Forward-merge release/26.02 into main#827
Merged
rgsl888prabhu merged 3 commits intomainfrom Feb 6, 2026
Merged
Conversation
## Issue Authors: - Ishika Roy (https://github.com/Iroy30) - Ramakrishnap (https://github.com/rgsl888prabhu) Approvers: - Ramakrishnap (https://github.com/rgsl888prabhu) - James Lamb (https://github.com/jameslamb) URL: #556
Contributor
Author
|
FAILURE - Unable to forward-merge due to an error, manual merge is necessary. Do not use the IMPORTANT: When merging this PR, do not use the auto-merger (i.e. the |
This PR adds cuts to the MIP solver. This includes the following:
1. Add constraints in the form C*x <= d to an LP that has already been solved to optimality (and has basis information).
- The constraints must be violated at the current relaxation solution x^star. That is, C*x^star > d.
- The constraint matrix, rhs, basis, and basis factorization, are all updated to include the additional constraints.
- Dual simplex is started in phase 2 from a dual feasible solution.
2. Remove constraints from an LP that has already been solved to optimality.
- The constraints must have slacks in the basis
- The basis is refactored from scratch
3. Add cut pass loop after solving the root relaxation
4. Add a cut pool to store cuts and select cuts
- We currently score cuts based on distance and orthogonality.
6. Add Mixed Integer Gomory Cuts
- These are computed via a MIR cut on a row of the simplex tableau
7. Add Mixed Integer Rounding (MIR) Cuts
- These are constructed by aggregating rows of the constraint matrix.
8. Add Strong Chvatal-Gomory Cuts
- These are constructed from a row of the tableau matrix and from rows of the constraint matrix.
9. Fixes to Handling of Steepest Edge Norms in Dual Simplex
- Ensure that all basic variables have a positive steepest edge norms
10. Reduced Costs Fixing at Root Node
- These are applied after each cut pass and after strong branching, if a heuristic solution is available.
12. Fix issues in Crossover when solving the dual problem
- We were not correctly populating slack variables when solving the dual. This issue appeared on graph20-80-1rand
14. Fix issue in Crossover when basis became rank-deficient in dual push
15. Fix issues across the code with handling and propagating concurrent halt.
16. New solver options: mip-cut-passes, mip-mixed-integer-gomory-cuts, mip-mir-cuts, mip-strong-chvatal-gomory-cuts, mip-knapsack-cuts, mip-cut-change-threshold, mip-cut-min-orthogonality.
Closes #698, #205
Results from a GH200 A/B test with 64 threads and a 300 second time limit. Further runs needed with larger time limit. Further work is needed to get the full benefit of cuts.
A: cuts PR with --mip-cut-passes=10
B: cuts PR with --mip-cut-passes=0
Geomean MIP GAP A / (B = baseline): 0.97
Geomean Time to Optimal A/B: 0.96
A optimal 45
B optimal 37
A problems with feasible solutions 225
B problems with feasible solutions 224
A wins
drayage-100-23 1.14 8.61
gfd-schedulen180f7d50m30k18 87.85 300.0
neos-827175 12.28 25.86
neos-1171448 28.85 98.2
n5-3 180.62 300.0
neos859080 0.78 1.38
seymour1 38.78 300.0
neos-860300 71.98 90.91
neos-3083819-nubu 30.48 300.0
neos-933966 248.26 300.0
neos-957323 46.36 300.0
neos-960392 50.12 115.68
netdiversion 121.59 300.0
ns1208400 162.15 300.0
nw04 39.89 45.98
piperout-27 8.22 14.09
supportcase7 30.48 300.0
supportcase6 70.22 300.0
uccase12 26.95 41.87
A wins 19
A losses
app1-1 2.3 0.74
cbs-cta 2.21 1.83
irp 17.59 12.94
istanbul-no-cutoff 18.26 13.56
mas76 300.0 20.98
neos-1122047 8.3 6.46
neos-1445765 1.86 1.52
neos-1582420 134.3 93.44
neos-3004026-krka 6.56 1.31
neos8 0.88 0.68
ns1952667 99.84 64.92
piperout-08 10.75 8.23
pk1 224.32 91.84
qap10 134.76 62.82
swath1 60.04 50.3
trento1 300.0 172.28
triptim1 246.65 130.92
A losses 17
Authors:
- Chris Maes (https://github.com/chris-maes)
Approvers:
- Ramakrishnap (https://github.com/rgsl888prabhu)
- Akif ÇÖRDÜK (https://github.com/akifcorduk)
- Alice Boucher (https://github.com/aliceb-nv)
URL: #814
…uts make it infeasible (#831) Fixes an issue on neos-4413714-turia where basis repair was called during dual push, and the slack needed was superbasic. Fixes an issue where cbs-cta was incorrectly classified as infeasible after cuts were added. Authors: - Chris Maes (https://github.com/chris-maes) Approvers: - Alice Boucher (https://github.com/aliceb-nv) URL: #831
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Forward-merge triggered by push to release/26.02 that creates a PR to keep main up-to-date. If this PR is unable to be immediately merged due to conflicts, it will remain open for the team to manually merge. See forward-merger docs for more info.