In step-32, it seems to be that data structure can not be deallocated completely in every local-refine step. In part of refine_mesh class. setup_dofs() is called after constructing new local refinement.
Could we guess with this one for sorting out? On Wed, March 25, 2009 4:06 pm, [email protected] wrote: > Dear All, > > In trying with step-32, I got a error like this one I reported beforehand > > > On Wed, March 18, 2009 10:05 am, Wolfgang Bangerth wrote: >> >>> p1_14900: p4_error: : 14 >>> rm_l_1_14901: (1683.000000) net_send: could not write to fd=5, errno = >>> 32 p0_14862: p4_error: net_recv read: probable EOF on socket: 1 >>> p2_14938: p4_error: : 14 >>> rm_l_2_14939: (1682.839844) net_send: could not write to fd=5, errno = >>> 32 p3_14976: p4_error: net_recv read: probable EOF on socket: 1 >>> rm_l_3_14977: (1682.675781) net_send: could not write to fd=5, errno = >>> 32 p0_14862: (1689.203125) net_send: could not write to fd=4, errno = >>> 32 >>> ------------------------------------------------------- > > > Afterwards, I have tried to run the code which is adapted to my current > research(viscoelastic flow) with local refinement and without > > Both are also with trilinos library with mpi. > > With local refinement code stopped with message above, but without local > refinement ( I use with only uniform grid over the whole of domain) code > runs very well (it does not give a any problem). Two code tests are from > with 4 processors on both two different clusters (Luckily, I can hold two > clusters.) > > And, trying with step-32 code on other cluster gave me a similar running > error message. > > ------------------------------------------------------------------- > MX:comp80:recv req:req status 5:Data truncated due to undersized buffer > type: 4 (recv) > state (0x8): > completed > mcp_handle : 65535 > seg:0x2c3ebd0,4 > match_mask:(0xffffffff_ffffffff) > match_val:(0x00000001_0000000e) > matched_val:(0x00000001_0000000e) > slength=8,rlength=4,accum_len=0 > unexpected=0 > > MX:Aborting > MX:comp80:recv req:req status 5:Data truncated due to undersized buffer > type: 4 (recv) > state (0x8): > completed > mcp_handle : 65535 > seg:0x26171c0,4 > match_mask:(0xffffffff_ffffffff) > match_val:(0x00000001_0000000e) > matched_val:(0x00000001_0000000e) > slength=8,rlength=4,accum_len=0 > ----------------------------------------------------------------------- > > > Doesn't you think that it is from relation local refinement and trilinos > library? Something may be not enough if this problem is from only cluster > environment. > > Anyone could give me a advice would be welcome please. > > Thank you > Sungho Yoon > > _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
