Embedding bugfix.

This commit is contained in:
laurent
2023-06-27 09:56:19 +01:00
parent 18707891b7
commit 6c4a960b15
2 changed files with 4 additions and 4 deletions

View File

@ -18,13 +18,13 @@ extern "C" __global__ void FN_NAME( \
const size_t *strides = info + num_dims; \
if (is_contiguous(num_dims, dims, strides)) { \
for (unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; i < numel; i += blockDim.x * gridDim.x) { \
memcpy(out + i * h_size, inp + ids[i], h_size); \
memcpy(&out[i * h_size], &inp[ids[i] * h_size], h_size * sizeof(TYPENAME)); \
} \
} \
else { \
for (unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; i < numel; i += blockDim.x * gridDim.x) { \
unsigned strided_i = get_strided_index(i, num_dims, dims, strides); \
memcpy(out + i * h_size, inp + ids[i], h_size); \
memcpy(&out[i * h_size], &inp[ids[strided_i] * h_size], h_size * sizeof(TYPENAME)); \
} \
} \
} \