API for Custom Solver

Main solver API

The main solver functions are in the table below.

Function	Description
`void set_default_settings(Workspace* work)`	Sets the default settings
`void load_data(Workspace* work)`	Loads the default problem data
`void qoco_custom_solve(Workspace* work)`	Solves the problem with qoco_custom

Settings

The settings are defined in the workspace.h file of qoco_custom and are identical to the QOCO settings which are described in settings. To change the settings, directly modify the settings struct as shown in the code snippet in calling.

Solution

The solution struct is within the workspace struct. To see how to access members fo the solution struct refer to the code snippet in calling (for example work.sol.obj returns the objective value). The members of the solution struct are given below.

Member	Description
`x`	Primal solution
`s`	Slack variable for conic constraints
`y`	Dual solution for equality constraints
`z`	Dual solution for conic constraints
`iters`	Number of iterations
`obj`	Objective value
`pres`	Primal residual
`dres`	Dual residual
`gap`	Duality gap
`status`	Solve status see exitflags

Status Flags

The solution struct contains the solve status flag. The various flags are given in the table below.

Status	Status Code	Value
Solver hasn’t been called	`QOCO_CUSTOM_UNSOLVED`	0
Solved to desired accuracy	`QOCO_CUSTOM_SOLVED`	1
Solved to low accuracy	`QOCO_CUSTOM_SOLVED_INACCURATE`	2
Numerical error or infeasible	`QOCO_CUSTOM_NUMERICAL_ERROR`	3
Iteration limit reached	`QOCO_CUSTOM_MAX_ITER`	4

Changing problem instance

Since the custom solvers are generated for a given problem family, users can solve any problem with identical sparsity patterns. This means that users can change the vector data c, b, and h, as well as the nonzero values in P, A, and G. To change the values, the user should directly modify the data within the Workspace struct as shown in calling.