Marek Kaluba

oh, and add `linearsolver=SCS.Direct` to `SCSSolver` kwords, which is ~3 times faster on this problem: ``` SCSSolver(alpha=1.5, acceleration_lookback=2, linearsolver=SCS.Direct) ``` (and You need to run master of `SCS.jl`)

I run SCS for `acceleration_lookback in 1:20` on the same problem: https://gist.github.com/kalmarek/13fcb870bc67caa02ffc1f96b8372b9e At the moment I'm also gathering logs with different alphas as well.

have a look at https://cloud.impan.pl/s/UPgXwH3EjkSkrvI The same problem runs on a grid of two parameters: `alpha` and `acceleration_lookback` (You can download the whole folder, but the computation hasn't finished yet)

in my case setting `OMP_NUM_THREADS` has no effect on the convergence/instability

@bodono, I tried https://github.com/cvxgrp/scs/tree/2.1.0 branch where it seems that You started to address the issue. ```julia using MathProgBase using ConicBenchmarkUtilities using SCS solver = SCSSolver(alpha=1.5, acceleration_lookback=20, linearsolver=SCS.Direct, eps=1e-10) dat =...

@bodono I had a look at the recent changes in 2.1.0 branch (as You can see I'd be very happy to have acceleration working stably ;-). As of `d0faaa079` things...

@bodono 2.1.1 is fantastic ;-) with `acceleration_lookback=50` on the test problem I get: ``` ---------------------------------------------------------------------------- SCS v2.1.1 - Splitting Conic Solver (c) Brendan O'Donoghue, Stanford University, 2012 ---------------------------------------------------------------------------- Lin-sys: sparse-direct,...

* the issue mentioned in https://github.com/cvxgrp/scs/issues/180#issuecomment-1301895062 seems to be solved by #251 * I can not reproduce the original issue anymore (probably solved by #246). I presume this issue can...

@blegat how to turn `MonomialBasis{Monomial{DynamicPolynomials.Commutative{DynamicPolynomials.CreationOrder}, Graded{LexOrder}}, MonomialVector{DynamicPolynomials.Commutative{DynamicPolynomials.CreationOrder}, Graded{LexOrder}}}` into an actual vector?

also something is rather odd here: ```julia basis = SumOfSquares.Certificate.gram_basis(cert.certificate, poly) @info eltype(basis) ``` shows `Any` ?