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 +++++++++
1 files changed, 9 insertions(+), 0 deletions(-)
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);
--
Gitblit v1.10.0