From f606b5456e4876da5f90e2902b2dff07516a03dc Mon Sep 17 00:00:00 2001
From: AlexeyAB <alexeyab84@gmail.com>
Date: Wed, 22 Aug 2018 14:52:48 +0000
Subject: [PATCH] XNOR-net 21 FPS on CPU yolov2-tiny.cfg
---
src/gemm.h | 9 +
src/convolutional_layer.c | 30 ++++-
src/gemm.c | 216 ++++++++++++++++++++++++++++++++++++++++--
src/layer.h | 1
4 files changed, 236 insertions(+), 20 deletions(-)
diff --git a/src/convolutional_layer.c b/src/convolutional_layer.c
index e7c8e3f..03aad92 100644
--- a/src/convolutional_layer.c
+++ b/src/convolutional_layer.c
@@ -132,6 +132,7 @@
return most;
}
#endif
+ if(l.xnor) return (size_t)l.bit_align*l.size*l.size*l.c * sizeof(float);
return (size_t)l.out_h*l.out_w*l.size*l.size*l.c*sizeof(float);
}
@@ -305,6 +306,10 @@
if(xnor){
l.binary_weights = calloc(c*n*size*size, sizeof(float));
l.binary_input = calloc(l.inputs*l.batch, sizeof(float));
+
+ int align = 8;
+ int src_align = l.out_h*l.out_w;
+ l.bit_align = src_align + (align - src_align % align);
}
if(batch_normalize){
@@ -622,7 +627,7 @@
}
// further optimizations: im2col_bin() for XNOR, and then transpose_aling_bin()
-size_t binary_transpose_align_input(int k, int n, float *b, char **t_bit_input, size_t ldb_align)
+size_t binary_transpose_align_input(int k, int n, float *b, char **t_bit_input, size_t ldb_align, int bit_align)
{
size_t new_ldb = k + (ldb_align - k%ldb_align); // (k / 8 + 1) * 8;
size_t t_intput_size = new_ldb * n;
@@ -637,12 +642,17 @@
//printf("\n align_weights_size = %d, k = %d, m = %d, lda = %d \n", align_weights_size, k, m, k);
//printf("\n align_bit_weights_size = %d, k = %d, m = %d, new_lda = %d \n", align_bit_weights_size, k, m, new_ldb);
- int blocksize = 64;
- transpose_block_SSE4x4(b, t_input, k, n, n, new_ldb, blocksize);
+ int src_size = k * bit_align;
- //printf("\n blocksize = %d \n", blocksize);
+ float_to_bit(b, t_input, src_size);
- float_to_bit(t_input, *t_bit_input, t_intput_size);
+ // b - [bit_align, k] - [l.bit_align, l.size*l.size*l.c] = src_size
+ // t_input - [bit_align, k] - [n', k]
+ // t_bit_input - [new_ldb, n] - [k', n]
+
+ transpose_bin(t_input, *t_bit_input, k, n, bit_align, new_ldb, 8);
+ //transpose_bin(b, *t_bit_input, k, n, bit_align, new_ldb, 8);
+
free(t_input);
return t_intput_size;
@@ -691,12 +701,16 @@
//if (l.xnor && l.size == 3 && l.stride == 1 && l.pad == 1) {}
//else
// further optimizations: im2col_bin() for XNOR, and then transpose_aling_bin()
- im2col_cpu_custom(state.input, l.c, l.h, l.w, l.size, l.stride, l.pad, b);
+ //im2col_cpu_custom(state.input, l.c, l.h, l.w, l.size, l.stride, l.pad, b);
//gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
//gemm_nn_custom(m, n, k, 1, a, k, b, n, c, n);
if (l.xnor) {
+ //im2col_cpu_custom(state.input, l.c, l.h, l.w, l.size, l.stride, l.pad, b);
+ memset(b, 0, l.bit_align*l.size*l.size*l.c * sizeof(float));
+ im2col_cpu_custom_bin(state.input, l.c, l.h, l.w, l.size, l.stride, l.pad, b, l.bit_align);
+
size_t output_size = l.outputs;
//float *count_output = calloc(output_size, sizeof(float));
//size_t bit_output_size = output_size / 8 + 1;
@@ -790,7 +804,7 @@
int ldb_align = l.lda_align;
size_t new_ldb = k + (ldb_align - k%ldb_align);
char *t_bit_input = NULL;
- size_t t_intput_size = binary_transpose_align_input(k, n, b, &t_bit_input, ldb_align);
+ size_t t_intput_size = binary_transpose_align_input(k, n, b, &t_bit_input, ldb_align, l.bit_align);
//char *t_bit_input = calloc(new_ldb * n, sizeof(char)); // for im2col_cpu_custom_transpose() only
//float_to_bit(t_input, t_bit_input, new_ldb * n); // for im2col_cpu_custom_transpose() only
@@ -825,6 +839,8 @@
//free(mean_arr);
}
else {
+ im2col_cpu_custom(state.input, l.c, l.h, l.w, l.size, l.stride, l.pad, b);
+
gemm(0, 0, m, n, k, 1, a, k, b, n, 1, c, n);
// bit-count to float
}
diff --git a/src/gemm.c b/src/gemm.c
index 7abfe25..6342252 100644
--- a/src/gemm.c
+++ b/src/gemm.c
@@ -6,6 +6,7 @@
#include <stdio.h>
#include <math.h>
#include <float.h>
+#include <string.h>
#if defined(_OPENMP)
#include <omp.h>
@@ -595,7 +596,7 @@
static int max_num_threads = 0;
if (max_num_threads == 0) {
max_num_threads = omp_get_max_threads();
- omp_set_num_threads( max_num_threads / 2);
+ //omp_set_num_threads( max_num_threads / 2);
}
#endif
@@ -1037,27 +1038,212 @@
}
}
-void transpose_8x8_bits(unsigned char A[8], unsigned char B[8], int m, int n)
+
+//From Berkeley Vision's Caffe!
+//https://github.com/BVLC/caffe/blob/master/LICENSE
+void im2col_cpu_custom_bin(float* data_im,
+ int channels, int height, int width,
+ int ksize, int stride, int pad, float* data_col, int bit_align)
+{
+ int c, h, w;
+ int height_col = (height + 2 * pad - ksize) / stride + 1;
+ int width_col = (width + 2 * pad - ksize) / stride + 1;
+ int channels_col = channels * ksize * ksize;
+
+ // optimized version
+ if (height_col == height && width_col == width && stride == 1 && pad == 1 && is_fma_avx2())
+ {
+ __m256i all256_sing1 = _mm256_set_epi32(0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000);
+
+ //int algin = 8;
+ //int ldb = width_col * height_col;
+ //int new_ldb = ldb + (algin - ldb % algin);
+ int new_ldb = bit_align;
+
+ #pragma omp parallel for
+ for (c = 0; c < channels_col; ++c) {
+ int w_offset = c % ksize;
+ int h_offset = (c / ksize) % ksize;
+ int c_im = c / ksize / ksize;
+ for (h = pad; h < height_col - pad; ++h) {
+ for (w = pad; w < width_col - pad - 8; w += 8) {
+ int im_row = h_offset + h - pad;
+ int im_col = w_offset + w - pad;
+ //int col_index = (c * height_col + h) * width_col + w;
+ int col_index = c * new_ldb + h * width_col + w;
+
+ //data_col[col_index] = data_im[im_col + width*(im_row + height*c_im)];
+ __m256 src256 = _mm256_loadu_ps((float *)(&data_im[im_col + width*(im_row + height*c_im)]));
+ _mm256_storeu_ps(&data_col[col_index], src256);
+
+ /*/
+ __m256i src256 = _mm256_loadu_si256((__m256i *)(&data_im[im_col + width*(im_row + height*c_im)]));
+ __m256i result256 = _mm256_and_si256(src256, all256_sing1); // check sign in 8 x 32-bit floats
+
+ uint32_t mask = _mm256_movemask_ps(_mm256_castsi256_ps(result256)); // (val >= 0) ? 0 : 1
+ mask = ~mask; // inverse mask, (val >= 0) ? 1 : 0
+
+ data_col[col_index / 8] = mask; // dst[i / 8] = mask;
+ */
+ }
+
+ for (; w < width_col - pad; ++w) {
+ int im_row = h_offset + h - pad;
+ int im_col = w_offset + w - pad;
+ //int col_index = (c * height_col + h) * width_col + w;
+ int col_index = c * new_ldb + h * width_col + w;
+
+ data_col[col_index] = data_im[im_col + width*(im_row + height*c_im)];
+ float val = data_im[im_col + width*(im_row + height*c_im)];
+ //if(val > 0) set_bit(data_col, col_index);
+ }
+ }
+
+ {
+ w = 0;
+ for (h = 0; h < height_col; ++h) {
+ int im_row = h_offset + h;
+ int im_col = w_offset + w;
+ //int col_index = (c * height_col + h) * width_col + w;
+ int col_index = c * new_ldb + h * width_col + w;
+
+ data_col[col_index] = im2col_get_pixel(data_im, height, width, channels, im_row, im_col, c_im, pad);
+ float val = im2col_get_pixel(data_im, height, width, channels, im_row, im_col, c_im, pad);
+ //if (val > 0) set_bit(data_col, col_index);
+ }
+ }
+
+ {
+ w = width_col - 1;
+ for (h = 0; h < height_col; ++h) {
+ int im_row = h_offset + h;
+ int im_col = w_offset + w;
+ //int col_index = (c * height_col + h) * width_col + w;
+ int col_index = c * new_ldb + h * width_col + w;
+
+ data_col[col_index] = im2col_get_pixel(data_im, height, width, channels, im_row, im_col, c_im, pad);
+ float val = im2col_get_pixel(data_im, height, width, channels, im_row, im_col, c_im, pad);
+ //if (val > 0) set_bit(data_col, col_index);
+ }
+ }
+
+ {
+ h = 0;
+ for (w = 0; w < width_col; ++w) {
+ int im_row = h_offset + h;
+ int im_col = w_offset + w;
+ //int col_index = (c * height_col + h) * width_col + w;
+ int col_index = c * new_ldb + h * width_col + w;
+
+ data_col[col_index] = im2col_get_pixel(data_im, height, width, channels, im_row, im_col, c_im, pad);
+ float val = im2col_get_pixel(data_im, height, width, channels, im_row, im_col, c_im, pad);
+ //if (val > 0) set_bit(data_col, col_index);
+ }
+ }
+
+ {
+ h = height_col - 1;
+ for (w = 0; w < width_col; ++w) {
+ int im_row = h_offset + h;
+ int im_col = w_offset + w;
+ //int col_index = (c * height_col + h) * width_col + w;
+ int col_index = c * new_ldb + h * width_col + w;
+
+ data_col[col_index] = im2col_get_pixel(data_im, height, width, channels, im_row, im_col, c_im, pad);
+ float val = im2col_get_pixel(data_im, height, width, channels, im_row, im_col, c_im, pad);
+ //if (val > 0) set_bit(data_col, col_index);
+ }
+ }
+ }
+
+ }
+ else {
+ printf("\n Error: is no non-optimized version \n");
+ //im2col_cpu(data_im, channels, height, width, ksize, stride, pad, data_col);
+ }
+
+ /*
+ int src_size = bit_align*channels_col;
+ char *bit_arr = calloc(src_size, sizeof(float));
+ float_to_bit(data_col, bit_arr, src_size);
+ memcpy(data_col, bit_arr, src_size * sizeof(float));
+ free(bit_arr);
+ */
+}
+
+
+void transpose_8x8_bits_my(unsigned char *A, unsigned char *B, int lda, int ldb)
+{
+ unsigned x, y, t;
+ for (y = 0; y < 8; ++y) {
+ for (x = 0; x < 8; ++x) {
+ if(A[y * lda] & (1 << x)) B[x * ldb] |= 1 << y;
+ }
+ }
+}
+
+unsigned char reverse_byte_1(char a)
+{
+ return ((a & 0x1) << 7) | ((a & 0x2) << 5) |
+ ((a & 0x4) << 3) | ((a & 0x8) << 1) |
+ ((a & 0x10) >> 1) | ((a & 0x20) >> 3) |
+ ((a & 0x40) >> 5) | ((a & 0x80) >> 7);
+}
+
+unsigned char reverse_byte_2(unsigned char a)
+{
+ return ((a * 0x0802LU & 0x22110LU) | (a * 0x8020LU & 0x88440LU)) * 0x10101LU >> 16;
+}
+
+static unsigned char lookup[16] = {
+ 0x0, 0x8, 0x4, 0xc, 0x2, 0xa, 0x6, 0xe,
+ 0x1, 0x9, 0x5, 0xd, 0x3, 0xb, 0x7, 0xf, };
+
+unsigned char reverse_byte(unsigned char n) {
+ // Reverse the top and bottom nibble then swap them.
+ return (lookup[n & 0b1111] << 4) | lookup[n >> 4];
+}
+
+
+void transpose8rS32_reversed_diagonale(unsigned char* A, int m, int n, unsigned char* B)
{
unsigned x, y, t;
// Load the array and pack it into x and y.
-
x = (A[0] << 24) | (A[m] << 16) | (A[2 * m] << 8) | A[3 * m];
y = (A[4 * m] << 24) | (A[5 * m] << 16) | (A[6 * m] << 8) | A[7 * m];
- t = (x ^ (x >> 7)) & 0x00AA00AA; x = x ^ t ^ (t << 7);
- t = (y ^ (y >> 7)) & 0x00AA00AA; y = y ^ t ^ (t << 7);
+ t = (x ^ (x >> 7)) & 0x00AA00AA; x = x ^ t ^ (t << 7);
+ t = (y ^ (y >> 7)) & 0x00AA00AA; y = y ^ t ^ (t << 7);
- t = (x ^ (x >> 14)) & 0x0000CCCC; x = x ^ t ^ (t << 14);
- t = (y ^ (y >> 14)) & 0x0000CCCC; y = y ^ t ^ (t << 14);
+ t = (x ^ (x >> 14)) & 0x0000CCCC; x = x ^ t ^ (t << 14);
+ t = (y ^ (y >> 14)) & 0x0000CCCC; y = y ^ t ^ (t << 14);
t = (x & 0xF0F0F0F0) | ((y >> 4) & 0x0F0F0F0F);
y = ((x << 4) & 0xF0F0F0F0) | (y & 0x0F0F0F0F);
x = t;
- B[0] = x >> 24; B[n] = x >> 16; B[2 * n] = x >> 8; B[3 * n] = x;
- B[4 * n] = y >> 24; B[5 * n] = y >> 16; B[6 * n] = y >> 8; B[7 * n] = y;
+ B[7 * n] = reverse_byte(x >> 24); B[6 * n] = reverse_byte(x >> 16); B[5 * n] = reverse_byte(x >> 8); B[4 * n] = reverse_byte(x);
+ B[3 * n] = reverse_byte(y >> 24); B[2 * n] = reverse_byte(y >> 16); B[1 * n] = reverse_byte(y >> 8); B[0 * n] = reverse_byte(y);
+}
+
+void transpose_bin(char *A, char *B, const int n, const int m,
+ const int lda, const int ldb, const int block_size)
+{
+ int i;
+ #pragma omp parallel for
+ for (i = 0; i < n; i += 8) {
+ int j;
+ for (j = 0; j < m - 8; j += 8) {
+ int a_index = i*lda + j;
+ int b_index = j*ldb + i;
+ //transpose_8x8_bits_my(&A[a_index/8], &B[b_index/8], lda/8, ldb/8);
+ transpose8rS32_reversed_diagonale(&A[a_index / 8], lda / 8, ldb / 8, &B[b_index / 8]);
+ }
+ for (; j < m; ++j) {
+ if (get_bit(A, i*lda + j)) set_bit(B, j*ldb + i);
+ }
+ }
}
void activate_array_cpu_custom(float *x, const int n, const ACTIVATION a)
@@ -1102,14 +1288,18 @@
size_t i;
__m256i all256_sing1 = _mm256_set_epi32(0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000);
+ __m256 float_zero256 = _mm256_set1_ps(0.00001);
for (i = 0; i < size; i+=8)
{
- __m256i src256 = _mm256_loadu_si256((__m256i *)(&src[i]));
- __m256i result256 = _mm256_and_si256(src256, all256_sing1); // check sign in 8 x 32-bit floats
+ //__m256i src256 = _mm256_loadu_si256((__m256i *)(&src[i]));
+ //__m256i result256 = _mm256_and_si256(src256, all256_sing1); // check sign in 8 x 32-bit floats
+ //uint32_t mask = _mm256_movemask_ps(_mm256_castsi256_ps(result256)); // (val >= 0) ? 0 : 1
+ //mask = ~mask; // inverse mask, (val >= 0) ? 1 : 0
- uint32_t mask = _mm256_movemask_ps(_mm256_castsi256_ps(result256)); // (val >= 0) ? 0 : 1
- mask = ~mask; // inverse mask, (val >= 0) ? 1 : 0
+ __m256 src256 = _mm256_loadu_ps((float *)(&src[i]));
+ __m256 result256 = _mm256_cmp_ps(src256, float_zero256, _CMP_GT_OS);
+ uint32_t mask = _mm256_movemask_ps(result256); // (val > 0) ? 0 : 1
dst[i / 8] = mask;
}
diff --git a/src/gemm.h b/src/gemm.h
index 3cdca99..d393b0d 100644
--- a/src/gemm.h
+++ b/src/gemm.h
@@ -11,12 +11,14 @@
size_t dst_i = index / 8;
int dst_shift = index % 8;
dst[dst_i] |= 1 << dst_shift;
+ //dst[dst_i] |= 1 << (8 - dst_shift);
}
static inline unsigned char get_bit(unsigned char const*const src, size_t index) {
size_t src_i = index / 8;
int src_shift = index % 8;
unsigned char val = (src[src_i] & (1 << src_shift)) > 0;
+ //unsigned char val = (src[src_i] & (1 << (8 - src_shift))) > 0;
return val;
}
@@ -25,6 +27,9 @@
void transpose_block_SSE4x4(float *A, float *B, const int n, const int m,
const int lda, const int ldb, const int block_size);
+void transpose_bin(char *A, char *B, const int n, const int m,
+ const int lda, const int ldb, const int block_size);
+
void gemm_nn_custom_bin_mean_transposed(int M, int N, int K, float ALPHA_UNUSED,
unsigned char *A, int lda,
unsigned char *B, int ldb,
@@ -34,6 +39,10 @@
int channels, int height, int width,
int ksize, int stride, int pad, float* data_col);
+void im2col_cpu_custom_bin(float* data_im,
+ int channels, int height, int width,
+ int ksize, int stride, int pad, float* data_col, int bit_align);
+
void im2col_cpu_custom_transpose(float* data_im,
int channels, int height, int width,
int ksize, int stride, int pad, float* data_col, int ldb_align);
diff --git a/src/layer.h b/src/layer.h
index bd8518a..72a63e9 100644
--- a/src/layer.h
+++ b/src/layer.h
@@ -182,6 +182,7 @@
char *align_bit_weights;
float *mean_arr;
int lda_align;
+ int bit_align;
float *col_image;
int * input_layers;
--
Gitblit v1.10.0