Skip to content

CUDA: use FP32 arithmetic for conv2d #15683

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 1 commit into from
Aug 30, 2025
Merged

CUDA: use FP32 arithmetic for conv2d #15683

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
14 changes: 4 additions & 10 deletions ggml/src/ggml-cuda/conv2d.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -82,7 +82,7 @@ static __global__ void conv2d_kernel(const float * __restrict__ input,
int64_t n, c_out, out_y, out_x;
Layout::unpack_indices(global_idx, P, n, c_out, out_y, out_x);

T acc = 0;
float acc = 0.0f;

for (int64_t c_in = 0; c_in < P.IC; ++c_in) {
kernel_bounds bounds = calculate_kernel_bounds(out_x, out_y, P);
Expand All @@ -93,21 +93,15 @@ static __global__ void conv2d_kernel(const float * __restrict__ input,
for (int64_t kx = bounds.x_min; kx < bounds.x_max; ++kx) {
const int64_t in_x = calculate_input_coord(out_x, kx, P.ST_X, P.DL_X, P.PD_X);

T input_val;
if (std::is_same<T, half>::value) {
input_val = __float2half(input[Layout::input_index(n, c_in, in_y, in_x, P)]);
} else {
input_val = input[Layout::input_index(n, c_in, in_y, in_x, P)];
}

T kernel_val = kernel[Layout::kernel_index(c_out, c_in, ky, kx, P)];
const float input_val = input[Layout::input_index(n, c_in, in_y, in_x, P)];
const float kernel_val = kernel[Layout::kernel_index(c_out, c_in, ky, kx, P)];
acc += (input_val * kernel_val);
}
}
}

// [N, OC, OH, OW]
output[Layout::output_index(n, c_out, out_y, out_x, P)] = (float) acc;
output[Layout::output_index(n, c_out, out_y, out_x, P)] = acc;
}

template <typename T>
Expand Down
Loading