Skip to content

Reduce allocation in IDRS inner loop #187

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
merged 1 commit into from
Jan 4, 2018
Merged

Reduce allocation in IDRS inner loop #187

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
14 changes: 6 additions & 8 deletions src/idrs.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,7 +45,7 @@ function idrs!(x, A, b;
history = ConvergenceHistory(partial=!log)
history[:tol] = tol
reserve!(history,:resnorm, maxiter)
idrs_method!(history, x, linsys_op, (A,), b, s, tol, maxiter; kwargs...)
idrs_method!(history, x, A, b, s, tol, maxiter; kwargs...)
log && shrink!(history)
log ? (x, history) : x
end
Expand All @@ -67,14 +67,12 @@ end
om
end

@inline linsys_op(x, A) = A*x

function idrs_method!(log::ConvergenceHistory, X, op, args, C::T,
function idrs_method!(log::ConvergenceHistory, X, A, C::T,
s::Number, tol::Number, maxiter::Number; smoothing::Bool=false, verbose::Bool=false
) where {T}

verbose && @printf("=== idrs ===\n%4s\t%7s\n","iter","resnorm")
R = C - op(X, args...)::T
R = C - A*X
normR = vecnorm(R)
iter = 1

Expand Down Expand Up @@ -123,7 +121,7 @@ function idrs_method!(log::ConvergenceHistory, X, op, args, C::T,
V .= R .- V

U[k] .= Q .+ om .* V
G[k] = op(U[k], args...)
A_mul_B!(G[k], A, U[k])

# Bi-orthogonalise the new basis vectors

Expand Down Expand Up @@ -164,15 +162,15 @@ function idrs_method!(log::ConvergenceHistory, X, op, args, C::T,
return X
end
if k < s
f[k+1:s] = f[k+1:s] - beta*M[k+1:s,k]
f[k+1:s] .-= beta*M[k+1:s,k]
end
iter += 1
end

# Now we have sufficient vectors in G_j to compute residual in G_j+1
# Note: r is already perpendicular to P so v = r
copy!(V, R)
Q = op(V, args...)::T
A_mul_B!(Q, A, V)
om = omega(Q, R)
R .-= om .* Q
X .+= om .* V
Expand Down