Chebyshev rangefinder #4
Conversation
src/rangefinder.jl
Outdated
| for i in 2:q | ||
| Yi = (4 / ν) * A * A' * Yi_1 - 2 * Yi_2 - Yi_2 | ||
| Y = hcat(Y, Yi) |
There was a problem hiding this comment.
Not sure if repeated use of hcat is bad practice in Julia (as opposed to e.g. allocating all the required space in a big matrix right at the start).
There was a problem hiding this comment.
Where possible, I would allocate at the beginning of the call since Y changes size and then overwrite columns.
Also use non-allocating matrix operations where possible (see https://docs.julialang.org/en/v1/stdlib/LinearAlgebra/#Low-level-matrix-operations).
For example of doing the multiply, see
src/rangefinder.jl
Outdated
| isnothing(Ω) && (Ω = GaussianTestMatrix(m, r)) | ||
|
|
||
| Y0 = qr(Ω.Ω).Q |
There was a problem hiding this comment.
Does this step make the Chebyshev rangefinder incompatible with SSFTTestMatrix?
src/rangefinder.jl
Outdated
| isnothing(Ω) && (Ω = GaussianTestMatrix(m, r)) | ||
|
|
||
| Y0 = qr(Ω.Ω).Q |
src/rangefinder.jl
Outdated
| for i in 2:q | ||
| Yi = (4 / ν) * A * A' * Yi_1 - 2 * Yi_2 - Yi_2 | ||
| Y = hcat(Y, Yi) |
There was a problem hiding this comment.
Where possible, I would allocate at the beginning of the call since Y changes size and then overwrite columns.
Also use non-allocating matrix operations where possible (see https://docs.julialang.org/en/v1/stdlib/LinearAlgebra/#Low-level-matrix-operations).
For example of doing the multiply, see
src/rangefinder.jl
Outdated
| Yi_1 = Y1 | ||
| for i in 2:q | ||
| Yi = (4 / ν) * A * A' * Yi_1 - 2 * Yi_2 - Yi_2 |
…mizedPreconditioners.jl into chebyshev_rangefinder
No description provided.