Imprecise inference when iterating a Pair with constant first type but varying second type. #37610

KristofferC · 2020-09-16T16:01:24Z

julia> a = ("foo" => "bar", "baz" => nothing);

julia> function f(a, i)
           y = iterate(a, i)
           if y !== nothing
               (k, v), st = y
               return k, v
           end
           return y
       end

julia> @code_warntype f(a, 1)

infers to Union{Nothing, Tuple{Union{Nothing, String},Union{Nothing, String}}} In particular, k is inferred as k::Union{Nothing, String} but it could maybe be inferrable as a String.

The text was updated successfully, but these errors were encountered:

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

…9305) The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes #37610 - some part of this code was taken from #37637 - this PR is originally supposed to be alternative and more generalized version of #39296

@time

…liaLang#39305) The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

@time

…liaLang#39305) The inference precision of certain functions really relies on constant propagation, but currently constant prop' won't happen when a call signature is union split and so sometimes inference ends up looser return type: e.g. ```julia julia> Base.return_types((Union{Tuple{Int,Nothing},Tuple{Int,Missing}},)) do t a, b = t a # I expected a::Int, but a::Union{Missing,Nothing,Int} end |> first Union{Missing, Nothing, Int64} ``` This PR: - enables constant prop' for each union signatures, by calling `abstract_call_method_with_const_args` just after each `abstract_call_method` - refactor `abstract_call_method_with_const_args` into two separate parts, 1.) heuristics to decide whether to do constant prop', 2.) try constant propagation The added test cases will should showcase the cases where the inference result could be improved by that. --- I've not seen notable regression in latency with this PR. Here is a sample benchmark of the impact of this PR on latency, from which I guess this PR is acceptable ? > build time: master (caeacef) ```bash Sysimage built. Summary: Total ─────── 61.615938 seconds Base: ─────── 26.575732 seconds 43.1313% Stdlibs: ──── 35.038024 seconds 56.8652% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1378/1378 Precompilation complete. Summary: Total ─────── 116.417013 seconds Generation ── 81.077365 seconds 69.6439% Execution ─── 35.339648 seconds 30.3561% LINK usr/lib/julia/sys.dylib ``` > build time: this PR ```bash Stdlibs total ──── 34.077962 seconds Sysimage built. Summary: Total ─────── 61.804573 seconds Base: ─────── 27.724077 seconds 44.8576% Stdlibs: ──── 34.077962 seconds 55.1383% JULIA usr/lib/julia/sys-o.a Generating REPL precompile statements... 30/30 Executing precompile statements... 1362/1362 Precompilation complete. Summary: Total ─────── 111.262672 seconds Generation ── 83.535305 seconds 75.0794% Execution ─── 27.727367 seconds 24.9206% LINK usr/lib/julia/sys.dylib ``` > first time to plot: master (caeacef) ```julia julia> using Plots; @time plot(rand(10,3)) 3.614168 seconds (5.47 M allocations: 324.564 MiB, 5.73% gc time, 53.02% compilation time) ``` > first time to plot: this PR ```julia julia> using Plots; @time plot(rand(10,3)) 3.557919 seconds (5.53 M allocations: 328.812 MiB, 2.89% gc time, 51.94% compilation time) ``` --- - fixes JuliaLang#37610 - some part of this code was taken from JuliaLang#37637 - this PR is originally supposed to be alternative and more generalized version of JuliaLang#39296

KristofferC added the compiler:inference Type inference label Sep 16, 2020

JeffBezanson self-assigned this Sep 16, 2020

JeffBezanson added a commit that referenced this issue Sep 18, 2020

fix #37610, allow constant prop on signatures with unions

7a92cf3

JeffBezanson added a commit that referenced this issue Sep 18, 2020

fix #37610, allow constant prop on signatures with unions

0417b40

JeffBezanson added a commit that referenced this issue Sep 18, 2020

fix #37610, allow constant prop on signatures with unions

e1c55ff

JeffBezanson added a commit that referenced this issue Sep 18, 2020

fix #37610, allow constant prop on signatures with unions

d6756a4

dkarrasch mentioned this issue Sep 18, 2020

fix #37610, allow constant prop on signatures with unions #37637

Closed

JeffBezanson added a commit that referenced this issue Sep 18, 2020

fix #37610, allow constant prop on signatures with unions

f1c9ce7

KristofferC mentioned this issue Jan 18, 2021

inference: enable constant propagation for union-split signatures #39305

Merged

aviatesk self-assigned this Jan 21, 2021

aviatesk mentioned this issue Jan 25, 2021

[feature] inference constant propagation even for union-split methods #33655

Closed

vtjnash closed this as completed in #39305 Mar 10, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Imprecise inference when iterating a Pair with constant first type but varying second type. #37610

Imprecise inference when iterating a Pair with constant first type but varying second type. #37610

KristofferC commented Sep 16, 2020 •

edited

Loading

Imprecise inference when iterating a Pair with constant first type but varying second type. #37610

Imprecise inference when iterating a Pair with constant first type but varying second type. #37610

Comments

KristofferC commented Sep 16, 2020 • edited Loading

KristofferC commented Sep 16, 2020 •

edited

Loading