Commit 71bc108d authored by Jing Zhang's avatar Jing Zhang
Browse files

clean

parent 20e6bc9d
...@@ -119,7 +119,7 @@ void device_convolution_forward_implicit_gemm_v5r1_dlops_nchw_kcyx_nkhw( ...@@ -119,7 +119,7 @@ void device_convolution_forward_implicit_gemm_v5r1_dlops_nchw_kcyx_nkhw(
constexpr index_t BlockSize = 64; constexpr index_t BlockSize = 64;
constexpr index_t KPerBlock = 16; constexpr index_t KPerBlock = 16;
constexpr index_t HoPerBlock = 4; constexpr index_t HoPerBlock = 8;
constexpr index_t WoPerBlock = 32; constexpr index_t WoPerBlock = 32;
constexpr index_t E1 = 2 * 9; constexpr index_t E1 = 2 * 9;
...@@ -127,7 +127,7 @@ void device_convolution_forward_implicit_gemm_v5r1_dlops_nchw_kcyx_nkhw( ...@@ -127,7 +127,7 @@ void device_convolution_forward_implicit_gemm_v5r1_dlops_nchw_kcyx_nkhw(
constexpr index_t K2 = 2; constexpr index_t K2 = 2;
constexpr index_t E1PerBlock = 2; constexpr index_t E1PerBlock = 2;
constexpr index_t KPerThread = 8; constexpr index_t KPerThread = 16;
constexpr index_t HoPerThread = 2; constexpr index_t HoPerThread = 2;
constexpr index_t WoPerThread = 2; constexpr index_t WoPerThread = 2;
constexpr index_t EPerThread = 1; constexpr index_t EPerThread = 1;
......
...@@ -3,7 +3,7 @@ ...@@ -3,7 +3,7 @@
#include <initializer_list> #include <initializer_list>
#include <cstdlib> #include <cstdlib>
#include <stdlib.h> #include <stdlib.h>
//#include <half.hpp> #include <half.hpp>
#include "config.hpp" #include "config.hpp"
#include "print.hpp" #include "print.hpp"
#include "device.hpp" #include "device.hpp"
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment