求助，使用GPU加速时报错error700：Illegal address during kernel execution

吴宗霖

使用超算CPU分区时计算一直没有问题，但是计算速度慢，就开了一个GPU分区的账户。在GPU分区计算两三百原子的体系时没有出现过问题。但是后面我模拟电解液分子AIMD的时候（格子大概是12埃的立方体），发现超过600原子的时候，就会报错Failing in Thread:1
call to cuStreamSynchronize returned error 700: Illegal address during kernel execution

Failing in Thread:1
call to cuStreamSynchronize returned error 700: Illegal address during kernel execution

--------------------------------------------------------------------------
Primary job  terminated normally, but 1 process returned
a non-zero exit code. Per user-direction, the job has been aborted.
--------------------------------------------------------------------------
--------------------------------------------------------------------------
mpirun detected that one or more processes exited with non-zero status, thus causing
the job to be terminated. The first process to do so was:

  Process name: [[5902,1],1]
  Exit code:    1
后面我有个界面模型AIMD想算，因为有真空层，z方向上比较大，格子大概是（20*20*35埃），超过450原子就也出现了这个错误。
对这个报错，超算工程师说是VASP6.3.2不支持linux编译（鬼扯），给我的方案是换到VASP6.2.1，尝试了一下，确实是可以算，但是每一步之间都会卡一会输出这样的信息：70327940 A   47132758406240 47132758406240 10000... 47132758406240 2... 2... 47132758406232 47132758406232 -212700068
70327940 A   47132758406240 47132758406240 10000... 47132758406240 2... 2... 47132758406232 47132758406232 -212700068
这样的话GPU加速的速度还不如CPU呢，所以请问上面那个报错是什么原因呢，有什么解决方法。谢谢各位大佬。