Issues

• Reactant cannot raise an atomic add in a KernelAbstractions kernel (MWE)

  JuliaDecisionFocusedLearning/CoolPDLP.jl#168 · Filter by author gdalle-bot opened on Oct 1, 2026

• Reactant raising miscompiles mod of a negative integer (MWE)

  JuliaDecisionFocusedLearning/CoolPDLP.jl#167 · Filter by author gdalle-bot opened on Oct 1, 2026

• Exporting CoolPDLP

  JuliaDecisionFocusedLearning/CoolPDLP.jl#152 · Filter by author klamike opened on Sep 8, 2026

• Reactant support

  JuliaDecisionFocusedLearning/CoolPDLP.jl#149 · Filter by author gdalle opened on Sep 7, 2026
  6

• Primal weight updates

  Feature

  JuliaDecisionFocusedLearning/CoolPDLP.jl#148 · Filter by author gdalle opened on Sep 7, 2026

• Halpern iteration

  Feature

  JuliaDecisionFocusedLearning/CoolPDLP.jl#147 · Filter by author gdalle opened on Sep 7, 2026

• Feasibility polishing

  Feature

  JuliaDecisionFocusedLearning/CoolPDLP.jl#146 · Filter by author gdalle opened on Sep 7, 2026

• Infeasibility detection

  Feature

  JuliaDecisionFocusedLearning/CoolPDLP.jl#145 · Filter by author gdalle opened on Sep 7, 2026

• PDLP vs JuMP+HiGHS on MIPLIB2017 benchmark LP relaxations (60s limit): convergence audit

  Bug

  JuliaDecisionFocusedLearning/CoolPDLP.jl#140 · Filter by author gdalle-bot opened on Aug 25, 2026

• Batched-mode profiling: a flat 2x launch tax, and a shared restart schedule that costs per-column convergence (10 medium MIPLIB2017 instances, B up to 1000)

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#139 · Filter by author gdalle-bot opened on Aug 25, 2026

• Profiling report: PDLP on the 10 largest MIPLIB2017 benchmark instances (L40S, CuSparseMatrixCSR, Float32)

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#138 · Filter by author gdalle-bot opened on Aug 25, 2026

• PDLP vs JuMP+HiGHS on full Netlib set (30s limit): convergence audit

  Bug

  JuliaDecisionFocusedLearning/CoolPDLP.jl#137 · Filter by author gdalle-bot opened on Aug 25, 2026
  1

• Minor / cosmetic cleanup items (bundled)

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#124 · Filter by author gdalle-bot opened on Aug 13, 2026

• Tutorial requests Float32 termination tolerance that is essentially unreachable

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#123 · Filter by author gdalle-bot opened on Aug 13, 2026

• Untested code paths worth adding coverage for

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#119 · Filter by author gdalle-bot opened on Aug 13, 2026

• Test asserts CPU and GPU results are bitwise different, which is backwards

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#117 · Filter by author gdalle-bot opened on Aug 13, 2026

• test/algorithms/all.jl has a stale no-op filter for Netlib instances

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#116 · Filter by author gdalle-bot opened on Aug 13, 2026

• Default var_names allocates n strings on every MILP construction, unused in the MOI path

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#114 · Filter by author gdalle-bot opened on Aug 13, 2026

• is_feasible computes A * x twice, and the batched version loops column-by-column

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#112 · Filter by author gdalle-bot opened on Aug 13, 2026

• copy!(sol_avg_last, sol_avg) runs every iteration but is only read occasionally

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#111 · Filter by author gdalle-bot opened on Aug 13, 2026

• Known dual inaccuracy in MOI tests is excluded rather than diagnosed

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#110 · Filter by author gdalle-bot opened on Aug 13, 2026

• MOI ResultCount returns 1 even when primal_status == NO_SOLUTION

  Bug

  JuliaDecisionFocusedLearning/CoolPDLP.jl#109 · Filter by author gdalle-bot opened on Aug 13, 2026

• Batched termination is all-or-nothing, wasting iterations on already-converged instances

  Task

  JuliaDecisionFocusedLearning/CoolPDLP.jl#108 · Filter by author gdalle-bot opened on Aug 13, 2026

• No infeasibility or unboundedness detection in the MOI wrapper

  Bug

  JuliaDecisionFocusedLearning/CoolPDLP.jl#107 · Filter by author gdalle-bot opened on Aug 13, 2026

• Base.isapprox(::MILP, ::MILP) silently only works for SparseMatrixCSC

  Bug

  JuliaDecisionFocusedLearning/CoolPDLP.jl#105 · Filter by author gdalle-bot opened on Aug 13, 2026