mirror/darknet
1
0
Fork 0
mirror of https://github.com/pjreddie/darknet.git synced 2023-08-10 21:13:14 +03:00
Code Issues Projects Releases Wiki Activity

smaller gridsize in bias

Browse Source
This commit is contained in:
Joseph Redmon 2015-03-24 18:27:12 -07:00
parent 56b6561ae4
commit e92f7d301c
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/convolutional_kernels.cu
View File

@ -11,16 +11,16 @@ extern "C" {
__global__ void bias_output_kernel(float *output, float *biases, int n, int size) __global__ void bias_output_kernel(float *output, float *biases, int n, int size)
{ {
int offset = blockIdx.x * blockDim.x + threadIdx.x; int offset = blockIdx.x * blockDim.x + threadIdx.x;
int filter = blockIdx.y; int filter = blockIdx.y % n;
int batch = blockIdx.z; int batch = blockIdx.y / n;
if(offset < size) output[(batch*n+filter)*size + offset] = biases[filter]; if(offset < size) output[(batch*n+filter)*size + offset] = biases[filter];
} }
void bias_output_gpu(float *output, float *biases, int batch, int n, int size) void bias_output_gpu(float *output, float *biases, int batch, int n, int size)
{ {
dim3 dimGrid((size-1)/BLOCK + 1, n*batch, 1);
dim3 dimBlock(BLOCK, 1, 1); dim3 dimBlock(BLOCK, 1, 1);
dim3 dimGrid((size-1)/BLOCK + 1, n, batch);
bias_output_kernel<<<dimGrid, dimBlock>>>(output, biases, n, size); bias_output_kernel<<<dimGrid, dimBlock>>>(output, biases, n, size);
check_error(cudaPeekAtLastError()); check_error(cudaPeekAtLastError());