From: Brian P. <bri...@tu...> - 2004-05-30 16:36:49
|
Wes Bethel wrote: > > While working on bringing the sort-first distributed memory > OpenRM+Chromium demos up to date, I ran into a snag. While running the > parallel isosurfacing demo, I received three instances of this error > message: > CR Warning(wahtahh:4103): crPackAppendBuffer: overflowed the destination! > CR Warning(wahtahh:4103): crPackAppendBuffer: overflowed the destination! > CR Warning(wahtahh:4103): crPackAppendBuffer: overflowed the destination! > p0_4103: p4_error: interrupt SIGSEGV: 11 > > After the third one, the parallel program wedged. > > Poking around a bit, I see the MTU size in the conf file I was using is > set to 1MB. I upped it to 8MB and the error went away. > > The parallel isosurface processes, which use OpenRM for scene graph > storage, are generating a bunch of triangles in each process. Then > OpenRM starts blasting triangles out through the SPU chain. It looks a > buffer was blown somewhere, resulting in catastrophic failure. > > I'm using the CR 1.7 stock release and RH 9. Wes, if you can tell me how to reproduce this with OpenRM I'll look into it. -Brian |