Speed improvement [🇷 for BE/BA]

posted by PharmCat  – Russia, 2020-08-08 19:27 (2126 d 04:33 ago) – Posting: # 21840
Views: 36,199

❝ 30%...It ain't much but it's honest work. :-)


It is very good boost. I donn't know how matrix multipication works in R, but some improovments can be done if make function A = B'*C*B or A += B'*C*B. At each step of this function there are many intermediate products that can be avoided with direct computation. (Or make it in C)

Example incremantal function:

"""
A' * B * A -> +θ (cache)
A' * B * C -> +β
"""
function mulθβinc!(θ, β, A::AbstractMatrix, B::AbstractMatrix, C::Vector)
    q = size(B, 1)
    p = size(A, 2)
    c = zeros(eltype(B), q)
    for i = 1:p
        c .= 0
        for n = 1:q
            for m = 1:q
                @inbounds c[n] += B[m, n] * A[m, i]
            end
            @inbounds β[i] += c[n] * C[n]
        end
        for n = 1:p
            for m = 1:q
                @inbounds θ[i, n] += A[m, n] * c[m]
            end
        end
    end
    θ, β
end

function mulsimple!(θ, β, A::AbstractMatrix, B::AbstractMatrix, C::Vector)
    AB = A'*B
    θ .+ AB*A, β .+ AB*C
end

θ = zeros(4, 4)
β = zeros(4)

A = [1 3 4 3; 2 3 2 3; 3 4 5 6; 3 4 5 6] B = [1 3 4 3; 2 3 2 3; 3 4 5 6; 3 4 5 6] C = [1, 3, 4 ,3]

julia> @time mulsimple!(θ, β, A, B, C);
0.000007 seconds (10 allocations: 1.172 KiB)

julia> @time mulθβinc!(θ, β, A, B, C);
0.000004 seconds (2 allocations: 144 bytes)

eightfold memory savings

Complete thread:

UA Flag
Activity
 Admin contact
23,653 posts in 4,991 threads, 1,570 registered users;
273 visitors (0 registered, 273 guests [including 47 identified bots]).
Forum time: 00:00 CEST (Europe/Vienna)

I’m all in favor of the democratic principle
that one idiot is as good as one genius, but I draw the line
when someone takes the next step and concludes
that two idiots are better than one genius.    Leo Szilard

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5