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

Add hpmv! to BLAS in stdlib/LinearAlgebra #34211

Merged
merged 1 commit into from
Jan 3, 2020
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
1 change: 1 addition & 0 deletions NEWS.md
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,7 @@ Standard library changes

#### LinearAlgebra

* The BLAS submodule now supports the level-2 BLAS subroutine `hpmv!` ([#34211]).

#### Markdown

Expand Down
77 changes: 77 additions & 0 deletions stdlib/LinearAlgebra/src/blas.jl
Original file line number Diff line number Diff line change
Expand Up @@ -28,6 +28,7 @@ export
gemv,
hemv!,
hemv,
hpmv!,
sbmv!,
sbmv,
symv!,
Expand Down Expand Up @@ -823,6 +824,82 @@ Only the [`ul`](@ref stdlib-blas-uplo) triangle of `A` is used.
"""
hemv(ul, A, x)

### hpmv!, (HP) Hermitian packed matrix-vector operation defined as y := alpha*A*x + beta*y.
for (fname, elty) in ((:zhpmv_, :ComplexF64),
(:chpmv_, :ComplexF32))
@eval begin
# SUBROUTINE ZHPMV(UPLO,N,ALPHA,AP,X,INCX,BETA,Y,INCY)
# Y <- ALPHA*AP*X + BETA*Y
# * .. Scalar Arguments ..
# DOUBLE PRECISION ALPHA,BETA
# INTEGER INCX,INCY,N
# CHARACTER UPLO
# * .. Array Arguments ..
# DOUBLE PRECISION A(N,N),X(N),Y(N)
function hpmv!(uplo::AbstractChar,
n::BlasInt,
α::$elty,
AP::Union{Ptr{$elty}, AbstractArray{$elty}},
x::Union{Ptr{$elty}, AbstractArray{$elty}},
incx::Integer,
β::$elty,
y::Union{Ptr{$elty}, AbstractArray{$elty}},
incy::Integer)

ccall((@blasfunc($fname), libblas), Cvoid,
(Ref{UInt8}, # uplo,
Ref{BlasInt}, # n,
Ref{$elty}, # α,
Ptr{$elty}, # AP,
Ptr{$elty}, # x,
Ref{BlasInt}, # incx,
Ref{$elty}, # β,
Ptr{$elty}, # y, output
Ref{BlasInt}), # incy
uplo,
n,
α,
AP,
x,
incx,
β,
y,
incy)
end
end
end

function hpmv!(uplo::AbstractChar,
α::Number, AP::Union{DenseArray{T}, AbstractVector{T}}, x::Union{DenseArray{T}, AbstractVector{T}},
β::Number, y::Union{DenseArray{T}, AbstractVector{T}}) where {T <: BlasComplex}
require_one_based_indexing(AP, x, y)
N = length(x)
if N != length(y)
throw(DimensionMismatch("x has length $(N), but y has length $(length(y))"))
end
if length(AP) < Int64(N*(N+1)/2)
throw(DimensionMismatch("Packed Hermitian matrix A has size smaller than length(x) = $(N)."))
end
GC.@preserve x y AP hpmv!(uplo, N, convert(T, α), AP, pointer(x), BlasInt(stride(x, 1)), convert(T, β), pointer(y), BlasInt(stride(y, 1)))
y
end

"""
hpmv!(uplo, α, AP, x, β, y)

Update vector `y` as `α*AP*x + β*y` where `AP` is a packed Hermitian matrix.
The storage layout for `AP` is described in the reference BLAS module, level-2 BLAS at
<https://www.netlib.org/lapack/explore-html/>.
iolaszlo marked this conversation as resolved.
Show resolved Hide resolved

The scalar inputs `α` and `β` shall be numbers.

The array inputs `x`, `y` and `AP` must be complex one-dimensional julia arrays of the
same type that is either `ComplexF32` or `ComplexF64`.

Return the updated `y`.
"""
hpmv!

### sbmv, (SB) symmetric banded matrix-vector multiplication
for (fname, elty) in ((:dsbmv_,:Float64),
(:ssbmv_,:Float32))
Expand Down
41 changes: 41 additions & 0 deletions stdlib/LinearAlgebra/test/blas.jl
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,7 @@ using LinearAlgebra: BlasReal, BlasComplex
Random.seed!(100)
## BLAS tests - testing the interface code to BLAS routines
@testset for elty in [Float32, Float64, ComplexF32, ComplexF64]

@testset "syr2k!" begin
U = randn(5,2)
V = randn(5,2)
Expand Down Expand Up @@ -200,6 +201,46 @@ Random.seed!(100)
@test_throws DimensionMismatch BLAS.trmm('R','U','N','N',one(elty),triu(Cnn),Cnm)
end

# hpmv!
if elty in (ComplexF32, ComplexF64)
@testset "hpmv!" begin
# Both matrix dimensions n coincide, as we have Hermitian matrices.
# Define the inputs and outputs of hpmv!, y = α*A*x+β*y
α = rand(elty)
M = rand(elty, n, n)
A = (M+M')/elty(2.0)
x = rand(elty, n)
β = rand(elty)
y = rand(elty, n)

y_result_julia = α*A*x+β*y

# Create lower triangular packing of A
AP = typeof(A[1,1])[]
for j in 1:n
for i in j:n
push!(AP, A[i,j])
end
end

y_result_blas_lower = copy(y)
BLAS.hpmv!('L', α, AP, x, β, y_result_blas_lower)
@test y_result_julia≈y_result_blas_lower

# Create upper triangular packing of A
AP = typeof(A[1,1])[]
for j in 1:n
for i in 1:j
push!(AP, A[i,j])
end
end

y_result_blas_upper = copy(y)
BLAS.hpmv!('U', α, AP, x, β, y_result_blas_upper)
@test y_result_julia≈y_result_blas_upper
end
end

#trsm
A = triu(rand(elty,n,n))
B = rand(elty,(n,n))
Expand Down