portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-04-04 11:15:55 +01:00
Code Releases Activity

Fix CoarsenedMatrix regarding illegal memory accesses

Browse Source 
Need a reference to geom since the lambda copies the this pointer which points to host memory, see
- https://docs.nvidia.com/cuda/cuda-c-programming-guide/#star-this-capture
- https://devblogs.nvidia.com/new-compiler-features-cuda-8/
This commit is contained in:
Daniel Richtmann 2020-08-24 16:54:36 +02:00
parent dd1ba266b2
commit b2087f14c4
No known key found for this signature in database
GPG Key ID: B33C490AF0772057
1 changed files with 8 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
Grid/algorithms/CoarsenedMatrix.h
View File

@ -310,6 +310,8 @@ public:
Stencil.HaloExchange(in,compressor);
autoView( in_v , in, AcceleratorRead);
autoView( out_v , out, AcceleratorWrite);
autoView( Stencil_v , Stencil, AcceleratorRead);
auto& geom_v = geom;
typedef LatticeView<Cobj> Aview;
Vector<Aview> AcceleratorViewContainer;
@ -331,14 +333,14 @@ public:
int ptype;
StencilEntry *SE;
for(int point=0;point<geom.npoint;point++){
for(int point=0;point<geom_v.npoint;point++){
SE=Stencil.GetEntry(ptype,point,ss);
SE=Stencil_v.GetEntry(ptype,point,ss);
if(SE->_is_local) {
nbr = coalescedReadPermute(in_v[SE->_offset],ptype,SE->_permute);
} else {
nbr = coalescedRead(Stencil.CommBuf()[SE->_offset]);
nbr = coalescedRead(Stencil_v.CommBuf()[SE->_offset]);
}
acceleratorSynchronise();
@ -382,6 +384,7 @@ public:
autoView( out_v , out, AcceleratorWrite);
autoView( in_v , in, AcceleratorRead);
autoView( Stencil_v , Stencil, AcceleratorRead);
const int Nsimd = CComplex::Nsimd();
typedef decltype(coalescedRead(in_v[0])) calcVector;
@ -395,12 +398,12 @@ public:
int ptype;
StencilEntry *SE;
SE=Stencil.GetEntry(ptype,point,ss);
SE=Stencil_v.GetEntry(ptype,point,ss);
if(SE->_is_local) {
nbr = coalescedReadPermute(in_v[SE->_offset],ptype,SE->_permute);
} else {
nbr = coalescedRead(Stencil.CommBuf()[SE->_offset]);
nbr = coalescedRead(Stencil_v.CommBuf()[SE->_offset]);
}
acceleratorSynchronise();