Skip to content

Querying dual of a FixRef is very slow on larger models #17

Closed
#18
Closed
Querying dual of a FixRef is very slow on larger models#17
#18
Assignees
ZedongPeng

Description

@dlcole3
dlcole3
opened on Jan 28, 2026

I am running an LP model with ~250K variables and ~300K constraints as a test case for CuPDLPx. When I tried to query a dual value from the fixed constraint reference on a variable, it took 9 minutes to query a single dual value. I did test querying the dual of a fixed constraint reference on a simpler model that was solved with CuPDLPx, and it worked fine. For reference, the .lp file of the model I am using can be found at the githup repo here (it won't upload directly on the issue, so apologies for the roundabout way of uploading). The following code loads the .lp file and recreates the problem. Any ideas why the dual is not being returned quickly?

using JuMP, CuPDLPx

m = read_from_file((@__DIR__)*"/test_model.lp")

set_optimizer(m, CuPDLPx.Optimizer)
set_optimizer_attribute(m, "verbose", true)
set_optimizer_attribute(m, "l_inf_ruiz_iterations", 0)
set_optimizer_attribute(m, "iteration_limit", 1e5)

optimize!(m)
println(solution_summary(m))

fixed_variables = []
for v in all_variables(m)
    if is_fixed(v)
        push!(fixed_variables, v)
    end
end

println("TRYING TO GET DUAL VALUE")
@time dual(FixRef(fixed_variables[1]))

Activity

  1. ZedongPeng commented on Jan 28, 2026

    @ZedongPeng
    ZedongPeng
    on Jan 28, 2026
    Collaborator

    Hi @dlcole3 — thanks for reporting this. I can reproduce it locally as well.

    Since cuPDLPx returns results on the CPU, this doesn’t seem to be caused by CPU↔GPU transfer overhead.

    My guess is that the bottleneck is in the following function in the MOI_wrapper—the current approach might not be very efficient.

    Hi @blegat and @odow: do you have suggestions or prior experience improving this?

    CuPDLPx.jl/src/MOI_wrapper.jl

    Lines 451 to 459 in bc521d6

    function MOI.get(
    optimizer::Optimizer,
    attr::MOI.ConstraintDual,
    ci::MOI.ConstraintIndex{MOI.ScalarAffineFunction{Float64}},
    )
    MOI.check_result_index_bounds(optimizer, attr)
    row = only(MOI.Utilities.rows(optimizer.sets, ci))
    return unsafe_load(optimizer.result.dual_solution, row)
    end 

    ---------------------------------------------------------------------------------------
                                    cuPDLPx v0.2.5                                    
                        A GPU-Accelerated First-Order LP Solver                        
               (c) Haihao Lu, Massachusetts Institute of Technology, 2025              
---------------------------------------------------------------------------------------
problem: 308001 rows, 234096 columns, 1137952 nonzeros
settings:
  iter_limit         : 100000
  time_limit         : 3600.00 sec
  eps_opt            : 1.0e-04
  eps_feas           : 1.0e-04
  l_inf_ruiz_iter    : 0

Running presolver (PSLP v0.0.4)...
  status          : REDUCED
  presolve time   : 0.261 sec
  reduced problem : 157221 rows, 130861 columns, 684356 nonzeros

Preconditioning
  Pock-Chambolle scaling (alpha=1.0000)
  Bound-objective scaling
---------------------------------------------------------------------------------------
   runtime     |     objective      |   absolute residuals    |   relative residuals    
  iter   time  |  pr obj    du obj  |  pr res  du res   gap   |  pr res  du res   gap   
---------------------------------------------------------------------------------------
     0 0.0e+00 |  0.0e+00   0.0e+00 | 0.0e+00 0.0e+00 0.0e+00 | 0.0e+00 0.0e+00 0.0e+00 
   200 0.0e+00 |  8.1e+02   1.3e+03 | 1.6e+01 7.0e-01 5.1e+02 | 4.9e-03 4.4e-03 2.4e-01 
   400 1.7e-02 |  8.6e+02   9.1e+02 | 6.9e+00 1.6e-01 4.8e+01 | 2.1e-03 1.0e-03 2.7e-02 
   600 2.4e-02 |  8.5e+02   8.5e+02 | 7.3e+00 6.0e-02 2.8e+00 | 2.2e-03 3.7e-04 1.6e-03 
   800 3.1e-02 |  8.3e+02   8.3e+02 | 4.6e+00 4.0e-02 2.7e+00 | 1.4e-03 2.5e-04 1.6e-03 
  1000 3.8e-02 |  8.2e+02   8.3e+02 | 4.1e+00 2.6e-02 2.1e+00 | 1.2e-03 1.6e-04 1.3e-03 
  1200 4.5e-02 |  8.2e+02   8.2e+02 | 3.8e+00 2.2e-02 2.9e+00 | 1.1e-03 1.4e-04 1.8e-03 
  1400 5.2e-02 |  8.1e+02   8.2e+02 | 2.7e+00 1.8e-02 5.2e+00 | 8.2e-04 1.1e-04 3.2e-03 
  1600 5.9e-02 |  8.1e+02   8.1e+02 | 3.2e+00 9.2e-03 1.1e+00 | 9.7e-04 5.7e-05 7.0e-04 
  1800 6.6e-02 |  8.0e+02   8.0e+02 | 2.8e+00 5.9e-03 2.1e+00 | 8.6e-04 3.7e-05 1.3e-03 
  2000 7.2e-02 |  8.0e+02   8.0e+02 | 2.4e+00 4.8e-03 1.6e+00 | 7.2e-04 3.0e-05 1.0e-03 
  2200 7.9e-02 |  8.0e+02   8.0e+02 | 1.7e+00 4.1e-03 7.6e-02 | 5.2e-04 2.5e-05 4.8e-05 
  2400 8.6e-02 |  8.0e+02   8.0e+02 | 2.3e+00 2.8e-03 4.7e-01 | 7.0e-04 1.7e-05 2.9e-04 
  2600 9.3e-02 |  8.0e+02   8.0e+02 | 2.0e+00 2.4e-03 1.0e+00 | 5.9e-04 1.5e-05 6.4e-04 
  2800 1.0e-01 |  8.0e+02   8.0e+02 | 1.5e+00 2.0e-03 1.0e+00 | 4.5e-04 1.3e-05 6.4e-04 
  3000 1.1e-01 |  8.0e+02   8.0e+02 | 1.3e+00 1.8e-03 1.1e-01 | 3.9e-04 1.1e-05 6.9e-05 
  3200 1.1e-01 |  8.0e+02   8.0e+02 | 1.2e+00 1.6e-03 1.5e-01 | 3.7e-04 9.8e-06 9.4e-05 
  3400 1.2e-01 |  8.0e+02   8.0e+02 | 1.1e+00 1.5e-03 3.0e-01 | 3.2e-04 9.3e-06 1.9e-04 
  3600 1.3e-01 |  8.0e+02   8.0e+02 | 9.8e-01 1.3e-03 2.3e-01 | 3.0e-04 8.3e-06 1.4e-04 
  3800 1.3e-01 |  8.0e+02   8.0e+02 | 9.2e-01 8.7e-04 4.0e-01 | 2.8e-04 5.4e-06 2.5e-04 
  4000 1.4e-01 |  8.0e+02   8.0e+02 | 8.9e-01 8.1e-04 5.3e-01 | 2.7e-04 5.0e-06 3.3e-04 
  4200 1.5e-01 |  8.0e+02   8.0e+02 | 9.1e-01 7.5e-04 7.4e-01 | 2.7e-04 4.6e-06 4.6e-04 
  4400 1.5e-01 |  8.0e+02   8.0e+02 | 9.3e-01 6.8e-04 8.3e-01 | 2.8e-04 4.2e-06 5.2e-04 
  4600 1.6e-01 |  8.0e+02   8.0e+02 | 9.3e-01 6.1e-04 9.3e-01 | 2.8e-04 3.8e-06 5.8e-04 
  4800 1.7e-01 |  8.0e+02   8.0e+02 | 9.2e-01 5.3e-04 8.3e-01 | 2.8e-04 3.3e-06 5.2e-04 
  5000 1.7e-01 |  8.0e+02   8.0e+02 | 9.2e-01 5.0e-04 7.6e-01 | 2.8e-04 3.1e-06 4.8e-04 
  5200 1.8e-01 |  8.0e+02   8.0e+02 | 9.2e-01 4.7e-04 8.5e-01 | 2.8e-04 2.9e-06 5.3e-04 
  5400 1.9e-01 |  8.0e+02   8.0e+02 | 9.0e-01 4.1e-04 9.6e-01 | 2.7e-04 2.5e-06 6.0e-04 
  5600 2.0e-01 |  8.0e+02   8.0e+02 | 8.6e-01 3.9e-04 9.7e-01 | 2.6e-04 2.4e-06 6.1e-04 
  5800 2.0e-01 |  8.0e+02   8.0e+02 | 8.0e-01 3.8e-04 8.1e-01 | 2.4e-04 2.4e-06 5.1e-04 
  6000 2.1e-01 |  8.0e+02   8.0e+02 | 7.8e-01 2.7e-04 7.4e-01 | 2.3e-04 1.7e-06 4.7e-04 
  6200 2.2e-01 |  8.0e+02   8.0e+02 | 7.8e-01 2.3e-04 8.4e-01 | 2.4e-04 1.4e-06 5.3e-04 
  6400 2.2e-01 |  8.0e+02   8.0e+02 | 7.7e-01 2.0e-04 9.2e-01 | 2.3e-04 1.3e-06 5.8e-04 
  6600 2.3e-01 |  8.0e+02   8.0e+02 | 7.5e-01 1.7e-04 1.0e+00 | 2.3e-04 1.1e-06 6.3e-04 
  6800 2.4e-01 |  8.0e+02   8.0e+02 | 7.3e-01 1.6e-04 1.0e+00 | 2.2e-04 1.0e-06 6.4e-04 
  7000 2.4e-01 |  8.0e+02   8.0e+02 | 7.1e-01 1.6e-04 1.1e+00 | 2.1e-04 9.8e-07 6.6e-04 
  7200 2.5e-01 |  8.0e+02   8.0e+02 | 7.1e-01 1.5e-04 1.1e+00 | 2.1e-04 9.6e-07 6.6e-04 
  7400 2.6e-01 |  8.0e+02   8.0e+02 | 7.0e-01 1.5e-04 1.0e+00 | 2.1e-04 9.2e-07 6.5e-04 
  7600 2.6e-01 |  8.0e+02   8.0e+02 | 6.9e-01 1.4e-04 1.0e+00 | 2.1e-04 8.7e-07 6.4e-04 
  7800 2.7e-01 |  8.0e+02   8.0e+02 | 6.9e-01 1.3e-04 1.0e+00 | 2.1e-04 8.1e-07 6.4e-04 
  8000 2.8e-01 |  8.0e+02   8.0e+02 | 6.8e-01 1.3e-04 9.6e-01 | 2.0e-04 7.9e-07 6.0e-04 
  8200 2.8e-01 |  8.0e+02   8.0e+02 | 6.7e-01 1.1e-04 9.4e-01 | 2.0e-04 7.0e-07 5.9e-04 
  8400 2.9e-01 |  8.0e+02   8.0e+02 | 6.7e-01 1.1e-04 1.0e+00 | 2.0e-04 6.6e-07 6.4e-04 
  8600 3.0e-01 |  8.0e+02   8.0e+02 | 6.5e-01 1.0e-04 8.8e-01 | 1.9e-04 6.4e-07 5.5e-04 
  8800 3.0e-01 |  8.0e+02   8.0e+02 | 6.3e-01 9.5e-05 8.0e-01 | 1.9e-04 5.9e-07 5.0e-04 
  9000 3.1e-01 |  8.0e+02   8.0e+02 | 6.3e-01 9.3e-05 8.4e-01 | 1.9e-04 5.8e-07 5.3e-04 
  9200 3.2e-01 |  8.0e+02   8.0e+02 | 6.3e-01 8.7e-05 9.3e-01 | 1.9e-04 5.4e-07 5.8e-04 
  9400 3.2e-01 |  8.0e+02   8.0e+02 | 6.3e-01 7.0e-05 9.9e-01 | 1.9e-04 4.4e-07 6.2e-04 
  9600 3.3e-01 |  8.0e+02   8.0e+02 | 6.2e-01 7.0e-05 9.5e-01 | 1.9e-04 4.3e-07 6.0e-04 
  9800 3.4e-01 |  8.0e+02   8.0e+02 | 6.1e-01 6.6e-05 9.3e-01 | 1.8e-04 4.1e-07 5.8e-04 
 10000 3.5e-01 |  8.0e+02   8.0e+02 | 6.1e-01 6.3e-05 9.1e-01 | 1.8e-04 3.9e-07 5.7e-04 
 11000 3.8e-01 |  8.0e+02   8.0e+02 | 5.9e-01 3.1e-05 8.9e-01 | 1.8e-04 1.9e-07 5.6e-04 
 12000 4.1e-01 |  8.0e+02   8.0e+02 | 5.7e-01 2.6e-05 8.6e-01 | 1.7e-04 1.6e-07 5.4e-04 
 13000 4.5e-01 |  8.0e+02   8.0e+02 | 5.7e-01 2.5e-05 8.6e-01 | 1.7e-04 1.5e-07 5.4e-04 
 14000 4.8e-01 |  8.0e+02   8.0e+02 | 5.6e-01 2.4e-05 8.3e-01 | 1.7e-04 1.5e-07 5.2e-04 
 15000 5.2e-01 |  8.0e+02   8.0e+02 | 5.4e-01 9.8e-05 7.9e-01 | 1.6e-04 6.1e-07 5.0e-04 
 16000 5.5e-01 |  8.0e+02   8.0e+02 | 4.7e-01 9.4e-05 6.2e-01 | 1.4e-04 5.8e-07 3.9e-04 
 17000 5.8e-01 |  8.0e+02   8.0e+02 | 3.4e-01 9.0e-05 4.8e-01 | 1.0e-04 5.6e-07 3.0e-04 
 18000 6.2e-01 |  8.0e+02   8.0e+02 | 3.1e-01 8.3e-05 3.7e-01 | 9.3e-05 5.1e-07 2.3e-04 
 19000 6.5e-01 |  8.0e+02   8.0e+02 | 3.0e-01 8.0e-05 3.4e-01 | 8.9e-05 5.0e-07 2.1e-04 
 20000 6.8e-01 |  8.0e+02   8.0e+02 | 2.9e-01 7.8e-05 3.1e-01 | 8.6e-05 4.8e-07 1.9e-04 
 21000 7.2e-01 |  8.0e+02   8.0e+02 | 2.5e-01 7.7e-05 2.3e-01 | 7.7e-05 4.8e-07 1.4e-04 
---------------------------------------------------------------------------------------
Solution Summary
  Status                 : OPTIMAL
  Presolve time          : 0.261 sec
  Precondition time      : 0.003004 sec
  Solve time             : 0.753 sec
  Iterations             : 21800
  Primal objective       : 796.3726271
  Dual objective         : 796.3513969
  Objective gap          : 1.332e-05
  Primal infeas          : 6.223e-05
  Dual infeas            : 4.619e-07
solution_summary(; result = 1, verbose = false)
├ solver_name          : CuPDLPx
├ Termination
│ ├ termination_status : OPTIMAL
│ ├ result_count       : 1
│ └ raw_status         : Solver finished
├ Solution (result = 1)
│ ├ primal_status        : FEASIBLE_POINT
│ ├ dual_status          : FEASIBLE_POINT
│ ├ objective_value      : 7.96373e+02
│ └ dual_objective_value : 7.96351e+02
└ Work counters
  └ solve_time (sec)   : 7.52827e-01
TRYING TO GET DUAL VALUE
285.565411 seconds (5.06 M allocations: 268.503 MiB, 0.42% gc time, 0.82% compilation time)
  2. ZedongPeng
    self-assigned this
    on Jan 28, 2026

  3. odow commented on Jan 28, 2026

    @odow
    odow
    on Jan 28, 2026
    Contributor

    I don't understand. Zedong the code you've referenced is for an affine constraint. Do you support returning duals from fixed variables? (VariableIndex-in-EqualTo)?

    And @dlcole3 does your example even return the correct solution?

  4. ZedongPeng commented on Jan 28, 2026

    @ZedongPeng
    ZedongPeng
    on Jan 28, 2026
    Collaborator

    My bad. I didn’t notice it was the reduced cost. cuPDLPx does return the reduced cost in the result, but the current MOI_wrapper doesn’t support it yet.

    Would that explain why it’s so slow right now?

    CuPDLPx.jl/test/MOI_wrapper.jl

    Lines 33 to 56 in bc521d6

    # some implementations not supported yet, such as reduced costs, TimeLimitSec, etc.
    exclude = [r"^test_infeasible_MAX_SENSE$",
    r"^test_infeasible_MAX_SENSE_offset$",
    r"^test_infeasible_MIN_SENSE$",
    r"^test_infeasible_MIN_SENSE_offset$",
    r"^test_infeasible_affine_MAX_SENSE$",
    r"^test_infeasible_affine_MAX_SENSE_offset$",
    r"^test_infeasible_affine_MIN_SENSE$",
    r"^test_infeasible_affine_MIN_SENSE_offset$",
    r"^test_linear_INFEASIBLE$",
    r"^test_linear_INFEASIBLE_2$",
    r"^test_linear_integration_delete_variables$",
    r"^test_solve_DualStatus_INFEASIBILITY_CERTIFICATE_EqualTo_lower$",
    r"^test_solve_DualStatus_INFEASIBILITY_CERTIFICATE_EqualTo_upper$",
    r"^test_solve_DualStatus_INFEASIBILITY_CERTIFICATE_GreaterThan$",
    r"^test_solve_DualStatus_INFEASIBILITY_CERTIFICATE_Interval_lower$",
    r"^test_solve_DualStatus_INFEASIBILITY_CERTIFICATE_Interval_upper$",
    r"^test_solve_DualStatus_INFEASIBILITY_CERTIFICATE_LessThan$",
    r"^test_solve_DualStatus_INFEASIBILITY_CERTIFICATE_VariableIndex_LessThan$",
    r"^test_solve_DualStatus_INFEASIBILITY_CERTIFICATE_VariableIndex_LessThan_max$",
    r"^test_solve_VariableIndex_ConstraintDual_MAX_SENSE$",
    r"^test_solve_VariableIndex_ConstraintDual_MIN_SENSE$",
    r"^test_variable_solve_with_lowerbound$",
    r"^test_variable_solve_with_upperbound$",

  5. odow commented on Jan 28, 2026

    @odow
    odow
    on Jan 28, 2026
    Contributor

    I assume it's probably computing via the fallback, which queries each row, forms the transpose of the matrix, etc.

  6. odow commented on Jan 28, 2026

    @odow
    odow
    on Jan 28, 2026
    Contributor

    I strongly encourage you to fix any tests which are currently skipped. The tests are there for a reason 😄

  7. ZedongPeng commented on Jan 28, 2026

    @ZedongPeng
    ZedongPeng
    on Jan 28, 2026
    Collaborator

    Thank you, @odow. I will fix it.
    Sorry @dlcole3 . You might need to wait until I fix it.

  8. odow commented on Jan 28, 2026

    @odow
    odow
    on Jan 28, 2026
    Contributor

    You just need to implement this method:

    function MOI.get( 
     optimizer::Optimizer, 
     attr::MOI.ConstraintDual, 
     ci::MOI.ConstraintIndex{MOI.VariableIndex,S}, 
 ) where {S<:MOI.AbstractSet}
     MOI.check_result_index_bounds(optimizer, attr) 
     return ???
 end

  9. blegat commented on Jan 29, 2026

    @blegat
    blegat
    on Jan 29, 2026
    Collaborator

    Indeed, it looks like this is going through: https://github.com/jump-dev/MathOptInterface.jl/blob/6c45e7b9f76426b690262d5880c6f3aae62062ba/src/Utilities/results.jl#L440-L449
    which will be quite slow, you can confirm this using @profview. @odow 's suggestion should fix it.

  10. ZedongPeng
    closed this as completedin #18on Jan 30, 2026

  11. ZedongPeng commented on Jan 30, 2026

    @ZedongPeng
    ZedongPeng
    on Jan 30, 2026
    Collaborator

    Hi @dlcole3 . The reduced cost is now supported. You will need to install CuPDLPx.jl from source at this moment.

    ---------------------------------------------------------------------------------------
                                    cuPDLPx v0.2.5                                    
                        A GPU-Accelerated First-Order LP Solver                        
               (c) Haihao Lu, Massachusetts Institute of Technology, 2025              
---------------------------------------------------------------------------------------
problem: 308001 rows, 234096 columns, 1137952 nonzeros
settings:
  iter_limit         : 100000
  time_limit         : 3600.00 sec
  eps_opt            : 1.0e-04
  eps_feas           : 1.0e-04
  l_inf_ruiz_iter    : 0

Running presolver (PSLP v0.0.4)...
  status          : REDUCED
  presolve time   : 0.209 sec
  reduced problem : 157221 rows, 130861 columns, 684356 nonzeros

Preconditioning
  Pock-Chambolle scaling (alpha=1.0000)
  Bound-objective scaling
---------------------------------------------------------------------------------------
   runtime     |     objective      |   absolute residuals    |   relative residuals    
  iter   time  |  pr obj    du obj  |  pr res  du res   gap   |  pr res  du res   gap   
---------------------------------------------------------------------------------------
     0 0.0e+00 |  0.0e+00   0.0e+00 | 0.0e+00 0.0e+00 0.0e+00 | 0.0e+00 0.0e+00 0.0e+00 
   200 0.0e+00 |  8.1e+02   1.3e+03 | 1.6e+01 7.0e-01 5.1e+02 | 4.9e-03 4.4e-03 2.4e-01 
   400 1.5e-02 |  8.6e+02   9.1e+02 | 6.9e+00 1.6e-01 4.8e+01 | 2.1e-03 1.0e-03 2.7e-02 
   600 2.1e-02 |  8.5e+02   8.5e+02 | 7.3e+00 6.0e-02 2.8e+00 | 2.2e-03 3.7e-04 1.6e-03 
   800 2.8e-02 |  8.3e+02   8.3e+02 | 4.6e+00 4.0e-02 2.7e+00 | 1.4e-03 2.5e-04 1.6e-03 
  1000 3.4e-02 |  8.2e+02   8.3e+02 | 4.1e+00 2.6e-02 2.1e+00 | 1.2e-03 1.6e-04 1.3e-03 
  1200 4.1e-02 |  8.2e+02   8.2e+02 | 3.8e+00 2.2e-02 2.9e+00 | 1.1e-03 1.4e-04 1.8e-03 
  1400 4.7e-02 |  8.1e+02   8.2e+02 | 2.7e+00 1.8e-02 5.2e+00 | 8.2e-04 1.1e-04 3.2e-03 
  1600 5.4e-02 |  8.1e+02   8.1e+02 | 3.2e+00 9.2e-03 1.1e+00 | 9.7e-04 5.7e-05 7.0e-04 
  1800 6.0e-02 |  8.0e+02   8.0e+02 | 2.8e+00 5.9e-03 2.1e+00 | 8.6e-04 3.7e-05 1.3e-03 
  2000 6.7e-02 |  8.0e+02   8.0e+02 | 2.4e+00 4.8e-03 1.6e+00 | 7.2e-04 3.0e-05 1.0e-03 
  2200 7.3e-02 |  8.0e+02   8.0e+02 | 1.7e+00 4.1e-03 7.6e-02 | 5.2e-04 2.5e-05 4.8e-05 
  2400 7.9e-02 |  8.0e+02   8.0e+02 | 2.3e+00 2.8e-03 4.7e-01 | 7.0e-04 1.7e-05 2.9e-04 
  2600 8.6e-02 |  8.0e+02   8.0e+02 | 2.0e+00 2.4e-03 1.0e+00 | 5.9e-04 1.5e-05 6.4e-04 
  2800 9.2e-02 |  8.0e+02   8.0e+02 | 1.5e+00 2.0e-03 1.0e+00 | 4.5e-04 1.3e-05 6.4e-04 
  3000 9.9e-02 |  8.0e+02   8.0e+02 | 1.3e+00 1.8e-03 1.1e-01 | 3.9e-04 1.1e-05 6.9e-05 
  3200 1.1e-01 |  8.0e+02   8.0e+02 | 1.2e+00 1.6e-03 1.5e-01 | 3.7e-04 9.8e-06 9.4e-05 
  3400 1.1e-01 |  8.0e+02   8.0e+02 | 1.1e+00 1.5e-03 3.0e-01 | 3.2e-04 9.3e-06 1.9e-04 
  3600 1.2e-01 |  8.0e+02   8.0e+02 | 9.8e-01 1.3e-03 2.3e-01 | 3.0e-04 8.3e-06 1.4e-04 
  3800 1.2e-01 |  8.0e+02   8.0e+02 | 9.2e-01 8.7e-04 4.0e-01 | 2.8e-04 5.4e-06 2.5e-04 
  4000 1.3e-01 |  8.0e+02   8.0e+02 | 8.9e-01 8.1e-04 5.3e-01 | 2.7e-04 5.0e-06 3.3e-04 
  4200 1.4e-01 |  8.0e+02   8.0e+02 | 9.1e-01 7.5e-04 7.4e-01 | 2.7e-04 4.6e-06 4.6e-04 
  4400 1.4e-01 |  8.0e+02   8.0e+02 | 9.3e-01 6.8e-04 8.3e-01 | 2.8e-04 4.2e-06 5.2e-04 
  4600 1.5e-01 |  8.0e+02   8.0e+02 | 9.3e-01 6.1e-04 9.3e-01 | 2.8e-04 3.8e-06 5.8e-04 
  4800 1.6e-01 |  8.0e+02   8.0e+02 | 9.2e-01 5.3e-04 8.3e-01 | 2.8e-04 3.3e-06 5.2e-04 
  5000 1.6e-01 |  8.0e+02   8.0e+02 | 9.2e-01 5.0e-04 7.6e-01 | 2.8e-04 3.1e-06 4.8e-04 
  5200 1.7e-01 |  8.0e+02   8.0e+02 | 9.2e-01 4.7e-04 8.5e-01 | 2.8e-04 2.9e-06 5.3e-04 
  5400 1.8e-01 |  8.0e+02   8.0e+02 | 9.0e-01 4.1e-04 9.6e-01 | 2.7e-04 2.5e-06 6.0e-04 
  5600 1.8e-01 |  8.0e+02   8.0e+02 | 8.6e-01 3.9e-04 9.7e-01 | 2.6e-04 2.4e-06 6.1e-04 
  5800 1.9e-01 |  8.0e+02   8.0e+02 | 8.0e-01 3.8e-04 8.1e-01 | 2.4e-04 2.4e-06 5.1e-04 
  6000 1.9e-01 |  8.0e+02   8.0e+02 | 7.8e-01 2.7e-04 7.4e-01 | 2.3e-04 1.7e-06 4.7e-04 
  6200 2.0e-01 |  8.0e+02   8.0e+02 | 7.8e-01 2.3e-04 8.4e-01 | 2.4e-04 1.4e-06 5.3e-04 
  6400 2.1e-01 |  8.0e+02   8.0e+02 | 7.7e-01 2.0e-04 9.2e-01 | 2.3e-04 1.3e-06 5.8e-04 
  6600 2.1e-01 |  8.0e+02   8.0e+02 | 7.5e-01 1.7e-04 1.0e+00 | 2.3e-04 1.1e-06 6.3e-04 
  6800 2.2e-01 |  8.0e+02   8.0e+02 | 7.3e-01 1.6e-04 1.0e+00 | 2.2e-04 1.0e-06 6.4e-04 
  7000 2.3e-01 |  8.0e+02   8.0e+02 | 7.1e-01 1.6e-04 1.1e+00 | 2.1e-04 9.8e-07 6.6e-04 
  7200 2.3e-01 |  8.0e+02   8.0e+02 | 7.1e-01 1.5e-04 1.1e+00 | 2.1e-04 9.6e-07 6.6e-04 
  7400 2.4e-01 |  8.0e+02   8.0e+02 | 7.0e-01 1.5e-04 1.0e+00 | 2.1e-04 9.2e-07 6.5e-04 
  7600 2.5e-01 |  8.0e+02   8.0e+02 | 6.9e-01 1.4e-04 1.0e+00 | 2.1e-04 8.7e-07 6.4e-04 
  7800 2.5e-01 |  8.0e+02   8.0e+02 | 6.9e-01 1.3e-04 1.0e+00 | 2.1e-04 8.1e-07 6.4e-04 
  8000 2.6e-01 |  8.0e+02   8.0e+02 | 6.8e-01 1.3e-04 9.6e-01 | 2.0e-04 7.9e-07 6.0e-04 
  8200 2.6e-01 |  8.0e+02   8.0e+02 | 6.7e-01 1.1e-04 9.4e-01 | 2.0e-04 7.0e-07 5.9e-04 
  8400 2.7e-01 |  8.0e+02   8.0e+02 | 6.7e-01 1.1e-04 1.0e+00 | 2.0e-04 6.6e-07 6.4e-04 
  8600 2.8e-01 |  8.0e+02   8.0e+02 | 6.5e-01 1.0e-04 8.8e-01 | 1.9e-04 6.4e-07 5.5e-04 
  8800 2.8e-01 |  8.0e+02   8.0e+02 | 6.3e-01 9.5e-05 8.0e-01 | 1.9e-04 5.9e-07 5.0e-04 
  9000 2.9e-01 |  8.0e+02   8.0e+02 | 6.3e-01 9.3e-05 8.4e-01 | 1.9e-04 5.8e-07 5.3e-04 
  9200 3.0e-01 |  8.0e+02   8.0e+02 | 6.3e-01 8.7e-05 9.3e-01 | 1.9e-04 5.4e-07 5.8e-04 
  9400 3.0e-01 |  8.0e+02   8.0e+02 | 6.3e-01 7.0e-05 9.9e-01 | 1.9e-04 4.4e-07 6.2e-04 
  9600 3.1e-01 |  8.0e+02   8.0e+02 | 6.2e-01 7.0e-05 9.5e-01 | 1.9e-04 4.3e-07 6.0e-04 
  9800 3.2e-01 |  8.0e+02   8.0e+02 | 6.1e-01 6.6e-05 9.3e-01 | 1.8e-04 4.1e-07 5.8e-04 
 10000 3.2e-01 |  8.0e+02   8.0e+02 | 6.1e-01 6.3e-05 9.1e-01 | 1.8e-04 3.9e-07 5.7e-04 
 11000 3.5e-01 |  8.0e+02   8.0e+02 | 5.9e-01 3.1e-05 8.9e-01 | 1.8e-04 1.9e-07 5.6e-04 
 12000 3.9e-01 |  8.0e+02   8.0e+02 | 5.7e-01 2.6e-05 8.6e-01 | 1.7e-04 1.6e-07 5.4e-04 
 13000 4.2e-01 |  8.0e+02   8.0e+02 | 5.7e-01 2.5e-05 8.6e-01 | 1.7e-04 1.5e-07 5.4e-04 
 14000 4.5e-01 |  8.0e+02   8.0e+02 | 5.6e-01 2.4e-05 8.3e-01 | 1.7e-04 1.5e-07 5.2e-04 
 15000 4.8e-01 |  8.0e+02   8.0e+02 | 5.4e-01 9.8e-05 7.9e-01 | 1.6e-04 6.1e-07 5.0e-04 
 16000 5.1e-01 |  8.0e+02   8.0e+02 | 4.7e-01 9.4e-05 6.2e-01 | 1.4e-04 5.8e-07 3.9e-04 
 17000 5.5e-01 |  8.0e+02   8.0e+02 | 3.4e-01 9.0e-05 4.8e-01 | 1.0e-04 5.6e-07 3.0e-04 
 18000 5.8e-01 |  8.0e+02   8.0e+02 | 3.1e-01 8.3e-05 3.7e-01 | 9.3e-05 5.1e-07 2.3e-04 
 19000 6.1e-01 |  8.0e+02   8.0e+02 | 3.0e-01 8.0e-05 3.4e-01 | 8.9e-05 5.0e-07 2.1e-04 
 20000 6.4e-01 |  8.0e+02   8.0e+02 | 2.9e-01 7.8e-05 3.1e-01 | 8.6e-05 4.8e-07 1.9e-04 
 21000 6.7e-01 |  8.0e+02   8.0e+02 | 2.5e-01 7.7e-05 2.3e-01 | 7.7e-05 4.8e-07 1.4e-04 
---------------------------------------------------------------------------------------
Solution Summary
  Status                 : OPTIMAL
  Presolve time          : 0.209 sec
  Precondition time      : 0.002739 sec
  Solve time             : 0.705 sec
  Iterations             : 21800
  Primal objective       : 796.3726271
  Dual objective         : 796.3513969
  Objective gap          : 1.332e-05
  Primal infeas          : 6.223e-05
  Dual infeas            : 4.619e-07
solution_summary(; result = 1, verbose = false)
├ solver_name          : CuPDLPx
├ Termination
│ ├ termination_status : OPTIMAL
│ ├ result_count       : 1
│ └ raw_status         : Solver finished
├ Solution (result = 1)
│ ├ primal_status        : FEASIBLE_POINT
│ ├ dual_status          : FEASIBLE_POINT
│ ├ objective_value      : 7.96373e+02
│ └ dual_objective_value : 7.96351e+02
└ Work counters
  └ solve_time (sec)   : 7.05285e-01
TRYING TO GET DUAL VALUE
  0.608581 seconds (2.01 M allocations: 128.601 MiB, 7.86% gc time, 99.98% compilation time)
7.9281443737842645

  12. dlcole3 commented on Feb 2, 2026

    @dlcole3
    dlcole3
    on Feb 2, 2026
    Author

    @ZedongPeng I have updated to use the main branch with the recent PR, and it is working great now. Thank you so much for your quick response in getting this working!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions