diff --git a/llvm/projects/hpvm-tensor-rt/CMakeLists.txt b/llvm/projects/hpvm-tensor-rt/CMakeLists.txt index 2434e773b0368b7126940eb75aa8a3d97136af50..85b1e3102077483943db89414cb65808db6b2de6 100644 --- a/llvm/projects/hpvm-tensor-rt/CMakeLists.txt +++ b/llvm/projects/hpvm-tensor-rt/CMakeLists.txt @@ -149,6 +149,9 @@ target_link_libraries(alexnet_cifar10 tensor_runtime ${GPU_PROFILER_LIB} ${SOC_ #add_executable(alexnet_cifar10_approx dnn_sources/src/alexnet_cifar10_approx.cc) #target_link_libraries(alexnet_cifar10_approx tensor_runtime ${GPU_PROFILER_LIB} ${SOC_SIMULATOR_LIB}) +add_executable(alexnet2_canny dnn_sources/src/alexnet2_canny.cc) +target_link_libraries(alexnet2_canny tensor_runtime ${GPU_PROFILER_LIB} ${SOC_SIMULATOR_LIB}) + add_executable(alexnet2_cifar10 dnn_sources/src/alexnet2_cifar10.cc) target_link_libraries(alexnet2_cifar10 tensor_runtime ${GPU_PROFILER_LIB} ${SOC_SIMULATOR_LIB}) diff --git a/llvm/projects/hpvm-tensor-rt/dnn_sources/src/alexnet_canny.cc b/llvm/projects/hpvm-tensor-rt/dnn_sources/src/alexnet2_canny.cc similarity index 96% rename from llvm/projects/hpvm-tensor-rt/dnn_sources/src/alexnet_canny.cc rename to llvm/projects/hpvm-tensor-rt/dnn_sources/src/alexnet2_canny.cc index 65d6335f75fb5f3e9469e42507e063a2b526aee8..628ce6616cde37a5eddde5ab6049001525203580 100644 --- a/llvm/projects/hpvm-tensor-rt/dnn_sources/src/alexnet_canny.cc +++ b/llvm/projects/hpvm-tensor-rt/dnn_sources/src/alexnet2_canny.cc @@ -119,6 +119,7 @@ int main(){ std::string dir_prefix = std::string("../model_params/alexnet2_cifar10/"); std::string input_path = dir_prefix + std::string("norm_cifar_input.bin"); + std::string canny_input_path = dir_prefix + std::string("canny_input.bin"); std::string labels_path = dir_prefix + std::string("test_labels.bin"); void* conv1_filter = readTrainedWeights("../model_params/alexnet2_cifar10/conv1.bin", @@ -171,6 +172,7 @@ int main(){ void* input = readInputBatch(input_path.c_str(), 0,start,end,3,32,32); + void* canny_input = readInputBatch(canny_input_path.c_str(), 0,start,end, 3, 128, 128); void* conv1out = tensorConvolution(input, conv1_filter, 1, 1, 1, 1, conv_mode, conv_precision); @@ -232,9 +234,8 @@ int main(){ void* argmax_out = tensorArgMax(result); void* select_out = tensorSelect2(argmax_out, index_vector); - void* reduced_input = tensorContract(input, select_out); + void* reduced_input = tensorContract(canny_input, select_out); - invoke_canny(reduced_input);