Merge branch 'develop' into gemm_layernorm_welford

library/src/tensor_operation_instance/gpu/quantization/CMakeLists.txt

 add_instance_library(device_quantization_instance
-    device_conv2d_xdl_bias_quant_int8_instance.cpp
-    device_conv2d_xdl_quant_int8_instance.cpp
+    device_conv2d_xdl_bias_perchannel_quantization_int8_instance.cpp
+    device_conv2d_xdl_bias_perlayer_quantization_int8_instance.cpp
+    device_conv2d_xdl_perchannel_quantization_int8_instance.cpp
+    device_conv2d_xdl_perlayer_quantization_int8_instance.cpp
 )

profiler/CMakeLists.txt

@@ -27,6 +27,7 @@ set(PROFILER_SOURCE
     src/profile_layernorm.cpp
     src/profile_softmax.cpp
     src/profile_batchnorm_fwd.cpp
+    src/profile_batchnorm_bwd.cpp
 )
 
 add_executable(ckProfiler ${PROFILER_SOURCE})

profiler/src/profiler.cpp

@@ -25,6 +25,7 @@ int profile_layernorm(int, char*[]);
 int profile_groupnorm(int, char*[]);
 int profile_reduce(int, char*[]);
 int profile_batchnorm_forward(int, char*[]);
+int profile_batchnorm_backward(int, char*[]);
 
 static void print_helper_message()
 {
@@ -148,6 +149,10 @@ int main(int argc, char* argv[])
     {
         return profile_batchnorm_forward(argc, argv);
     }
+    else if(strcmp(argv[1], "bnorm_bwd") == 0)
+    {
+        return profile_batchnorm_backward(argc, argv);
+    }
     else
     {
         print_helper_message();

test/CMakeLists.txt

@@ -53,4 +53,4 @@ add_subdirectory(softmax)
 add_subdirectory(normalization)
 add_subdirectory(data_type)
 add_subdirectory(elementwise_normalization)
-add_subdirectory(batchnorm_fwd)
+add_subdirectory(batchnorm)

test/batchnorm_fwd/CMakeLists.txt → test/batchnorm/CMakeLists.txt

 add_gtest_executable(test_batchnorm_fwd_rank_4 batchnorm_fwd_rank_4.cpp)
+add_gtest_executable(test_batchnorm_bwd_rank_4 batchnorm_bwd_rank_4.cpp)
 target_link_libraries(test_batchnorm_fwd_rank_4 PRIVATE utility device_batchnorm_instance)
+target_link_libraries(test_batchnorm_bwd_rank_4 PRIVATE utility device_batchnorm_instance)