Skip to content

optimize index computation #33909

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
merged 5 commits into from
Jul 7, 2021
Merged

optimize index computation #33909

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
673a2ae
optimize index computation
Jul 1, 2021
08ddf3c
Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…
Jul 2, 2021
52e7ef1
fix review
Jul 6, 2021
3bbb80f
Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…
Jul 6, 2021
2332646
Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…
Jul 6, 2021
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
13 changes: 8 additions & 5 deletions paddle/fluid/operators/roll_op.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,13 +36,16 @@ __global__ void RollCudaKernel(const T* input, T* output, int64_t N,
}

int64_t output_idx = idx;
int64_t dim_idx, dim_idx_shift;
int64_t new_dim_idx = 0;

#pragma unroll Rank
#pragma unroll
for (size_t i = 0; i < Rank; i++) {
dim_idx = (idx / strides[i]) % sizes[i];
dim_idx_shift = (dim_idx + shifts[i]) % sizes[i];
output_idx = output_idx + (dim_idx_shift - dim_idx) * strides[i];
new_dim_idx = (idx / strides[i]) % sizes[i] + shifts[i];
if (new_dim_idx >= sizes[i]) {
output_idx += (shifts[i] - sizes[i]) * strides[i];
} else {
output_idx += shifts[i] * strides[i];
}
}
output[output_idx] = input[idx];
}
Expand Down