Skip to content

Fix @fastmath x^2 inlining regression for Float32 and Float16 #60640

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.

Sign up for GitHub

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

Merged
oscardssmith merged 3 commits into from
Jan 14, 2026
Merged

Fix @fastmath x^2 inlining regression for Float32 and Float16 #60640

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6a90b4b
fastmath:
Yashagarwal9798 Jan 11, 2026
bd9c0e9
Base.IEEEFloat
oscardssmith Jan 12, 2026
1da88f1
fix ccall
oscardssmith Jan 14, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 4 additions & 3 deletions base/fastmath.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -297,12 +297,13 @@ exp10_fast(x::Union{Float32,Float64}) = Base.Math.exp10_fast(x)

# builtins

@inline function pow_fast(x::Float64, y::Integer)
@inline function pow_fast(x::T, y::Integer) where T <: Base.IEEEFloat
z = y % Int32
z == y ? pow_fast(x, z) : x^y
end
pow_fast(x::Float32, y::Integer) = x^y
pow_fast(x::Float64, y::Int32) = ccall("llvm.powi.f64.i32", llvmcall, Float64, (Float64, Int32), x, y)
pow_fast(x::Float16, y::Int32) = ccall("llvm.powi", llvmcall, Float16, (Float16, Int32), x, y)
pow_fast(x::Float32, y::Int32) = ccall("llvm.powi", llvmcall, Float32, (Float32, Int32), x, y)
pow_fast(x::Float64, y::Int32) = ccall("llvm.powi", 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)

Expand Down
25 changes: 25 additions & 0 deletions test/fastmath.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -303,6 +303,31 @@ end
@test @fastmath (1 + 1 / n) ^ 4503599627370496 ≈ ℯ
end

# Test that x^2 is inlined to fmul for all float types (issue #60639)
@testset "pow_fast inlining for literal powers" begin
for T in (Float16, Float32, Float64)
f(x) = @fastmath x^2
llvm = sprint(code_llvm, f, (T,))
# Should be inlined to fmul, not call power_by_squaring
@test occursin("fmul", llvm)
@test !occursin("power_by_squaring", llvm)
end
end

# Test correctness of pow_fast for Float32/Float16 with various exponents (issue #60639)
@testset "pow_fast correctness" begin
for T in (Float16, Float32)
x = T(2.5)
# Exponents that fit in Int32
@test (@fastmath x^2) ≈ x^2
@test (@fastmath x^10) ≈ x^10
@test (@fastmath x^(-3)) ≈ x^(-3)
# Exponents that don't fit in Int32
big_exp = Int64(2)^40
@test (@fastmath x^big_exp) ≈ x^big_exp
end
end

@testset "sincos fall-backs" begin
struct FloatWrapper
inner::Float64
Expand Down