"git@developer.sourcefind.cn:modelzoo/resnet50_tensorflow.git" did not exist on "9b6cd10635f6662d8ce4afaa29adb63fadc2972b"
Commit 59328f3a authored by Jing Zhang's avatar Jing Zhang
Browse files

merge master

parents 063342ca fcbb9788
...@@ -164,7 +164,7 @@ void device_dynamic_convolution_forward_implicit_gemm_v5r1_nchw_kcyx_nkhw( ...@@ -164,7 +164,7 @@ void device_dynamic_convolution_forward_implicit_gemm_v5r1_nchw_kcyx_nkhw(
#endif #endif
constexpr auto conv_driver = constexpr auto conv_driver =
//DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad< // DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad<
DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_outpad< DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_outpad<
BlockSize, BlockSize,
typename vector_type<TInWei, InWeiVectorSize>::type, typename vector_type<TInWei, InWeiVectorSize>::type,
......
...@@ -50,7 +50,7 @@ int main(int argc, char* argv[]) ...@@ -50,7 +50,7 @@ int main(int argc, char* argv[])
using LeftPads = Sequence<0, 0>; using LeftPads = Sequence<0, 0>;
using RightPads = Sequence<0, 0>; using RightPads = Sequence<0, 0>;
#elif 1 #elif 0
constexpr index_t N = 1; constexpr index_t N = 1;
constexpr index_t C = 16; constexpr index_t C = 16;
constexpr index_t HI = 270; constexpr index_t HI = 270;
...@@ -64,21 +64,7 @@ int main(int argc, char* argv[]) ...@@ -64,21 +64,7 @@ int main(int argc, char* argv[])
using LeftPads = Sequence<0, 0>; using LeftPads = Sequence<0, 0>;
using RightPads = Sequence<0, 0>; using RightPads = Sequence<0, 0>;
#elif 0 #elif 1
constexpr index_t N = 1;
constexpr index_t C = 16;
constexpr index_t HI = 2048;
constexpr index_t WI = 2048;
constexpr index_t K = 16;
constexpr index_t Y = 3;
constexpr index_t X = 3;
using ConvStrides = Sequence<1, 1>;
using ConvDilations = Sequence<1, 1>;
using LeftPads = Sequence<1, 1>;
using RightPads = Sequence<1, 1>;
#elif 0
constexpr index_t N = 1; constexpr index_t N = 1;
constexpr index_t C = 16; constexpr index_t C = 16;
constexpr index_t HI = 1080; constexpr index_t HI = 1080;
...@@ -118,20 +104,6 @@ int main(int argc, char* argv[]) ...@@ -118,20 +104,6 @@ int main(int argc, char* argv[])
using ConvStrides = Sequence<1, 1>; using ConvStrides = Sequence<1, 1>;
using ConvDilations = Sequence<1, 1>; using ConvDilations = Sequence<1, 1>;
using LeftPads = Sequence<1, 1>;
using RightPads = Sequence<1, 1>;
#elif 0
constexpr index_t N = 1;
constexpr index_t C = 16;
constexpr index_t HI = 135;
constexpr index_t WI = 240;
constexpr index_t K = 16;
constexpr index_t Y = 3;
constexpr index_t X = 3;
using ConvStrides = Sequence<1, 1>;
using ConvDilations = Sequence<1, 1>;
using LeftPads = Sequence<1, 1>; using LeftPads = Sequence<1, 1>;
using RightPads = Sequence<1, 1>; using RightPads = Sequence<1, 1>;
#elif 0 #elif 0
...@@ -663,17 +635,12 @@ int main(int argc, char* argv[]) ...@@ -663,17 +635,12 @@ int main(int argc, char* argv[])
constexpr index_t in_vector_size = 1; constexpr index_t in_vector_size = 1;
using acc_data_t = float; using acc_data_t = float;
using out_data_t = float; using out_data_t = float;
#elif 1
using in_data_t = half_t;
constexpr index_t in_vector_size = 16;
using acc_data_t = float;
using out_data_t = half_t;
#elif 0 #elif 0
using in_data_t = float; using in_data_t = float;
constexpr index_t in_vector_size = 1; constexpr index_t in_vector_size = 1;
using acc_data_t = float; using acc_data_t = float;
using out_data_t = int8_t; using out_data_t = int8_t;
#elif 0 #elif 1
using in_data_t = int8_t; using in_data_t = int8_t;
constexpr index_t in_vector_size = 16; constexpr index_t in_vector_size = 16;
using acc_data_t = int32_t; using acc_data_t = int32_t;
...@@ -774,7 +741,7 @@ int main(int argc, char* argv[]) ...@@ -774,7 +741,7 @@ int main(int argc, char* argv[])
LeftPads{}, LeftPads{},
RightPads{}, RightPads{},
nrepeat); nrepeat);
#elif 1 #elif 0
device_dynamic_convolution_forward_implicit_gemm_v4r4_nhwc_kyxc_nhwk<in_data_t, device_dynamic_convolution_forward_implicit_gemm_v4r4_nhwc_kyxc_nhwk<in_data_t,
in_vector_size, in_vector_size,
acc_data_t, acc_data_t,
...@@ -821,7 +788,6 @@ int main(int argc, char* argv[]) ...@@ -821,7 +788,6 @@ int main(int argc, char* argv[])
check_error(out_nkhw_host, out_nkhw_device); check_error(out_nkhw_host, out_nkhw_device);
#if 0
if(do_log) if(do_log)
{ {
LogRange(std::cout << "in_nchw : ", in_nchw.mData, ",") << std::endl; LogRange(std::cout << "in_nchw : ", in_nchw.mData, ",") << std::endl;
...@@ -829,6 +795,5 @@ int main(int argc, char* argv[]) ...@@ -829,6 +795,5 @@ int main(int argc, char* argv[])
LogRange(std::cout << "out_nkhw_host : ", out_nkhw_host.mData, ",") << std::endl; LogRange(std::cout << "out_nkhw_host : ", out_nkhw_host.mData, ",") << std::endl;
LogRange(std::cout << "out_nkhw_device: ", out_nkhw_device.mData, ",") << std::endl; LogRange(std::cout << "out_nkhw_device: ", out_nkhw_device.mData, ",") << std::endl;
} }
#endif
} }
} }
...@@ -10,7 +10,7 @@ cmake ...@@ -10,7 +10,7 @@ cmake
-D CMAKE_INSTALL_PREFIX=${MY_PROJECT_INSTALL} \ -D CMAKE_INSTALL_PREFIX=${MY_PROJECT_INSTALL} \
-D CMAKE_BUILD_TYPE=Release \ -D CMAKE_BUILD_TYPE=Release \
-D DEVICE_BACKEND="AMD" \ -D DEVICE_BACKEND="AMD" \
-D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx906 -gline-tables-only -save-temps=$CWD -ftemplate-backtrace-limit=0" \ -D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx1030 -gline-tables-only -save-temps=$CWD -ftemplate-backtrace-limit=0" \
-D CMAKE_CXX_COMPILER=/opt/rocm/bin/hipcc \ -D CMAKE_CXX_COMPILER=/opt/rocm/bin/hipcc \
-D CMAKE_PREFIX_PATH="/opt/rocm" \ -D CMAKE_PREFIX_PATH="/opt/rocm" \
-D CMAKE_VERBOSE_MAKEFILE:BOOL=ON \ -D CMAKE_VERBOSE_MAKEFILE:BOOL=ON \
......
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