ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-20 04:49:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Use DynamicBuffer instead of raw pointer (#32)

Browse Source 
* Use DynamicBuffer to hold raw pointer (to global and LDS memory)

* add workaround for compiler issue (inefficient ISA) of ds_write for int8x4, int8x8, int8x16

[ROCm/composable_kernel commit: 78b987fbd6]
This commit is contained in:
Chao Liu
2021-05-12 13:10:42 -05:00
committed by GitHub
parent ce4662dbda
commit da8f48c6d6
16 changed files with 564 additions and 490 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
driver/include/device_dynamic_convolution_forward_implicit_gemm_v5r1_nchw_kcyx_nkhw.hpp
View File

@@ -63,12 +63,12 @@ void device_dynamic_convolution_forward_implicit_gemm_v5r1_nchw_kcyx_nkhw(
#if 0
// run-time variables
const auto in_n_c_hi_wi_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(to_multi_index(InDesc::GetLengths()));
const auto wei_k_c_y_x_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(to_multi_index(WeiDesc::GetLengths()));
const auto out_n_k_ho_wo_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(to_multi_index(OutDesc::GetLengths()));
const auto in_n_c0_hi_wi_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_multi_index(N, C0, Hi, Wi));
const auto wei_k_c0_y_x_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_multi_index(K, C0, Y, X));
const auto out_n_k0_ho_wo_k1_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_multi_index(N, K0, Ho, Wo, K1));
const auto conv_strides = to_multi_index(ConvStrides{});
const auto conv_dilations = to_multi_index(ConvDilations{});