From 0dab894a5be9f7d10d85e89dea91d02c71bae84d Mon Sep 17 00:00:00 2001
From: Edmond Yoo <hj3yoo@uwaterloo.ca>
Date: Sun, 16 Sep 2018 03:24:45 +0000
Subject: [PATCH] Moving files from MTGCardDetector repo
---
src/gemm.h | 80 ++++++++++++++++++++++++++++++++++-----
1 files changed, 69 insertions(+), 11 deletions(-)
diff --git a/src/gemm.h b/src/gemm.h
index f0231bf..90b8ce9 100644
--- a/src/gemm.h
+++ b/src/gemm.h
@@ -1,32 +1,90 @@
#ifndef GEMM_H
#define GEMM_H
+#include "activations.h"
+#include <stdint.h>
+#include <stddef.h>
-void gemm_bin(int M, int N, int K, float ALPHA,
- char *A, int lda,
+void convolution_2d(int w, int h, int ksize, int n, int c, int pad, int stride,
+ float *weights, float *input, float *output, float *mean);
+
+static inline void set_bit(unsigned char *const dst, size_t index) {
+ 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;
+}
+
+void float_to_bit(float *src, unsigned char *dst, size_t size);
+
+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,
+ float *C, int ldc, float *mean_arr);
+
+void im2col_cpu_custom(float* data_im,
+ int channels, int height, int width,
+ int ksize, int stride, int pad, float* data_col);
+
+void im2col_cpu_custom_align(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_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);
+
+void activate_array_cpu_custom(float *x, const int n, const ACTIVATION a);
+
+
+void gemm_bin(int M, int N, int K, float ALPHA,
+ char *A, int lda,
float *B, int ldb,
float *C, int ldc);
-
-void gemm(int TA, int TB, int M, int N, int K, float ALPHA,
- float *A, int lda,
+
+
+void forward_maxpool_layer_avx(float *src, float *dst, int *indexes, int size, int w, int h, int out_w, int out_h, int c,
+ int pad, int stride, int batch);
+
+
+void gemm(int TA, int TB, int M, int N, int K, float ALPHA,
+ float *A, int lda,
float *B, int ldb,
float BETA,
float *C, int ldc);
-void gemm_cpu(int TA, int TB, int M, int N, int K, float ALPHA,
- float *A, int lda,
+void gemm_cpu(int TA, int TB, int M, int N, int K, float ALPHA,
+ float *A, int lda,
float *B, int ldb,
float BETA,
float *C, int ldc);
#ifdef GPU
-void gemm_ongpu(int TA, int TB, int M, int N, int K, float ALPHA,
- float *A_gpu, int lda,
+void gemm_ongpu(int TA, int TB, int M, int N, int K, float ALPHA,
+ float *A_gpu, int lda,
float *B_gpu, int ldb,
float BETA,
float *C_gpu, int ldc);
-void gemm_gpu(int TA, int TB, int M, int N, int K, float ALPHA,
- float *A, int lda,
+void gemm_gpu(int TA, int TB, int M, int N, int K, float ALPHA,
+ float *A, int lda,
float *B, int ldb,
float BETA,
float *C, int ldc);
--
Gitblit v1.10.0