Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Revised calcHTranspose operator #68

Merged
merged 16 commits into from
Jul 15, 2024
Merged

Revised calcHTranspose operator #68

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
a4b45fd
Initial version with revised H^t operator
johnmauff Jul 2, 2024
bb4b7c6
Next step in Memory reduction.
johnmauff Jul 2, 2024
9933c80
More explicit definitions of ints.
johnmauff Jul 3, 2024
ac504c2
Code cleanup for new CSR format for H^t matrix
johnmauff Jul 3, 2024
c9a61af
Verified that this works for the hurricane_4panel case.
johnmauff Jul 8, 2024
7b826dd
More bug fixes for new-H^t operator.
johnmauff Jul 9, 2024
5bb4f40
Reduced int length for JH array
johnmauff Jul 9, 2024
32ad127
Change integer size
johnmauff Jul 9, 2024
1bdea2a
Cleanup calculation of nnz
johnmauff Jul 10, 2024
a6e2f5b
Removed remaining old H^t code.
johnmauff Jul 10, 2024
2c22b05
Minor cleanup of the code.
johnmauff Jul 10, 2024
033a6cf
Merge remote-tracking branch 'origin/main' into revised-Ht
johnmauff Jul 10, 2024
720e92c
Cleanup
johnmauff Jul 10, 2024
c4b683b
Clarification on the size of int used in the code.
johnmauff Jul 11, 2024
2e5fb52
Fixed OpenACC directive
johnmauff Jul 11, 2024
1367543
Merge branch 'main' into revised-Ht
sjsprecious Jul 15, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 7 additions & 9 deletions src/CostFunction3D.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -806,8 +806,8 @@ void CostFunction3D::calcInnovation()

void CostFunction3D::calcHTranspose(const real* yhat, real* Astate)
{
uint64_t n,m; // uint64_t
uint64_t j,begin,end; // uint32_t
uint64_t n,m;
uint64_t j,begin,end;
sjsprecious marked this conversation as resolved.
Show resolved Hide resolved
real tmp,val;

#pragma acc data present(yhat,Astate)
Expand Down Expand Up @@ -2578,17 +2578,15 @@ void CostFunction3D::calcHmatrix()
//GPTLstop("CostFunction3D::calcHmatrix:nnz");
IH[mObs] = nnz;

std::cout << "CostFunction3D:: nnz " << nnz << std::endl;
std::cout << "CostFunction3D:: nonzeros " << nnz << std::endl;
std::cout << "Memory usage for [H] (Mbytes): " << sizeof(real)*(nnz)/(1024.0*1024.0) << std::endl;
H = new real[nnz];
JH = new uint32_t [nnz]; // uint32_t
JH = new uint32_t [nnz];
Ht = new real[nnz];
JHt = new uint32_t [nnz]; // uint32_t
JHt = new uint32_t [nnz];
std::cout << "CostFunction3D::calcHmatrix: before big loop" << std::endl;
hi=0;

//#pragma omp parallel for private(m,mi,i,j,k,ii,iis,iie,jj,jjs,jje,kk,kks,kke,ibasis,jbasis,kbasis,iiNode,jjNode,kkNode,iNode,jNode,kNode,var,d,wgt_index,weight,cIndex) //[8.1]
//#pragma acc parallel loop vector gang vector_length(32) private(m,mi,i,j,k,ii,iis,iie,jj,jjs,jje,kk,kks,kke,ibasis,jbasis,kbasis,iiNode,jjNode,kkNode,iNode,jNode,kNode,var,d,wgt_index,weight,cIndex)
for (m = 0; m < mObs; m++) {
IH[m]=hi;
mi = m*(7+varDim*derivDim);
Expand Down Expand Up @@ -2670,8 +2668,8 @@ void CostFunction3D::calcHmatrix()

void CostFunction3D::Htransform(const real* Cstate, real* Hstate)
{
uint64_t i; // uint32_t
uint64_t j,begin,end; // uint64_t
uint64_t i;
uint64_t j,begin,end;
sjsprecious marked this conversation as resolved.
Show resolved Hide resolved
real tmp;

#pragma acc data present(Cstate,Hstate)
Expand Down
8 changes: 4 additions & 4 deletions src/CostFunction3D.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -154,13 +154,13 @@ CostFunction3D(const Projection& proj, const int& numObs = 0, const int& stateSi

// explicitly store the H matrix in CSR format
real *H;
uint64_t *IH; // uint64_t
uint32_t *JH; // uint32_t
uint64_t *IH; // Array with extent:(nState+1) can take on values [0 to nonzeros]
uint32_t *JH; // Array with extent:(nonzeros) can take on values [0 to nState-1]

// explicity store the H^t matrix in CSR format
real *Ht;
uint64_t *IHt; // uint64_t
uint32_t *JHt; // uint32_t
uint64_t *IHt; // Array with extent(mObs+1_ can take on values [0 to nonzeros]
uint32_t *JHt; // Array with extent(nonzeros) can take on values [0 to mObs-1]

int basisappx;
real* basis0;
Expand Down
Loading