Project

General

Profile

Bug #1488

GPU manager runs out of memory on talent

Added by Dong Hun Lee over 2 years ago. Updated about 2 years ago.

Status:
Merged
Priority:
High
Category:
GPU Support
Target version:
Start date:
04/07/2017
Due date:
% Done:

90%

Spent time:
Tags:

Description

Fatal CUDA Error out of memory at cuda-hybrid-api.cu:790.
Return value 2 from 'cudaMallocHost(&pinnedChunk, bufSize * numBuffers)'.------------- Processor 0 Exiting: Called CmiAbort ------------
Reason:  Exiting!

[0] Stack Traceback:
  [0:0] CmiAbortHelper+0xb3  [0x5e78df]
  [0:1] CmiAbort+0x2d  [0x5e791a]
  [0:2] cudaErrorDie+0x5f  [0x61e300]
  [0:3] _Z10createPoolPiiR5CkVecI11_bufferPoolE+0x483  [0x61f2ea]
  [0:4] initHybridAPI+0xaa  [0x61ebad]
  [0:5] _Z10_initCharmiPPc+0x650  [0x511c19]
  [0:6]   [0x5e76cd]
  [0:7] ConverseInit+0x324  [0x5e75eb]
  [0:8] main+0x3f  [0x50f980]
  [0:9] __libc_start_main+0xf5  [0x7f9a68c28f45]
  [0:10]   [0x50a9f9]
Charm++ fatal error:
 Exiting!

[0] Stack Traceback:
  [0:0]   [0x5e84fd]
  [0:1] LrtsAbort+0x68  [0x5e7e6a]
  [0:2] CmiAbortHelper+0xbf  [0x5e78eb]
  [0:3] CmiAbort+0x2d  [0x5e791a]
  [0:4] cudaErrorDie+0x5f  [0x61e300]
  [0:5] _Z10createPoolPiiR5CkVecI11_bufferPoolE+0x483  [0x61f2ea]
  [0:6] initHybridAPI+0xaa  [0x61ebad]
  [0:7] _Z10_initCharmiPPc+0x650  [0x511c19]
  [0:8]   [0x5e76cd]
  [0:9] ConverseInit+0x324  [0x5e75eb]
  [0:10] main+0x3f  [0x50f980]
  [0:11] __libc_start_main+0xf5  [0x7f9a68c28f45]
  [0:12]   [0x50a9f9]
Aborted (core dumped)

History

#1 Updated by Michael Robson over 2 years ago

  • % Done changed from 0 to 90
  • Status changed from New to Implemented
  • Category set to GPU Support

#2 Updated by Phil Miller about 2 years ago

  • Priority changed from Normal to High

#3 Updated by Michael Robson about 2 years ago

  • Tags set to openatom

#4 Updated by Ronak Buch about 2 years ago

  • Status changed from Implemented to Merged

Also available in: Atom PDF