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

gonna change im2col

Browse Source
This commit is contained in:
Joseph Redmon
2015-03-21 12:25:14 -07:00
parent dcb000b553
commit 4af116e996
16 changed files with 153 additions and 78 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
src/convolutional_kernels.cu
View File

@ -56,6 +56,7 @@ extern "C" void backward_bias_gpu(float *bias_updates, float *delta, int batch,
extern "C" void forward_convolutional_layer_gpu(convolutional_layer layer, network_state state)
{
clock_t time = clock();
int i;
int m = layer.n;
int k = layer.size*layer.size*layer.c;
@ -63,15 +64,31 @@ extern "C" void forward_convolutional_layer_gpu(convolutional_layer layer, netwo
convolutional_out_width(layer);
bias_output_gpu(layer.output_gpu, layer.biases_gpu, layer.batch, layer.n, n);
cudaDeviceSynchronize();
printf("bias %f\n", sec(clock() - time));
time = clock();
float imt=0;
float gemt = 0;
for(i = 0; i < layer.batch; ++i){
time = clock();
im2col_ongpu(state.input + i*layer.c*layer.h*layer.w, layer.c, layer.h, layer.w, layer.size, layer.stride, layer.pad, layer.col_image_gpu);
cudaDeviceSynchronize();
imt += sec(clock()-time);
time = clock();
float * a = layer.filters_gpu;
float * b = layer.col_image_gpu;
float * c = layer.output_gpu;
gemm_ongpu(0,0,m,n,k,1.,a,k,b,n,1.,c+i*m*n,n);
cudaDeviceSynchronize();
gemt += sec(clock()-time);
time = clock();
}
activate_array_ongpu(layer.output_gpu, m*n*layer.batch, layer.activation);
cudaDeviceSynchronize();
printf("activate %f\n", sec(clock() - time));
printf("im2col %f\n", imt);
printf("gemm %f\n", gemt);
}
extern "C" void backward_convolutional_layer_gpu(convolutional_layer layer, network_state state)