Backports for 1.8-rc2/1.8.0

Make.inc

@@ -900,10 +900,6 @@ OPENBLAS_DYNAMIC_ARCH:=0
 OPENBLAS_TARGET_ARCH:=ARMV8
 USE_BLAS64:=1
 BINARY:=64
-ifeq ($(OS),Darwin)
-# Apple Chips are all at least A12Z
-MCPU:=apple-a12
-endif
 endif
 
 # Set MARCH-specific flags

NEWS.md

@@ -11,8 +11,8 @@ New language features
   e.g. `[;;;]` creates a 0×0×0 `Array` ([#41618]).
 * `try`-blocks can now optionally have an `else`-block which is executed right after the main body only if
   no errors were thrown ([#42211]).
-* `@inline` and `@noinline` annotations can now be placed within a function body ([#41312]).
-* `@inline` and `@noinline` annotations can now be applied to a function call site or block
+* `@inline` and `@noinline` can now be placed within a function body, allowing one to annotate anonymous function ([#41312]).
+* `@inline` and `@noinline` can now be applied to a function at callsite or block
   to enforce the involved function calls to be (or not to be) inlined ([#41328]).
 * `∀`, `∃`, and `∄` are now allowed as identifier characters ([#42314]).
 * Support for Unicode 14.0.0 ([#43443]).
@@ -43,7 +43,9 @@ Compiler/Runtime improvements
   `libjulia-codegen`. It is loaded by default, so normal usage should see no changes.
   In deployments that do not need the compiler (e.g. system images where all needed code
   is precompiled), this library (and its LLVM dependency) can simply be excluded ([#41936]).
-* Conditional type constraints can now be forwarded interprocedurally (i.e. propagated from caller to callee) ([#42529]).
+* Conditional type constraints are now be forwarded interprocedurally (i.e. propagated from caller to callee).
+  This allows inference to understand e.g. `Base.ifelse(isa(x, Int), x, 0)` returns `::Int`-value
+  even if the type of `x` is not known ([#42529]).
 * Julia-level SROA (Scalar Replacement of Aggregates) has been improved: allowing elimination of
   `getfield` calls with constant global fields ([#42355]), enabling elimination of mutable structs with
   uninitialized fields ([#43208]), improving performance ([#43232]), and handling more nested `getfield`
@@ -53,7 +55,7 @@ Compiler/Runtime improvements
 * Inference now tracks various effects such as side-effectful-ness and nothrow-ness on a per-specialization basis.
   Code heavily dependent on constant propagation should see significant compile-time performance improvements and
   certain cases (e.g. calls to uninlinable functions that are nevertheless effect free) should see runtime performance
-  improvements. Effects may be overwritten manually with the `@Base.assume_effects` macro ([#43852]).
+  improvements. Effects may be overwritten manually with the `Base.@assume_effects` macro ([#43852]).
 * Precompilation (with explicit `precompile` directives or representative workloads) now saves more type-inferred code,
   resulting in reduced time-to-first task for packages that use precompilation.  This change also eliminates the
   runtime performance degradation occasionally triggered by precompilation on older Julia versions. More specifically,
@@ -108,6 +110,8 @@ New library features
 * TCP socket objects now expose `closewrite` functionality and support half-open mode usage ([#40783]).
 * `extrema` now accepts an `init` keyword argument ([#36265], [#43604]).
 * `Iterators.countfrom` now accepts any type that defines `+` ([#37747]).
+* `@time` now separates out % time spent recompiling invalidated methods ([#45015]).
+* `@time_imports` now shows any compilation and recompilation time percentages per import ([#45064]).
 
 Standard library changes
 ------------------------

base/abstractarray.jl

@@ -2478,7 +2478,7 @@ function _typed_hvncat_shape(::Type{T}, shape::NTuple{N, Tuple}, row_first, as::
     shapelength == lengthas || throw(ArgumentError("number of elements does not match shape; expected $(shapelength), got $lengthas)"))
     # discover dimensions
     nd = max(N, cat_ndims(as[1]))
-    outdims = zeros(Int, nd)
+    outdims = fill(-1, nd)
     currentdims = zeros(Int, nd)
     blockcounts = zeros(Int, nd)
     shapepos = ones(Int, nd)
@@ -2503,7 +2503,7 @@ function _typed_hvncat_shape(::Type{T}, shape::NTuple{N, Tuple}, row_first, as::
 
             isendblock = blockcounts[d] == shapev[d][shapepos[d]]
             if isendblock
-                if outdims[d] == 0
+                if outdims[d] == -1
                     outdims[d] = currentdims[d]
                 elseif outdims[d] != currentdims[d]
                     throw(ArgumentError("argument $i has a mismatched number of elements along axis $ad; \

base/compiler/abstractinterpretation.jl

@@ -720,7 +720,7 @@ function concrete_eval_eligible(interp::AbstractInterpreter,
     isoverlayed(method_table(interp)) && !is_nonoverlayed(result.edge_effects) && return false
     return f !== nothing &&
            result.edge !== nothing &&
-           is_concrete_eval_eligible(result.edge_effects) &&
+           is_foldable(result.edge_effects) &&
            is_all_const_arg(arginfo)
 end
 

base/compiler/ssair/inlining.jl

@@ -1257,11 +1257,7 @@ function handle_const_call!(
             result = results[j]
             if isa(result, ConstResult)
                 case = const_result_item(result, state)
-                if case === nothing
-                    fully_covered = false
-                else
-                    push!(cases, InliningCase(result.mi.specTypes, case))
-                end
+                push!(cases, InliningCase(result.mi.specTypes, case))
             elseif isa(result, InferenceResult)
                 fully_covered &= handle_inf_result!(result, argtypes, flag, state, cases)
             else
@@ -1314,7 +1310,9 @@ end
 
 function const_result_item(result::ConstResult, state::InliningState)
     if !isdefined(result, :result) || !is_inlineable_constant(result.result)
-        return compileable_specialization(state.et, result.mi, result.effects)
+        case = compileable_specialization(state.et, result.mi, result.effects)
+        @assert case !== nothing "concrete evaluation should never happen for uncompileable callsite"
+        return case
     end
     @assert result.effects === EFFECTS_TOTAL
     return ConstantCase(quoted(result.result))

base/compiler/tfuncs.jl

@@ -1515,6 +1515,10 @@ function tuple_tfunc(argtypes::Vector{Any})
             params[i] = typeof(x.val)
         else
             x = isvarargtype(x) ? x : widenconst(x)
+            # since there don't exist any values whose runtime type are `Tuple{Type{...}}`,
+            # here we should turn such `Type{...}`-parameters to valid parameters, e.g.
+            # (::Type{Int},) -> Tuple{DataType} (or PartialStruct for more accuracy)
+            # (::Union{Type{Int32},Type{Int64}}) -> Tuple{Type}
             if isType(x)
                 anyinfo = true
                 xparam = x.parameters[1]
@@ -1523,6 +1527,10 @@ function tuple_tfunc(argtypes::Vector{Any})
                 else
                     params[i] = Type
                 end
+            elseif iskindtype(x)
+                params[i] = x
+            elseif !isvarargtype(x) && hasintersect(x, Type)
+                params[i] = Union{x, Type}
             else
                 params[i] = x
             end

base/compiler/types.jl

@@ -113,13 +113,13 @@ is_nothrow(effects::Effects)      = effects.nothrow === ALWAYS_TRUE
 is_terminates(effects::Effects)   = effects.terminates === ALWAYS_TRUE
 is_nonoverlayed(effects::Effects) = effects.nonoverlayed
 
-is_concrete_eval_eligible(effects::Effects) =
+is_foldable(effects::Effects) =
     is_consistent(effects) &&
     is_effect_free(effects) &&
     is_terminates(effects)
 
 is_total(effects::Effects) =
-    is_concrete_eval_eligible(effects) &&
+    is_foldable(effects) &&
     is_nothrow(effects)
 
 is_removable_if_unused(effects::Effects) =

base/essentials.jl

@@ -766,7 +766,7 @@ function invoke_in_world(world::UInt, @nospecialize(f), @nospecialize args...; k
 end
 
 # TODO: possibly make this an intrinsic
-inferencebarrier(@nospecialize(x)) = Ref{Any}(x)[]
+inferencebarrier(@nospecialize(x)) = RefValue{Any}(x).x
 
 """
     isempty(collection) -> Bool

base/expr.jl

@@ -381,6 +381,9 @@ end
 `@assume_effects` overrides the compiler's effect modeling for the given method.
 `ex` must be a method definition or `@ccall` expression.
 
+!!! compat "Julia 1.8"
+    Using `Base.@assume_effects` requires Julia version 1.8.
+
 ```jldoctest
 julia> Base.@assume_effects :terminates_locally function pow(x)
            # this :terminates_locally allows `pow` to be constant-folded
@@ -402,7 +405,7 @@ julia> code_typed() do
 1 ─     return 479001600
 ) => Int64
 
-julia> Base.@assume_effects :total_may_throw @ccall jl_type_intersection(Vector{Int}::Any, Vector{<:Integer}::Any)::Any
+julia> Base.@assume_effects :total !:nothrow @ccall jl_type_intersection(Vector{Int}::Any, Vector{<:Integer}::Any)::Any
 Vector{Int64} (alias for Array{Int64, 1})
 ```
 
@@ -423,12 +426,15 @@ The following `setting`s are supported.
 - `:nothrow`
 - `:terminates_globally`
 - `:terminates_locally`
+- `:foldable`
 - `:total`
 
+# Extended help
+
 ---
-# `:consistent`
+## `:consistent`
 
-The `:consistent` setting asserts that for egal inputs:
+The `:consistent` setting asserts that for egal (`===`) inputs:
 - The manner of termination (return value, exception, non-termination) will always be the same.
 - If the method returns, the results will always be egal.
 
@@ -461,7 +467,7 @@ The `:consistent` setting asserts that for egal inputs:
     itself is not required to meet the egality requirement specified above.
 
 ---
-# `:effect_free`
+## `:effect_free`
 
 The `:effect_free` setting asserts that the method is free of externally semantically
 visible side effects. The following is an incomplete list of externally semantically
@@ -491,7 +497,7 @@ were not executed.
     valid for all world ages and limit use of this assertion accordingly.
 
 ---
-# `:nothrow`
+## `:nothrow`
 
 The `:nothrow` settings asserts that this method does not terminate abnormally
 (i.e. will either always return a value or never return).
@@ -505,7 +511,7 @@ The `:nothrow` settings asserts that this method does not terminate abnormally
     `MethodErrors` and similar exceptions count as abnormal termination.
 
 ---
-# `:terminates_globally`
+## `:terminates_globally`
 
 The `:terminates_globally` settings asserts that this method will eventually terminate
 (either normally or abnormally), i.e. does not loop indefinitely.
@@ -520,7 +526,7 @@ The `:terminates_globally` settings asserts that this method will eventually ter
     on a method that *technically*, but not *practically*, terminates.
 
 ---
-# `:terminates_locally`
+## `:terminates_locally`
 
 The `:terminates_locally` setting is like `:terminates_globally`, except that it only
 applies to syntactic control flow *within* the annotated method. It is thus
@@ -531,59 +537,79 @@ non-termination if the method calls some other method that does not terminate.
     `:terminates_globally` implies `:terminates_locally`.
 
 ---
-# `:total`
+## `:foldable`
+
+This setting is a convenient shortcut for the set of effects that the compiler
+requires to be guaranteed to constant fold a call at compile time. It is
+currently equivalent to the following `setting`s:
 
-This `setting` combines the following other assertions:
 - `:consistent`
 - `:effect_free`
-- `:nothrow`
 - `:terminates_globally`
-and is a convenient shortcut.
+
+!!! note
+    This list in particular does not include `:nothrow`. The compiler will still
+    attempt constant propagation and note any thrown error at compile time. Note
+    however, that by the `:consistent`-cy requirements, any such annotated call
+    must consistently throw given the same argument values.
 
 ---
-# `:total_may_throw`
+## `:total`
 
-This `setting` combines the following other assertions:
+This `setting` is the maximum possible set of effects. It currently implies
+the following other `setting`s:
 - `:consistent`
 - `:effect_free`
+- `:nothrow`
 - `:terminates_globally`
-and is a convenient shortcut.
 
-!!! note
-    This setting is particularly useful since it allows the compiler to evaluate a call of
-    the applied method when all the call arguments are fully known to be constant, no matter
-    if the call results in an error or not.
-
-    `@assume_effects :total_may_throw` is similar to [`@pure`](@ref) with the primary
-    distinction that the `:consistent`-cy requirement applies world-age wise rather
-    than globally as described above. However, in particular, a method annotated
-    `@pure` should always be `:total` or `:total_may_throw`.
-    Another advantage is that effects introduced by `@assume_effects` are propagated to
-    callers interprocedurally while a purity defined by `@pure` is not.
+!!! warning
+    `:total` is a very strong assertion and will likely gain additional semantics
+    in future versions of Julia (e.g. if additional effects are added and included
+    in the definition of `:total`). As a result, it should be used with care.
+    Whenever possible, prefer to use the minimum possible set of specific effect
+    assertions required for a particular application. In cases where a large
+    number of effect overrides apply to a set of functions, a custom macro is
+    recommended over the use of `:total`.
+
+---
+## Negated effects
+
+Effect names may be prefixed by `!` to indicate that the effect should be removed
+from an earlier meta effect. For example, `:total !:nothrow` indicates that while
+the call is generally total, it may however throw.
+
+---
+## Comparison to `@pure`
+
+`@assume_effects :foldable` is similar to [`@pure`](@ref) with the primary
+distinction that the `:consistent`-cy requirement applies world-age wise rather
+than globally as described above. However, in particular, a method annotated
+`@pure` should always be at least `:foldable`.
+Another advantage is that effects introduced by `@assume_effects` are propagated to
+callers interprocedurally while a purity defined by `@pure` is not.
 """
 macro assume_effects(args...)
     (consistent, effect_free, nothrow, terminates_globally, terminates_locally) =
         (false, false, false, false, false, false)
-    for setting in args[1:end-1]
-        if isa(setting, QuoteNode)
-            setting = setting.value
-        end
+    for org_setting in args[1:end-1]
+        (setting, val) = compute_assumed_setting(org_setting)
         if setting === :consistent
-            consistent = true
+            consistent = val
         elseif setting === :effect_free
-            effect_free = true
+            effect_free = val
         elseif setting === :nothrow
-            nothrow = true
+            nothrow = val
         elseif setting === :terminates_globally
-            terminates_globally = true
+            terminates_globally = val
         elseif setting === :terminates_locally
-            terminates_locally = true
+            terminates_locally = val
+        elseif setting === :foldable
+            consistent = effect_free = terminates_globally = val
         elseif setting === :total
-            consistent = effect_free = nothrow = terminates_globally = true
-        elseif setting === :total_may_throw
-            consistent = effect_free = terminates_globally = true
+            consistent = effect_free = nothrow = terminates_globally = val
         else
-            throw(ArgumentError("@assume_effects $setting not supported"))
+            throw(ArgumentError("@assume_effects $org_setting not supported"))
         end
     end
     ex = args[end]
@@ -598,6 +624,16 @@ macro assume_effects(args...)
     return esc(pushmeta!(ex, :purity, consistent, effect_free, nothrow, terminates_globally, terminates_locally))
 end
 
+function compute_assumed_setting(@nospecialize(setting), val::Bool=true)
+    if isexpr(setting, :call) && setting.args[1] === :(!)
+        return compute_assumed_setting(setting.args[2], !val)
+    elseif isa(setting, QuoteNode)
+        return compute_assumed_setting(setting.value, val)
+    else
+        return (setting, val)
+    end
+end
+
 """
     @propagate_inbounds
 

base/fastmath.jl

@@ -279,8 +279,8 @@ exp10_fast(x::Union{Float32,Float64}) = Base.Math.exp10_fast(x)
 
 # builtins
 
-pow_fast(x::Float32, y::Integer) = ccall("llvm.powi.f32", llvmcall, Float32, (Float32, Int32), x, y)
-pow_fast(x::Float64, y::Integer) = ccall("llvm.powi.f64", llvmcall, Float64, (Float64, Int32), x, y)
+pow_fast(x::Float32, y::Integer) = ccall("llvm.powi.f32.i32", llvmcall, Float32, (Float32, Int32), x, y)
+pow_fast(x::Float64, y::Integer) = ccall("llvm.powi.f64.i32", llvmcall, Float64, (Float64, Int32), x, y)
 pow_fast(x::FloatTypes, ::Val{p}) where {p} = pow_fast(x, p) # inlines already via llvm.powi
 @inline pow_fast(x, v::Val) = Base.literal_pow(^, x, v)