WIP: explicitly separate offset component into compile-time, block-invariant and per-thread components

This commit is contained in:
Chao Liu
2019-09-21 22:53:03 -05:00
parent 740da00aa2
commit 51884fc214
11 changed files with 337 additions and 76 deletions

View File

@@ -9,17 +9,17 @@
#include "conv_common.hpp"
#include "host_conv.hpp"
#include "device_convolution_direct_v2_nchw_kcyx_nkhw.hpp"
#include "device_convolution_implicit_gemm_v1_chwn_cyxk_khwn.hpp"
#include "device_convolution_implicit_gemm_v1_chwn_cyxk_khwn_padded.hpp"
//#include "device_convolution_implicit_gemm_v1_chwn_cyxk_khwn.hpp"
//#include "device_convolution_implicit_gemm_v1_chwn_cyxk_khwn_padded.hpp"
//#include "device_convolution_implicit_gemm_v1_nchw_cyxk_nkhw.hpp"
//#include "device_convolution_implicit_gemm_v2_chwn_cyxk_khwn.hpp"
//#include "device_convolution_implicit_gemm_v3_nchw_cyxk_nkhw.hpp"
#include "device_convolution_implicit_gemm_v4r1_nchw_kcyx_nkhw.hpp"
//#include "device_convolution_implicit_gemm_v4r1_nchw_kcyx_nkhw.hpp"
#include "device_convolution_implicit_gemm_v4r1_nchw_kcyx_nkhw_padded.hpp"
//#include "device_convolution_implicit_gemm_v4r2_nchw_kcyx_nkhw.hpp"
//#include "device_convolution_implicit_gemm_v4r3_nchw_kcyx_nkhw.hpp"
#include "device_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw.hpp"
#include "device_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw_padded.hpp"
//#include "device_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw.hpp"
//#include "device_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw_padded.hpp"
struct GeneratorTensor_1
{