[mpich-discuss] _get_addr error while running application using MPICH

Zhou, Hui zhouh at anl.gov
Tue Nov 20 08:06:14 CST 2018


Hi Zhifeng,

I searched the mpich-3.2.1 codebase and there is no `get_addr` symbol, 
so it appears it is a function in your codebase. Try search that see 
if it yields any clues. The error is during MPI_INIT, so try identify 
the codes before MPI_INIT, hopefully (as it should), it comes quite 
early. Then you may try a smaller code that copies the code up to 
MPI_INIT and see if you can reproduce the errror. In the case that 
MPI_INIT is burried in your codebase, could you try put your code in a 
debugger and obtain a backtrace at the point of MPI_INIT? 

Also in your original email, was there some texts inside those 
eclipses?

>> MPIR_Init_thread(784).....:
>> MPID_Init(1323)...........: channel initialization failed
>> MPIDI_CH3_Init(120).......:
>> MPID_nem_init_ckpt(852)...:
>> MPIDI_CH3I_Seg_commit(364): PMI_Barrier returned -1

-- 
Hui Zhou

On Mon, Nov 19, 2018 at 01:49:04PM -0500, Zhifeng Yang wrote:
>Hi Hui,
>
>I just searched the whole code. There is no MPI_T_* name in the code. I may
>tried the newer version later on. Thank you very much
>
>Zhifeng
>



More information about the discuss mailing list