In fairly recent past, it was not possible to run desktop systems using vm.overcommit_memory=2 if they used Intel graphics because without the OOM killer, the shrinker would never run. Instead, regular memory allocations would fail eventually. In addition, the i915 Mesa driver assumed malloc would never fail. So eventually, the whole desktop infrastructure would crash with a NULL pointer dereference somewhere in Mesa. Is running with vm.overcommit_memory=2 a supported configuration? I expect that vm.overcommit_memory=2 mode could be an interesting operating mode for OpenCL workloads. Thanks, Florian _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
