github
/
darknet
mirror of https://github.com/AlexeyAB/darknet.git
1
0
Fork 0
Code Issues Releases Wiki Activity

Try to fuse conv_xnor+shortcut -> conv_xnor

Browse Source
pull/2383/head
AlexeyAB 6 years ago
parent 9e138adf09
commit 5448e07445
5 changed files with 51 additions and 18 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 2
      include/darknet.h
  2. 6
      src/convolutional_kernels.cu
  3. 3
      src/im2col.h
  4. 43
      src/im2col_kernels.cu
  5. 15
      src/network.c

2
include/darknet.h
Unescape View File

@ -457,6 +457,8 @@ struct layer {
float *binary_input_gpu; float *binary_input_gpu;
float *binary_weights_gpu; float *binary_weights_gpu;
float *bin_conv_shortcut_in_gpu;
float *bin_conv_shortcut_out_gpu;
float * mean_gpu; float * mean_gpu;
float * variance_gpu; float * variance_gpu;

6
src/convolutional_kernels.cu
Unescape View File

@ -300,7 +300,8 @@ void forward_convolutional_layer_gpu(convolutional_layer l, network_state state)
//start_timer(); //start_timer();
gemm_nn_custom_bin_mean_transposed_gpu(m, n, k, gemm_nn_custom_bin_mean_transposed_gpu(m, n, k,
(unsigned char *)l.align_bit_weights_gpu, new_ldb, (unsigned char *)l.transposed_align_workspace_gpu, (unsigned char *)l.align_bit_weights_gpu, new_ldb, (unsigned char *)l.transposed_align_workspace_gpu,
new_ldb, l.output_gpu, n, l.mean_arr_gpu, l.biases_gpu, l.activation == LEAKY); new_ldb, l.output_gpu, n, l.mean_arr_gpu, l.biases_gpu, l.activation == LEAKY,
l.bin_conv_shortcut_in_gpu, l.bin_conv_shortcut_out_gpu);
//cudaDeviceSynchronize(); //cudaDeviceSynchronize();
//stop_timer_and_show_name("gemm_nn_custom_bin_mean_transposed_gpu"); //stop_timer_and_show_name("gemm_nn_custom_bin_mean_transposed_gpu");
@ -370,7 +371,8 @@ void forward_convolutional_layer_gpu(convolutional_layer l, network_state state)
//start_timer(); //start_timer();
gemm_nn_custom_bin_mean_transposed_gpu(m, n, k, gemm_nn_custom_bin_mean_transposed_gpu(m, n, k,
(unsigned char *)l.align_bit_weights_gpu, new_ldb, (unsigned char *)l.transposed_align_workspace_gpu, (unsigned char *)l.align_bit_weights_gpu, new_ldb, (unsigned char *)l.transposed_align_workspace_gpu,
new_ldb, l.output_gpu, n, l.mean_arr_gpu, l.biases_gpu, l.activation == LEAKY); new_ldb, l.output_gpu, n, l.mean_arr_gpu, l.biases_gpu, l.activation == LEAKY,
l.bin_conv_shortcut_in_gpu, l.bin_conv_shortcut_out_gpu);
//cudaDeviceSynchronize(); //cudaDeviceSynchronize();
//stop_timer_and_show_name("gemm_nn_custom_bin_mean_transposed_gpu"); //stop_timer_and_show_name("gemm_nn_custom_bin_mean_transposed_gpu");
//} //}

3
src/im2col.h
Unescape View File

@ -44,7 +44,8 @@ void fill_int8_gpu(unsigned char *src, unsigned char val, size_t size);
void gemm_nn_custom_bin_mean_transposed_gpu(int M, int N, int K, void gemm_nn_custom_bin_mean_transposed_gpu(int M, int N, int K,
unsigned char *A, int lda, unsigned char *A, int lda,
unsigned char *B, int ldb, unsigned char *B, int ldb,
float *C, int ldc, float *mean_arr, float *bias, int leaky_activation); float *C, int ldc, float *mean_arr, float *bias, int leaky_activation,
float *shortcut_in_gpu, float *shortcut_out_gpu);
// sequentially - BAD // sequentially - BAD
void gemm_nn_custom_bin_mean_transposed_sequentially_gpu(int M, int N, int K, void gemm_nn_custom_bin_mean_transposed_sequentially_gpu(int M, int N, int K,

43
src/im2col_kernels.cu
Unescape View File

@ -1247,7 +1247,8 @@ int warpAllReduceSum(int val) {
__global__ void gemm_nn_custom_bin_mean_transposed_tensor_kernel(int M, int N, int K, __global__ void gemm_nn_custom_bin_mean_transposed_tensor_kernel(int M, int N, int K,
unsigned char *A, int lda, unsigned char *A, int lda,
unsigned char *B, int ldb, unsigned char *B, int ldb,
float *C, int ldc, float *mean_arr, float *bias_arr, int leaky_activation) float *C, int ldc, float *mean_arr, float *bias_arr, int leaky_activation,
float *shortcut_in_gpu, float *shortcut_out_gpu)
{ {
// total 57% // total 57%
int index = blockIdx.x*blockDim.x + threadIdx.x; int index = blockIdx.x*blockDim.x + threadIdx.x;
@ -1411,9 +1412,14 @@ __global__ void gemm_nn_custom_bin_mean_transposed_tensor_kernel(int M, int N, i
float bias_val = bias_arr[i + i_d]; float bias_val = bias_arr[i + i_d];
float dst_val = count *mean_val + bias_val; float dst_val = count *mean_val + bias_val;
if (leaky_activation) if (leaky_activation)
dst_val = (dst_val > 0) ? (dst_val) : (0.1f*dst_val); // Leaky activation dst_val = (dst_val >= 0) ? (dst_val) : (0.1f*dst_val); // Leaky activation
size_t out_index = (i + i_d)*ldc + (c_x * 8 + j + j_d);
C[out_index] = dst_val;
C[(i + i_d)*ldc + (c_x*8 + j + j_d)] = dst_val; if (shortcut_out_gpu) {
shortcut_out_gpu[out_index] = shortcut_in_gpu[out_index] + dst_val;
}
} }
} }
@ -1574,7 +1580,8 @@ __global__ void gemm_nn_custom_bin_mean_transposed_tensor_kernel(int M, int N, i
__global__ void gemm_nn_custom_bin_mean_transposed_gpu_kernel(int M, int N, int K, __global__ void gemm_nn_custom_bin_mean_transposed_gpu_kernel(int M, int N, int K,
unsigned char *A, int lda, unsigned char *A, int lda,
unsigned char *B, int ldb, unsigned char *B, int ldb,
float *C, int ldc, float *mean_arr, float *bias_arr, int leaky_activation) float *C, int ldc, float *mean_arr, float *bias_arr, int leaky_activation,
float *shortcut_in_gpu, float *shortcut_out_gpu)
{ {
// total 57% // total 57%
int index = blockIdx.x*blockDim.x + threadIdx.x; int index = blockIdx.x*blockDim.x + threadIdx.x;
@ -1628,7 +1635,7 @@ __global__ void gemm_nn_custom_bin_mean_transposed_gpu_kernel(int M, int N, int
//ulonglong4 a_bit256 = *((ulonglong4 *)(A + A_i)); // weights //ulonglong4 a_bit256 = *((ulonglong4 *)(A + A_i)); // weights
ulonglong4 a_bit256 = *((ulonglong4 *)(A_s + A_i)); // weights ulonglong4 a_bit256 = *((ulonglong4 *)(A_s + A_i)); // weights
ulonglong4 b_bit256 = *((ulonglong4 *)(B + B_i)); // input ulonglong4 b_bit256 = *((ulonglong4 *)(B + B_i)); // input
c_bit256 = xnor_int256(a_bit256, b_bit256); c_bit256 = xor_int256(a_bit256, b_bit256);
int tmp_count = __popcll(c_bit256.w) + __popcll(c_bit256.x) + int tmp_count = __popcll(c_bit256.w) + __popcll(c_bit256.x) +
__popcll(c_bit256.y) + __popcll(c_bit256.z); __popcll(c_bit256.y) + __popcll(c_bit256.z);
@ -1661,7 +1668,7 @@ __global__ void gemm_nn_custom_bin_mean_transposed_gpu_kernel(int M, int N, int
//uint64_t a_bit64 = *((uint64_t *)(A + A_i)); // weights //uint64_t a_bit64 = *((uint64_t *)(A + A_i)); // weights
uint64_t a_bit64 = *((uint64_t *)(A_s + A_i)); // weights uint64_t a_bit64 = *((uint64_t *)(A_s + A_i)); // weights
uint64_t b_bit64 = *((uint64_t *)(B + B_i)); // input uint64_t b_bit64 = *((uint64_t *)(B + B_i)); // input
c_bit64 = xnor_int64(a_bit64, b_bit64); c_bit64 = xor_int64(a_bit64, b_bit64);
int tmp_count = __popcll(c_bit64); int tmp_count = __popcll(c_bit64);
int sum_count = warpAllReduceSum(tmp_count); int sum_count = warpAllReduceSum(tmp_count);
@ -1691,7 +1698,7 @@ __global__ void gemm_nn_custom_bin_mean_transposed_gpu_kernel(int M, int N, int
//uint64_t a_bit64 = *((uint64_t *)(A + A_i)); // weights //uint64_t a_bit64 = *((uint64_t *)(A + A_i)); // weights
uint32_t a_bit32 = *((uint32_t *)(A_s + A_i)); // weights uint32_t a_bit32 = *((uint32_t *)(A_s + A_i)); // weights
uint32_t b_bit32 = *((uint32_t *)(B + B_i)); // input uint32_t b_bit32 = *((uint32_t *)(B + B_i)); // input
uint32_t c_bit32 = xnor_int32(a_bit32, b_bit32); uint32_t c_bit32 = xor_int32(a_bit32, b_bit32);
int tmp_count = __popc(c_bit32); int tmp_count = __popc(c_bit32);
int sum_count = warpAllReduceSum(tmp_count); int sum_count = warpAllReduceSum(tmp_count);
@ -1712,7 +1719,7 @@ __global__ void gemm_nn_custom_bin_mean_transposed_gpu_kernel(int M, int N, int
//ulonglong4 a_bit256 = *((ulonglong4 *)(A + (i*lda + k) / 8)); // weights //ulonglong4 a_bit256 = *((ulonglong4 *)(A + (i*lda + k) / 8)); // weights
ulonglong4 a_bit256 = *((ulonglong4 *)(A_s + (local_i*lda + k) / 8)); // weights ulonglong4 a_bit256 = *((ulonglong4 *)(A_s + (local_i*lda + k) / 8)); // weights
ulonglong4 b_bit256 = *((ulonglong4 *)(B + (j*ldb + k) / 8)); // input ulonglong4 b_bit256 = *((ulonglong4 *)(B + (j*ldb + k) / 8)); // input
ulonglong4 c_bit256 = xnor_int256(a_bit256, b_bit256); ulonglong4 c_bit256 = xor_int256(a_bit256, b_bit256);
count += __popcll(c_bit256.w) + __popcll(c_bit256.x) + count += __popcll(c_bit256.w) + __popcll(c_bit256.x) +
__popcll(c_bit256.y) + __popcll(c_bit256.z); __popcll(c_bit256.y) + __popcll(c_bit256.z);
@ -1724,7 +1731,7 @@ __global__ void gemm_nn_custom_bin_mean_transposed_gpu_kernel(int M, int N, int
//uint64_t a_bit64 = *((uint64_t *)(A + (i*lda + k) / 8)); // weights //uint64_t a_bit64 = *((uint64_t *)(A + (i*lda + k) / 8)); // weights
uint64_t a_bit64 = *((uint64_t *)(A_s + (local_i*lda + k) / 8)); // weights uint64_t a_bit64 = *((uint64_t *)(A_s + (local_i*lda + k) / 8)); // weights
uint64_t b_bit64 = *((uint64_t *)(B + (j*ldb + k) / 8)); // input uint64_t b_bit64 = *((uint64_t *)(B + (j*ldb + k) / 8)); // input
uint64_t c_bit64 = xnor_int64(a_bit64, b_bit64); uint64_t c_bit64 = xor_int64(a_bit64, b_bit64);
count += __popcll(c_bit64); count += __popcll(c_bit64);
} }
@ -1735,8 +1742,13 @@ __global__ void gemm_nn_custom_bin_mean_transposed_gpu_kernel(int M, int N, int
count = count - f1; // remove extra bits (from empty space for align only) count = count - f1; // remove extra bits (from empty space for align only)
float dst_val = (2 * count - K) *mean_val + bias_val; float dst_val = (2 * count - K) *mean_val + bias_val;
if(leaky_activation) if(leaky_activation)
dst_val = (dst_val > 0) ? (dst_val) : (0.1f*dst_val); // Leaky activation dst_val = (dst_val >= 0) ? (dst_val) : (0.1f*dst_val); // Leaky activation
C[i*ldc + j] = dst_val; size_t out_index = i*ldc + j;
C[out_index] = dst_val;
if (shortcut_out_gpu) {
shortcut_out_gpu[out_index] = shortcut_in_gpu[out_index] + dst_val;
}
} }
} }
} }
@ -1762,7 +1774,8 @@ __global__ void gemm_nn_custom_bin_mean_transposed_gpu_kernel(int M, int N, int
void gemm_nn_custom_bin_mean_transposed_gpu(int M, int N, int K, void gemm_nn_custom_bin_mean_transposed_gpu(int M, int N, int K,
unsigned char *A, int lda, unsigned char *A, int lda,
unsigned char *B, int ldb, unsigned char *B, int ldb,
float *C, int ldc, float *mean_arr, float *bias, int leaky_activation) float *C, int ldc, float *mean_arr, float *bias, int leaky_activation,
float *shortcut_in_gpu, float *shortcut_out_gpu)
{ {
int size = M*N; int size = M*N;
const int num_blocks = get_number_of_blocks(size, BLOCK); const int num_blocks = get_number_of_blocks(size, BLOCK);
@ -1794,7 +1807,8 @@ void gemm_nn_custom_bin_mean_transposed_gpu(int M, int N, int K,
A, lda, A, lda,
B, ldb, B, ldb,
C, ldc, C, ldc,
mean_arr, bias, leaky_activation); mean_arr, bias, leaky_activation,
shortcut_in_gpu, shortcut_out_gpu);
//cudaDeviceSynchronize(); //cudaDeviceSynchronize();
//getchar(); //getchar();
@ -1807,7 +1821,8 @@ void gemm_nn_custom_bin_mean_transposed_gpu(int M, int N, int K,
A, lda, A, lda,
B, ldb, B, ldb,
C, ldc, C, ldc,
mean_arr, bias, leaky_activation); mean_arr, bias, leaky_activation,
shortcut_in_gpu, shortcut_out_gpu);
} }
CHECK_CUDA(cudaPeekAtLastError()); CHECK_CUDA(cudaPeekAtLastError());
} }

15
src/network.c
Unescape View File

@ -975,7 +975,7 @@ void fuse_conv_batchnorm(network net)
} }
} }
void forward_blank_layer(layer l, network_state state) {}
void calculate_binary_weights(network net) void calculate_binary_weights(network net)
{ {
@ -996,6 +996,19 @@ void calculate_binary_weights(network net)
if (net.layers[j].use_bin_output) { if (net.layers[j].use_bin_output) {
l->activation = LINEAR; l->activation = LINEAR;
} }
// fuse conv_xnor + shortcut -> conv_xnor
if ((j + 1) < net.n && net.layers[j].type == CONVOLUTIONAL) {
layer *sc = &net.layers[j + 1];
if (sc->type == SHORTCUT && sc->w == sc->out_w && sc->h == sc->out_h && sc->c == sc->out_c)
{
l->bin_conv_shortcut_in_gpu = net.layers[net.layers[j + 1].index].output_gpu;
l->bin_conv_shortcut_out_gpu = net.layers[j + 1].output_gpu;
net.layers[j + 1].type = BLANK;
net.layers[j + 1].forward_gpu = forward_blank_layer;
}
}
} }
} }
} }

Write Preview
Loading…
Cancel
Save