Segmentation fault in libgurobi130.so (13.0.1) during MIP solve — reproducible at fixed offset, invalid array index

Summary

We are experiencing deterministic segmentation faults inside Gurobi 13.0.1 when solving MIP models. The crash always occurs at the same instruction and offset in libgurobi130.so, suggesting an internal bug rather than application-level misuse. The issue is more frequent under multi-threaded and multi-instance workloads.

Environment

ItemValue

Gurobi version

13.0.1 (gurobipy)

Python

3.12.3

Ubuntu 6.8.0-101-generic, x86_64

CPU

AMD Ryzen 9 9955HX, 32 logical cores

Memory

60 GB total, ~58 GB available

NUMA

Single node

Threads

Default (auto, uses all cores)

Crash Details

Location: libgurobi130.so, offset 0x600ada+2617
Instruction: movsd (%rdi,%r15,8),%xmm5 — r15 holds invalid index causing out-of-bounds access
Reproducibility: Single instance ~3–5 min; two instances in parallel ~40 sec

What We Tried (all still crash)

Threads=8/16, OBBT=0, Method=1 (simplex), gurobi.env threads=8

Questions

Is this known in 13.0.1? Any workaround?
Parameters to avoid this path (Presolve=0, Heuristics=0, etc.)?
Should we open a support ticket with core dump and model?

Vassilios Yfantis

Gurobi Staff

2026年03月19日 08:32

Hi Songping,

I went ahead and created a ticket so we can investigate this further.

Best regards,

Vassilios

songping gao

2026年03月19日 08:41

Thank you. I am currently trying to reproduce this issue and preparing the model export file.

Gwyneth Butera

2026年07月23日 12:59

This issue was investigated and not reproduced. It is likely that the crash is caused by a hardware issue on your machine, e.g., unstable memory. We suggest that you test the memory. A useful way to do this is to create a bootable device (e.g., USB) with memtest86(+), see https://en.wikipedia.org/wiki/MemTest86.