Skip to content

Fix GEMM dispatch for complex-real matmul #1520

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
jishnub merged 3 commits into from
Dec 13, 2025
Merged

Fix GEMM dispatch for complex-real matmul #1520

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
b6ad3d7
Fix GEMM dispatch for complex-real matmul
jishnub Dec 9, 2025
c8bb596
Update outdated flag in `generic_matmatmul_wrapper!`
jishnub Dec 10, 2025
cbdb717
Reorder methods
jishnub Dec 10, 2025
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
18 changes: 8 additions & 10 deletions src/matmul.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -317,7 +317,7 @@ end
BlasFlag.SYRK
elseif (tA_uc == 'C' && tB_uc == 'N') || (tA_uc == 'N' && tB_uc == 'C')
BlasFlag.HERK
else isntc
else
BlasFlag.GEMM
end
else
Expand Down Expand Up @@ -499,7 +499,7 @@ function matmul2x2or3x3_nonzeroalpha!(C, tA, tB, A, B, α::Bool, β)
return false
end

# THE one big BLAS dispatch. This is split into two methods to improve latency
# THE one big BLAS dispatch. This is split into syrk/herk/gemm and symm/hemm/none methods to improve latency
Base.@constprop :aggressive function generic_matmatmul_wrapper!(C::StridedMatrix{T}, tA, tB, A::StridedVecOrMat{T}, B::StridedVecOrMat{T},
α::Number, β::Number, val::BlasFlag.SyrkHerkGemm) where {T<:Number}
mA, nA = lapack_size(tA, A)
Expand All @@ -511,6 +511,12 @@ Base.@constprop :aggressive function generic_matmatmul_wrapper!(C::StridedMatrix
_syrk_herk_gemm_wrapper!(C, tA, tB, A, B, α, β, val)
return C
end

function generic_matmatmul_wrapper!(C::StridedVecOrMat{Complex{T}}, tA, tB, A::StridedVecOrMat{Complex{T}}, B::StridedVecOrMat{T},
α::Number, β::Number, ::Val{BlasFlag.GEMM}) where {T<:BlasReal}
gemm_wrapper!(C, tA, tB, A, B, α, β)
end

Base.@constprop :aggressive function _syrk_herk_gemm_wrapper!(C, tA, tB, A, B, α, β, ::Val{BlasFlag.SYRK})
if A === B
tA_uc = uppercase(tA) # potentially strip a WrapperChar
Expand Down Expand Up @@ -657,14 +663,6 @@ Base.@constprop :aggressive generic_matmatmul!(C::StridedMatrix{T}, tA, tB, A::S
_add::MulAddMul = MulAddMul()) where {T<:BlasFloat} =
generic_matmatmul!(C, tA, tB, A, B, _add.alpha, _add.beta)

function generic_matmatmul_wrapper!(C::StridedVecOrMat{Complex{T}}, tA, tB, A::StridedVecOrMat{Complex{T}}, B::StridedVecOrMat{T},
α::Number, β::Number, ::Val{true}) where {T<:BlasReal}
gemm_wrapper!(C, tA, tB, A, B, α, β)
end
Base.@constprop :aggressive function generic_matmatmul_wrapper!(C::StridedVecOrMat{Complex{T}}, tA, tB, A::StridedVecOrMat{Complex{T}}, B::StridedVecOrMat{T},
Copy link
Member Author

@jishnub jishnub Dec 10, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think this method is ever taken. BlasReal matrices would always use GEMM as the flag, and gemm_wrapper! may delegate to _generic_matmatmul! if necessary.

alpha::Number, beta::Number, ::Val{false}) where {T<:BlasReal}
_generic_matmatmul!(C, wrap(A, tA), wrap(B, tB), alpha, beta)
end
# legacy method
Base.@constprop :aggressive generic_matmatmul!(C::StridedVecOrMat{Complex{T}}, tA, tB, A::StridedVecOrMat{Complex{T}}, B::StridedVecOrMat{T},
_add::MulAddMul = MulAddMul()) where {T<:BlasReal} =
Expand Down