Skip to content

fix bugs of gpu kernel threading #162

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

Merged
amcadmus merged 2 commits into from
Jan 23, 2020
Merged

fix bugs of gpu kernel threading #162

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ac68881
fix bug of 0 output when sec_a.back() is lager than 1024
Jan 23, 2020
d1fa044
fix bug of 0 output when sec_a.back() is lager than 1024
Jan 23, 2020
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
15 changes: 10 additions & 5 deletions source/op/cuda/descrpt_se_a.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -208,13 +208,15 @@ __global__ void compute_descriptor_se_a (VALUETYPE* descript,
const VALUETYPE* coord,
const VALUETYPE rmin,
const VALUETYPE rmax,
compute_t* sel_a_diff_dev)
compute_t* sel_a_diff_dev,
const int sec_a_size)
{
// <<<nloc, sec_a.back()>>>
const unsigned int idx = blockIdx.x;
const unsigned int idy = threadIdx.x;
const unsigned int idx = blockIdx.y;
const unsigned int idy = blockIdx.x * blockDim.x + threadIdx.x;
const int idx_deriv = idy * 4 * 3; // 4 components time 3 directions
const int idx_value = idy * 4; // 4 components
if (idy >= sec_a_size) {return;}

// else {return;}
VALUETYPE * row_descript = descript + idx * ndescrpt;
Expand Down Expand Up @@ -355,7 +357,9 @@ void DescrptSeALauncher(const VALUETYPE* coord,
);
}

compute_descriptor_se_a<<<nloc, sec_a.back()>>> (
const int nblock_ = (sec_a.back() + LEN -1) / LEN;
dim3 block_grid(nblock_, nloc);
compute_descriptor_se_a<<<block_grid, LEN>>> (
descript,
ndescrpt,
descript_deriv,
Expand All @@ -370,7 +374,8 @@ void DescrptSeALauncher(const VALUETYPE* coord,
coord,
rcut_r_smth,
rcut_r,
sel_a_diff
sel_a_diff,
sec_a.back()
);
////
// res = cudaFree(sec_a_dev); cudaErrcheck(res);
Expand Down
17 changes: 11 additions & 6 deletions source/op/cuda/descrpt_se_r.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -209,14 +209,16 @@ __global__ void compute_descriptor_se_r (VALUETYPE* descript,
const VALUETYPE* coord,
const VALUETYPE rmin,
const VALUETYPE rmax,
compute_t* sel_diff_dev)
compute_t* sel_diff_dev,
const int sec_size)
{
// <<<nloc, sec.back()>>>
const unsigned int idx = blockIdx.x;
const unsigned int idy = threadIdx.x;
const unsigned int idx = blockIdx.y;
const unsigned int idy = blockIdx.x * blockDim.x + threadIdx.x;
const int idx_deriv = idy * 3; // 1 components time 3 directions
const int idx_value = idy; // 1 components

if (idy >= sec_size) {return;}

// else {return;}
VALUETYPE * row_descript = descript + idx * ndescrpt;
VALUETYPE * row_descript_deriv = descript_deriv + idx * descript_deriv_size;
Expand Down Expand Up @@ -324,7 +326,9 @@ void DescrptSeRLauncher(const VALUETYPE* coord,
nei_iter
);
}
compute_descriptor_se_r<<<nloc, sec.back()>>> (
const int nblock_ = (sec.back() + LEN -1) / LEN;
dim3 block_grid(nblock_, nloc);
compute_descriptor_se_r<<<block_grid, LEN>>> (
descript,
ndescrpt,
descript_deriv,
Expand All @@ -339,6 +343,7 @@ void DescrptSeRLauncher(const VALUETYPE* coord,
coord,
rcut_smth,
rcut,
sel_diff
sel_diff,
sec.back()
);
}