| | |
| | | dim3 dimGrid((size-1)/BLOCK + 1, n, batch); |
| | | dim3 dimBlock(BLOCK, 1, 1); |
| | | |
| | | scale_bias_kernel<<<dimGrid, dimBlock>>>(output, biases, n, size); |
| | | scale_bias_kernel<<<dimGrid, dimBlock, 0, get_cuda_stream()>>>(output, biases, n, size); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | |
| | | dim3 dimGrid((size-1)/BLOCK + 1, n, batch); |
| | | dim3 dimBlock(BLOCK, 1, 1); |
| | | |
| | | add_bias_kernel<<<dimGrid, dimBlock>>>(output, biases, n, size); |
| | | add_bias_kernel<<<dimGrid, dimBlock, 0, get_cuda_stream()>>>(output, biases, n, size); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | |
| | | extern "C" void normalize_gpu(float *x, float *mean, float *variance, int batch, int filters, int spatial) |
| | | { |
| | | size_t N = batch*filters*spatial; |
| | | normalize_kernel<<<cuda_gridsize(N), BLOCK>>>(N, x, mean, variance, batch, filters, spatial); |
| | | normalize_kernel<<<cuda_gridsize(N), BLOCK, 0, get_cuda_stream()>>>(N, x, mean, variance, batch, filters, spatial); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | |
| | | |
| | | extern "C" void fast_mean_gpu(float *x, int batch, int filters, int spatial, float *mean) |
| | | { |
| | | fast_mean_kernel<<<filters, BLOCK>>>(x, batch, filters, spatial, mean); |
| | | fast_mean_kernel<<<filters, BLOCK, 0, get_cuda_stream()>>>(x, batch, filters, spatial, mean); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | | extern "C" void fast_variance_gpu(float *x, float *mean, int batch, int filters, int spatial, float *variance) |
| | | { |
| | | fast_variance_kernel<<<filters, BLOCK>>>(x, mean, batch, filters, spatial, variance); |
| | | fast_variance_kernel<<<filters, BLOCK, 0, get_cuda_stream() >>>(x, mean, batch, filters, spatial, variance); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | |
| | | |
| | | extern "C" void pow_ongpu(int N, float ALPHA, float * X, int INCX, float * Y, int INCY) |
| | | { |
| | | pow_kernel<<<cuda_gridsize(N), BLOCK>>>(N, ALPHA, X, INCX, Y, INCY); |
| | | pow_kernel<<<cuda_gridsize(N), BLOCK, 0, get_cuda_stream() >>>(N, ALPHA, X, INCX, Y, INCY); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | | extern "C" void axpy_ongpu_offset(int N, float ALPHA, float * X, int OFFX, int INCX, float * Y, int OFFY, int INCY) |
| | | { |
| | | axpy_kernel<<<cuda_gridsize(N), BLOCK>>>(N, ALPHA, X, OFFX, INCX, Y, OFFY, INCY); |
| | | axpy_kernel<<<cuda_gridsize(N), BLOCK, 0, get_cuda_stream()>>>(N, ALPHA, X, OFFX, INCX, Y, OFFY, INCY); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | |
| | | |
| | | extern "C" void copy_ongpu_offset(int N, float * X, int OFFX, int INCX, float * Y, int OFFY, int INCY) |
| | | { |
| | | copy_kernel<<<cuda_gridsize(N), BLOCK>>>(N, X, OFFX, INCX, Y, OFFY, INCY); |
| | | copy_kernel<<<cuda_gridsize(N), BLOCK, 0, get_cuda_stream()>>>(N, X, OFFX, INCX, Y, OFFY, INCY); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | |
| | | extern "C" void flatten_ongpu(float *x, int spatial, int layers, int batch, int forward, float *out) |
| | | { |
| | | int size = spatial*batch*layers; |
| | | flatten_kernel<<<cuda_gridsize(size), BLOCK>>>(size, x, spatial, layers, batch, forward, out); |
| | | flatten_kernel<<<cuda_gridsize(size), BLOCK, 0, get_cuda_stream()>>>(size, x, spatial, layers, batch, forward, out); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | | extern "C" void reorg_ongpu(float *x, int w, int h, int c, int batch, int stride, int forward, float *out) |
| | | { |
| | | int size = w*h*c*batch; |
| | | reorg_kernel<<<cuda_gridsize(size), BLOCK>>>(size, x, w, h, c, batch, stride, forward, out); |
| | | reorg_kernel<<<cuda_gridsize(size), BLOCK, 0, get_cuda_stream()>>>(size, x, w, h, c, batch, stride, forward, out); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | | extern "C" void mask_ongpu(int N, float * X, float mask_num, float * mask) |
| | | { |
| | | mask_kernel<<<cuda_gridsize(N), BLOCK>>>(N, X, mask_num, mask); |
| | | mask_kernel<<<cuda_gridsize(N), BLOCK, 0, get_cuda_stream() >>>(N, X, mask_num, mask); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | |
| | | |
| | | extern "C" void scal_ongpu(int N, float ALPHA, float * X, int INCX) |
| | | { |
| | | scal_kernel<<<cuda_gridsize(N), BLOCK>>>(N, ALPHA, X, INCX); |
| | | scal_kernel<<<cuda_gridsize(N), BLOCK, 0, get_cuda_stream()>>>(N, ALPHA, X, INCX); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | |
| | | |
| | | extern "C" void fill_ongpu(int N, float ALPHA, float * X, int INCX) |
| | | { |
| | | fill_kernel<<<cuda_gridsize(N), BLOCK>>>(N, ALPHA, X, INCX); |
| | | fill_kernel<<<cuda_gridsize(N), BLOCK, 0, get_cuda_stream()>>>(N, ALPHA, X, INCX); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |
| | | |
| | |
| | | { |
| | | int inputs = n; |
| | | int batch = groups; |
| | | softmax_kernel<<<cuda_gridsize(batch), BLOCK>>>(inputs, offset, batch, input, temp, output); |
| | | softmax_kernel<<<cuda_gridsize(batch), BLOCK, 0, get_cuda_stream()>>>(inputs, offset, batch, input, temp, output); |
| | | check_error(cudaPeekAtLastError()); |
| | | } |