d)
f (generic function with 1 method)

julia> @inferred f(Diagonal(rand(3)))
3×3 Diagonal{Int64,Array{Int64,1}}:
 0  ⋅  ⋅
 ⋅  0  ⋅
 ⋅  ⋅  1

julia> @inferred Diagonal(rand(3)) .* 3
ERROR: return type Diagonal{Float64,Array{Float64,1}} does not match inferred return type Union{Array{Float64,2}, Diagonal{Float64,Array{Float64,1}}}
Stacktrace:
 [1] error(::String) at ./error.jl:33
 [2] top-level scope

julia> @inferred Diagonal(1:4) .+ Bidiagonal(rand(4), rand(3), 'U') .* Tridiagonal(1:3, 1:4, 1:3)
4×4 Tridiagonal{Float64,Array{Float64,1}}:
 1.30771  0.838589   ⋅          ⋅
 0.0      3.89109   0.0459757   ⋅
  ⋅       0.0       4.48033    2.51508
  ⋅        ⋅        0.0        6.23739
```

In addition to the issues referenced above, it fixes:

* Fixes #19313, #22053, #23445, and #24586: Literals are no longer treated
  specially in a fused broadcast; they're just arguments in a `Broadcasted`
  object like everything else.

* Fixes #21094: Since broadcasting is now represented by a pure Julia
  datastructure it can be created within `@generated` functions and serialized.

* Fixes #26097: The fallback destination-array specialization method of
  `copyto!` is specifically implemented as `Broadcasted{Nothing}` and will not
  be confused by `nothing` arguments.

* Fixes the broadcast-specific element of #25499: The default base broadcast
  implementation no longer depends upon `Base._return_type` to allocate its
  array (except in the empty or concretely-type cases). Note that the sparse
  implementation (#19595) is still dependent upon inference and is _not_ fixed.

* Fixes #25340: Functions are treated like normal values just like arguments
  and only evaluated once.

* Fixes #22255, and is performant with 12+ fused broadcasts. Okay, that one was
  fixed on master already, but this fixes it now, too.

* Fixes #25521.

* The performance of this patch has been thoroughly tested through its
  iterative development process in #25377. There remain [two classes of
  performance regressions](#25377) that Nanosoldier flagged.

* #25691: Propagation of constant literals sill lose their constant-ness upon
  going through the broadcast machinery. I believe quite a large number of
  functions would need to be marked as `@pure` to support this -- including
  functions that are intended to be specialized.

(For bookkeeping, this is the squashed version of the [teh-jn/lazydotfuse](#25377)
branch as of a1d4e7e. Squashed and separated
out to make it easier to review and commit)

Co-authored-by: Tim Holy <tim.holy@gmail.com>
Co-authored-by: Jameson Nash <vtjnash@gmail.com>
Co-authored-by: Andrew Keller <ajkeller34@users.noreply.github.com>
Keno pushed a commit that referenced this issue Apr 27, 2018
This patch represents the combined efforts of four individuals, over 60
commits, and an iterated design over (at least) three pull requests that
spanned nearly an entire year (closes #22063, #23692, #25377 by superceding
them).

This introduces a pure Julia data structure that represents a fused broadcast
expression.  For example, the expression `2 .* (x .+ 1)` lowers to:

```julia
julia> Meta.@lower 2 .* (x .+ 1)
:($(Expr(:thunk, CodeInfo(:(begin
      Core.SSAValue(0) = (Base.getproperty)(Base.Broadcast, :materialize)
      Core.SSAValue(1) = (Base.getproperty)(Base.Broadcast, :make)
      Core.SSAValue(2) = (Base.getproperty)(Base.Broadcast, :make)
      Core.SSAValue(3) = (Core.SSAValue(2))(+, x, 1)
      Core.SSAValue(4) = (Core.SSAValue(1))(*, 2, Core.SSAValue(3))
      Core.SSAValue(5) = (Core.SSAValue(0))(Core.SSAValue(4))
      return Core.SSAValue(5)
  end)))))
```

Or, slightly more readably as:

```julia
using .Broadcast: materialize, make
materialize(make(*, 2, make(+, x, 1)))
```

The `Broadcast.make` function serves two purposes. Its primary purpose is to
construct the `Broadcast.Broadcasted` objects that hold onto the function, the
tuple of arguments (potentially including nested `Broadcasted` arguments), and
sometimes a set of `axes` to include knowledge of the outer shape. The
secondary purpose, however, is to allow an "out" for objects that _don't_ want
to participate in fusion. For example, if `x` is a range in the above `2 .* (x
.+ 1)` expression, it needn't allocate an array and operate elementwise — it
can just compute and return a new range. Thus custom structures are able to
specialize `Broadcast.make(f, args...)` just as they'd specialize on `f`
normally to return an immediate result.

`Broadcast.materialize` is identity for everything _except_ `Broadcasted`
objects for which it allocates an appropriate result and computes the
broadcast. It does two things: it `initialize`s the outermost `Broadcasted`
object to compute its axes and then `copy`s it.

Similarly, an in-place fused broadcast like `y .= 2 .* (x .+ 1)` uses the exact
same expression tree to compute the right-hand side of the expression as above,
and then uses `materialize!(y, make(*, 2, make(+, x, 1)))` to `instantiate` the
`Broadcasted` expression tree and then `copyto!` it into the given destination.

All-together, this forms a complete API for custom types to extend and
customize the behavior of broadcast (fixes #22060). It uses the existing
`BroadcastStyle`s throughout to simplify dispatch on many arguments:

* Custom types can opt-out of broadcast fusion by specializing
  `Broadcast.make(f, args...)` or `Broadcast.make(::BroadcastStyle, f, args...)`.

* The `Broadcasted` object computes and stores the type of the combined
  `BroadcastStyle` of its arguments as its first type parameter, allowing for
  easy dispatch and specialization.

* Custom Broadcast storage is still allocated via `broadcast_similar`, however
  instead of passing just a function as a first argument, the entire
  `Broadcasted` object is passed as a final argument. This potentially allows
  for much more runtime specialization dependent upon the exact expression
  given.

* Custom broadcast implmentations for a `CustomStyle` are defined by
  specializing `copy(bc::Broadcasted{CustomStyle})` or
  `copyto!(dest::AbstractArray, bc::Broadcasted{CustomStyle})`.

* Fallback broadcast specializations for a given output object of type `Dest`
  (for the `DefaultArrayStyle` or another such style that hasn't implemented
  assignments into such an object) are defined by specializing
  `copyto(dest::Dest, bc::Broadcasted{Nothing})`.

As it fully supports range broadcasting, this now deprecates `(1:5) + 2` to
`.+`, just as had been done for all `AbstractArray`s in general.

As a first-mover proof of concept, LinearAlgebra uses this new system to
improve broadcasting over structured arrays. Before, broadcasting over a
structured matrix would result in a sparse array. Now, broadcasting over a
structured matrix will _either_ return an appropriately structured matrix _or_
a dense array. This does incur a type instability (in the form of a
discriminated union) in some situations, but thanks to type-based introspection
of the `Broadcasted` wrapper commonly used functions can be special cased to be
type stable.  For example:

```julia
julia> f(d) = round.(Int, d)
f (generic function with 1 method)

julia> @inferred f(Diagonal(rand(3)))
3×3 Diagonal{Int64,Array{Int64,1}}:
 0  ⋅  ⋅
 ⋅  0  ⋅
 ⋅  ⋅  1

julia> @inferred Diagonal(rand(3)) .* 3
ERROR: return type Diagonal{Float64,Array{Float64,1}} does not match inferred return type Union{Array{Float64,2}, Diagonal{Float64,Array{Float64,1}}}
Stacktrace:
 [1] error(::String) at ./error.jl:33
 [2] top-level scope

julia> @inferred Diagonal(1:4) .+ Bidiagonal(rand(4), rand(3), 'U') .* Tridiagonal(1:3, 1:4, 1:3)
4×4 Tridiagonal{Float64,Array{Float64,1}}:
 1.30771  0.838589   ⋅          ⋅
 0.0      3.89109   0.0459757   ⋅
  ⋅       0.0       4.48033    2.51508
  ⋅        ⋅        0.0        6.23739
```

In addition to the issues referenced above, it fixes:

* Fixes #19313, #22053, #23445, and #24586: Literals are no longer treated
  specially in a fused broadcast; they're just arguments in a `Broadcasted`
  object like everything else.

* Fixes #21094: Since broadcasting is now represented by a pure Julia
  datastructure it can be created within `@generated` functions and serialized.

* Fixes #26097: The fallback destination-array specialization method of
  `copyto!` is specifically implemented as `Broadcasted{Nothing}` and will not
  be confused by `nothing` arguments.

* Fixes the broadcast-specific element of #25499: The default base broadcast
  implementation no longer depends upon `Base._return_type` to allocate its
  array (except in the empty or concretely-type cases). Note that the sparse
  implementation (#19595) is still dependent upon inference and is _not_ fixed.

* Fixes #25340: Functions are treated like normal values just like arguments
  and only evaluated once.

* Fixes #22255, and is performant with 12+ fused broadcasts. Okay, that one was
  fixed on master already, but this fixes it now, too.

* Fixes #25521.

* The performance of this patch has been thoroughly tested through its
  iterative development process in #25377. There remain [two classes of
  performance regressions](#25377) that Nanosoldier flagged.

* #25691: Propagation of constant literals sill lose their constant-ness upon
  going through the broadcast machinery. I believe quite a large number of
  functions would need to be marked as `@pure` to support this -- including
  functions that are intended to be specialized.

(For bookkeeping, this is the squashed version of the [teh-jn/lazydotfuse](#25377)
branch as of a1d4e7e. Squashed and separated
out to make it easier to review and commit)

Co-authored-by: Tim Holy <tim.holy@gmail.com>
Co-authored-by: Jameson Nash <vtjnash@gmail.com>
Co-authored-by: Andrew Keller <ajkeller34@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
domain:broadcast Applying a function over a collection
Projects
None yet
Development

Successfully merging a pull request may close this issue.

6 participants