Skip to content

Optimization to prevent overflow when handling large activation tensors #1639

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

Closed
wants to merge 1 commit into from
Closed

Optimization to prevent overflow when handling large activation tensors #1639

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 3 additions & 3 deletions csrc/activation_kernels.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,8 +18,8 @@ __global__ void silu_and_mul_kernel(
const int d) {
const int64_t token_idx = blockIdx.x;
for (int64_t idx = threadIdx.x; idx < d; idx += blockDim.x) {
const scalar_t x = __ldg(&input[token_idx * 2 * d + idx]);
const scalar_t y = __ldg(&input[token_idx * 2 * d + d + idx]);
const scalar_t x = __ldg(&input[(int64_t) token_idx * 2 * d + idx]);
const scalar_t y = __ldg(&input[(int64_t) token_idx * 2 * d + d + idx]);
out[token_idx * d + idx] = silu(x) * y;
}
}
Expand Down Expand Up @@ -57,7 +57,7 @@ __global__ void activation_kernel(
const int d) {
const int64_t token_idx = blockIdx.x;
for (int64_t idx = threadIdx.x; idx < d; idx += blockDim.x) {
const scalar_t x = __ldg(&input[token_idx * d + idx]);
const scalar_t x = __ldg(&input[(int64_t) token_idx * d + idx]);
out[token_idx * d + idx] = ACT_FN(x);
}
}
Expand Down