Add direct CUDA resident-state sync path and profiling hooks
This commit is contained in:
@@ -100,12 +100,14 @@ namespace Parallel
|
||||
MyList<gridseg> **combined_dst;
|
||||
int *send_lengths;
|
||||
int *recv_lengths;
|
||||
double **send_bufs;
|
||||
double **recv_bufs;
|
||||
int *send_buf_caps;
|
||||
int *recv_buf_caps;
|
||||
MPI_Request *reqs;
|
||||
MPI_Status *stats;
|
||||
double **send_bufs;
|
||||
double **recv_bufs;
|
||||
int *send_buf_caps;
|
||||
int *recv_buf_caps;
|
||||
unsigned char *send_buf_pinned;
|
||||
unsigned char *recv_buf_pinned;
|
||||
MPI_Request *reqs;
|
||||
MPI_Status *stats;
|
||||
int max_reqs;
|
||||
bool lengths_valid;
|
||||
int *tc_req_node;
|
||||
@@ -116,10 +118,11 @@ namespace Parallel
|
||||
void destroy();
|
||||
};
|
||||
|
||||
void Sync_cached(MyList<Patch> *PatL, MyList<var> *VarList, int Symmetry, SyncCache &cache);
|
||||
void transfer_cached(MyList<gridseg> **src, MyList<gridseg> **dst,
|
||||
MyList<var> *VarList1, MyList<var> *VarList2,
|
||||
int Symmetry, SyncCache &cache);
|
||||
void Sync_cached(MyList<Patch> *PatL, MyList<var> *VarList, int Symmetry, SyncCache &cache);
|
||||
void Sync_ensure_cache(MyList<Patch> *PatL, int Symmetry, SyncCache &cache);
|
||||
void transfer_cached(MyList<gridseg> **src, MyList<gridseg> **dst,
|
||||
MyList<var> *VarList1, MyList<var> *VarList2,
|
||||
int Symmetry, SyncCache &cache);
|
||||
|
||||
struct AsyncSyncState {
|
||||
int req_no;
|
||||
|
||||
Reference in New Issue
Block a user