Managing memory for multiple LP instances

[zuzhaoye]

Hi GLPK team,

First, thanks for providing such a powerful library! I’m currently using glpk.js in a project and looking for some advice with a memory issue.

1. Context:

• Using GLPK.js in a web application (React)
• Each run involves solving 25 LP instances
• Each instance has approximately 200 decision variables and 200 constraints
• The application runs perfectly for the first 4 runs
• Everytime, on the 5th run, I encounter the following error:

ERROR Uncaught RangeError: WebAssembly.Instance(): Out of memory: Cannot allocate Wasm memory for new instance at handleError (http://localhost:3000/static/js/bundle.js:394110:58) at http://localhost:3000/static/js/bundle.js:394129:7

2. Current Approach: (not working)
I tried to assign the problem instance to null after solving, but it did not work. Here is how my code looks like in general:

import GLPK from 'glpk.js';

export const runOP = async ({ inputs }) => {

  // Extract input values
 (details skipped)

  // Load GLPK
  let glpk = await GLPK();

  // Prepare the model
  let lp = {
    name: 'MyOP',
    objective: {
      direction: glpk.GLP_MIN,
      name: 'cost',
      vars: []
    },
    subjectTo: [],
    vars: {}
  };

  // Add variables
  (details skipped)

  // Solve the problem
  try {
    const result = await glpk.solve(lp);
    // Extract results
    (details skipped)

    return {
      status: status,
      optimalCost: optimalCost,
    };
  } catch (error) {
    console.error('Error solving LP:', error);
    return {
      status: 'Error',
      optimalCost: null,
    };
  } finally {
    // Try to clean up, but did not work
    lp = null;
    glpk = null;
  }
};

3. Question:
Do you have any suggestions for managing memory in this scenario to avoid the out-of-memory error?

4. Additional Information:
Browser: Chrome, 128.0.6613.138
GLPK.js version: 4.0.2
Node.js version: v20.15.1

[jvail]

Hi @zuzhaoye, I am afraid there is no team - just me and my ~10 tiny fingers. If you could put together a working example with a sufficiently large LP then I can give it a try. Maye take an example from here https://github.com/Ivordir/YALPS?