Skip to content

Fix compilation error while in a ROCm environment #1628

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
Apr 16, 2022
Merged

Fix compilation error while in a ROCm environment #1628

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
3c1104d
fix compilation error while in a ROCm environment
denghuilu Apr 12, 2022
97c99e6
fix error of variable declaration
denghuilu Apr 13, 2022
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
9 changes: 0 additions & 9 deletions source/lib/src/rocm/tabulate.hip.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -524,15 +524,6 @@ __global__ void tabulate_fusion_se_r_grad_fifth_order_polynomial(
int warp_idx = __shfl(threadIdx.x / 64, 0);
int lane_idx = threadIdx.x % 64;

bool unloop = false;
FPTYPE * iteratorA = (FPTYPE *)&_data[0]; // dy
for (int ii = 0; ii < MTILE; ii++) {
for (int jj = thread_idx; jj < last_layer_size; jj += blockDim.x) {
iteratorA[ii * last_layer_size + jj] = ;
}
}
__syncthreads();
Copy link
Member

@njzjz njzjz Apr 13, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Are you sure that this line should also be removed?

Copy link
Member Author

@denghuilu denghuilu Apr 13, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, there are no shared memory access operations

Copy link
Member

@njzjz njzjz Apr 13, 2022 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I am wondering - I do see this line in CUDA implementation

deepmd-kit/source/lib/src/cuda/tabulate.cu

Line 557 in 97c99e6

__syncthreads();

Copy link
Member Author

@denghuilu denghuilu Apr 14, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I am wondering - I do see this line in CUDA implementation

deepmd-kit/source/lib/src/cuda/tabulate.cu

Line 557 in 97c99e6

__syncthreads();

The __syncthreads function is used to synchronise threads within the entire CUDA block, but these threads do not conflict with each other under the current implementation(There is a conflict in the previous implementation, and it is a waste to keep this __syncthreads), so this function is no longer needed.


for (int ii = 0; ii < nnei; ii += KTILE) {
FPTYPE xx = em[block_idx * nnei + ii + warp_idx];

Expand Down
2 changes: 1 addition & 1 deletion source/lib/tests/test_tabulate_se_r.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -170,7 +170,7 @@ TEST_F(TestTabulateSeR, tabulate_fusion_se_r_grad_gpu_rocm)
std::vector<double> dy_dem(em.size(), 0.0);
std::vector<double> dy(nloc * nnei * last_layer_size, 1.0);

* dy_dem_dev = NULL, * table_dev = NULL, * em_dev = NULL, * dy_dev = NULL;
double * dy_dem_dev = NULL, * table_dev = NULL, * em_dev = NULL, * dy_dev = NULL;
deepmd::malloc_device_memory_sync(dy_dem_dev, dy_dem);
deepmd::malloc_device_memory_sync(table_dev, table);
deepmd::malloc_device_memory_sync(em_dev, em);
Expand Down