There is a bug in __alloc_gpu (https://github.com/lkoshale/DA_STAR/blob/f17db9039e393f7b67956df51c579790bbd33f23/include/a_star.cu#L72C1-L72C40) rendering all sssp with starting node other than 0 fail to proceed.
Adding gpuErrchk ( cudaMemcpy(d_PQ,PQ,sizeof(unsigned int)*N,cudaMemcpyHostToDevice) ); in __alloc_gpu should fix this issue.
By the way, several cudaFree and free should be added to address device & host memory leaks.
Refer to master...turtleizzy:DA_STAR:master . Several api-breaking features have been added in my fork making it difficult to submit a PR directly from my fork.
There is a bug in
__alloc_gpu(https://github.com/lkoshale/DA_STAR/blob/f17db9039e393f7b67956df51c579790bbd33f23/include/a_star.cu#L72C1-L72C40) rendering all sssp with starting node other than 0 fail to proceed.Adding
gpuErrchk ( cudaMemcpy(d_PQ,PQ,sizeof(unsigned int)*N,cudaMemcpyHostToDevice) );in__alloc_gpushould fix this issue.By the way, several
cudaFreeandfreeshould be added to address device & host memory leaks.Refer to master...turtleizzy:DA_STAR:master . Several api-breaking features have been added in my fork making it difficult to submit a PR directly from my fork.