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.

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

Jump to bottom

Remove use of views for performance, and use scaling + BLAS call #8

Merged
merged 2 commits into from
Jun 14, 2018
Merged

Remove use of views for performance, and use scaling + BLAS call #8

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b1da0e9
Remove use of views for performance, and use scaling + BLAS call
mtanneau Jun 14, 2018
1257156
remove useless include
mtanneau Jun 14, 2018
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
37 changes: 21 additions & 16 deletions src/Cholesky/denseBlockAngular.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -236,19 +236,23 @@ function CpBDBt!(C::StridedMatrix{T}, B::StridedMatrix{T}, d::StridedVector{T})

# compute symmetric rank-k update
# only the upper-triangular part of C is modified
temp = zero(T)
for j=1:m
for l = 1:k
@inbounds temp = d[l] * B[j, l]
if temp == zero(T)
# skip loop if temp is zero
continue
end
for i=1:j
@inbounds C[i, j] = C[i, j] + temp * B[i, l]
end
end
end
# temp = zero(T)
# for j=1:m
# for l = 1:k
# @inbounds temp = d[l] * B[j, l]
# if temp == zero(T)
# # skip loop if temp is zero
# continue
# end
# for i=1:j
# @inbounds C[i, j] = C[i, j] + temp * B[i, l]
# end
# end
# end

# # Linear scaling + BLAS call is more efficient
B_ = B * Diagonal(sqrt.(d))
Base.BLAS.syrk!('U', 'N', 1.0, B_, 1.0, C)

return C
end
Expand All @@ -267,8 +271,9 @@ function cholesky!(
C = zeros(Tv, A.m, A.m)

for r in 1:A.R
# copying ararys uses more memory, but is faster
θ_ = view(θ, A.colptr[r]:(A.colptr[r+1]-1))
A_ = view(A.cols, :, A.colptr[r]:(A.colptr[r+1]-1))
A_ = A.cols[:, A.colptr[r]:(A.colptr[r+1]-1)]
η_ = view(F.η, :, r)

# diagonal elements
Expand Down Expand Up @@ -299,8 +304,8 @@ function cholesky(

for r in 1:A.R
θ_ = view(θ, A.colptr[r]:(A.colptr[r+1]-1))
A_ = view(A.cols, :, A.colptr[r]:(A.colptr[r+1]-1))
η_ = view(η, :, r)
A_ = A.cols[:, A.colptr[r]:(A.colptr[r+1]-1)]
η_ = view(η, :, r) # use view because η is modified in-place

# diagonal elements
d[r] = oneunit(Tv) / sum(θ_)
Expand Down
2 changes: 0 additions & 2 deletions test/runtests.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,8 +5,6 @@ else
using Test
end

include("../src/Tulip.jl")

# write your own tests here
const testdir = dirname(@__FILE__)

Expand Down