portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-10 07:55:35 +00:00
Code Releases Activity

Minor improvements

Browse Source
This commit is contained in:
Peter Boyle 2019-06-04 20:51:48 +01:00
parent 45b15d10d3
commit 24bff6dbe6
1 changed files with 6 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
Grid/stencil/Stencil.h
View File

@ -121,10 +121,10 @@ void Gather_plane_exchange_table(Vector<std::pair<int,int> >& table,const Lattic
struct StencilEntry { struct StencilEntry {
#ifdef GRID_NVCC #ifdef GRID_NVCC
uint64_t _byte_offset; // 8 bytes uint64_t _byte_offset; // 8 bytes
uint32_t _offset; // 8 bytes uint32_t _offset; // 4 bytes
#else #else
uint64_t _byte_offset; // 8 bytes uint64_t _byte_offset; // 8 bytes
uint64_t _offset; // 4 bytes (8 ever required?) uint64_t _offset; // 8 bytes (8 ever required?)
#endif #endif
uint8_t _is_local; // 1 bytes uint8_t _is_local; // 1 bytes
uint8_t _permute; // 1 bytes uint8_t _permute; // 1 bytes
@ -1270,10 +1270,13 @@ public:
PRINTIT(shm_bytes); // X bytes + R bytes PRINTIT(shm_bytes); // X bytes + R bytes
// Double this to include spin projection overhead with 2:1 ratio in wilson // Double this to include spin projection overhead with 2:1 ratio in wilson
auto gatheralltime = gathertime+gathermtime; auto gatheralltime = gathertime+gathermtime;
auto allbytes = comms_bytes+shm_bytes;
std::cout << GridLogMessage << " Stencil SHM " << (shm_bytes)/gatheralltime/1000. << " GB/s per rank"<<std::endl; std::cout << GridLogMessage << " Stencil SHM " << (shm_bytes)/gatheralltime/1000. << " GB/s per rank"<<std::endl;
std::cout << GridLogMessage << " Stencil SHM " << (shm_bytes)/gatheralltime/1000.*NP/NN << " GB/s per node"<<std::endl; std::cout << GridLogMessage << " Stencil SHM " << (shm_bytes)/gatheralltime/1000.*NP/NN << " GB/s per node"<<std::endl;
auto all_bytes = comms_bytes+shm_bytes;
std::cout << GridLogMessage << " Stencil SHM all" << (all_bytes)/gatheralltime/1000. << " GB/s per rank"<<std::endl;
std::cout << GridLogMessage << " Stencil SHM all" << (all_bytes)/gatheralltime/1000.*NP/NN << " GB/s per node"<<std::endl;
auto membytes = (shm_bytes + comms_bytes/2) // read/write auto membytes = (shm_bytes + comms_bytes/2) // read/write
+ (shm_bytes+comms_bytes)/2 * sizeof(vobj)/sizeof(cobj); + (shm_bytes+comms_bytes)/2 * sizeof(vobj)/sizeof(cobj);
std::cout << GridLogMessage << " Stencil SHM mem " << (membytes)/gatheralltime/1000. << " GB/s per rank"<<std::endl; std::cout << GridLogMessage << " Stencil SHM mem " << (membytes)/gatheralltime/1000. << " GB/s per rank"<<std::endl;