Merge remote-tracking branch 'origin/develop' into gpu-invoker

example/15_grouped_gemm/grouped_gemm_xdl_bfp16.cpp

@@ -9,13 +9,14 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/device_grouped_gemm_xdl.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/literals.hpp"
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
 
 template <ck::index_t... Is>

example/15_grouped_gemm/grouped_gemm_xdl_fp16.cpp

@@ -9,13 +9,14 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/device_grouped_gemm_xdl.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/literals.hpp"
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
 
 template <ck::index_t... Is>

example/15_grouped_gemm/grouped_gemm_xdl_fp32.cpp

@@ -9,13 +9,14 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/device_grouped_gemm_xdl.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/literals.hpp"
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
 
 template <ck::index_t... Is>

example/15_grouped_gemm/grouped_gemm_xdl_int4.cpp

@@ -9,13 +9,14 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/device_grouped_gemm_xdl.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/literals.hpp"
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
 
 template <ck::index_t... Is>

example/15_grouped_gemm/grouped_gemm_xdl_int8.cpp

@@ -9,13 +9,14 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/device_grouped_gemm_xdl.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/literals.hpp"
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
 
 template <ck::index_t... Is>

example/15_grouped_gemm/run_grouped_gemm_example.inc

@@ -52,15 +52,15 @@ bool run_grouped_gemm(const ProblemSize& problem_size, const ExecutionConfig& co
 
     auto f_host_tensor_descriptor =
         [](std::size_t row, std::size_t col, std::size_t stride, auto layout) {
+            using namespace ck::literals;
+
             if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::value)
             {
-                return HostTensorDescriptor(std::vector<std::size_t>({row, col}),
-                                            std::vector<std::size_t>({stride, 1}));
+                return HostTensorDescriptor({row, col}, {stride, 1_uz});
             }
             else
             {
-                return HostTensorDescriptor(std::vector<std::size_t>({row, col}),
-                                            std::vector<std::size_t>({1, stride}));
+                return HostTensorDescriptor({row, col}, {1_uz, stride});
             }
         };
 
@@ -208,10 +208,10 @@ bool run_grouped_gemm(const ProblemSize& problem_size, const ExecutionConfig& co
 
 #ifdef BUILD_INT4_EXAMPLE
             const Tensor<EDataType> c_device_result_converted(c_device_tensors[i]);
-            pass &= ck::utils::check_err(c_device_result_converted.mData, c_host_tensors[i].mData);
+            pass &= ck::utils::check_err(c_device_result_converted, c_host_tensors[i]);
 
 #else
-            pass &= ck::utils::check_err(c_device_tensors[i].mData, c_host_tensors[i].mData);
+            pass &= ck::utils::check_err(c_device_tensors[i], c_host_tensors[i]);
 #endif
         }
     }

example/16_gemm_multi_d_multi_reduces/gemm_add_add_mean_meansquare_xdl_fp16.cpp

@@ -9,12 +9,13 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/literals.hpp"
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
 #include "ck/library/utility/check_err.hpp"
 
@@ -109,21 +110,20 @@ void DumpPerf(float ave_time, int M, int N, int K)
 }
 
 auto f_host_tensor_descriptor1d = [](std::size_t len, std::size_t stride) {
-    return HostTensorDescriptor(std::vector<std::size_t>({len}),
-                                std::vector<std::size_t>({stride}));
+    return HostTensorDescriptor({len}, {stride});
 };
 
 auto f_host_tensor_descriptor2d =
     [](std::size_t row, std::size_t col, std::size_t stride, auto layout) {
+        using namespace ck::literals;
+
         if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::value)
         {
-            return HostTensorDescriptor(std::vector<std::size_t>({row, col}),
-                                        std::vector<std::size_t>({stride, 1}));
+            return HostTensorDescriptor({row, col}, {stride, 1_uz});
         }
         else
         {
-            return HostTensorDescriptor(std::vector<std::size_t>({row, col}),
-                                        std::vector<std::size_t>({1, stride}));
+            return HostTensorDescriptor({row, col}, {1_uz, stride});
         }
     };
 
@@ -259,12 +259,9 @@ int main()
         r0_device_buf.FromDevice(r0_m.mData.data());
         r1_device_buf.FromDevice(r1_m.mData.data());
 
-        pass = ck::utils::check_err(
-            e_m_n.mData, e_m_n_host.mData, "Error: Incorrect results c", 1e-2, 1e-2);
-        pass &= ck::utils::check_err(
-            r0_m.mData, r0_m_host.mData, "Error: Incorrect results d0", 1e-2, 1e-2);
-        pass &= ck::utils::check_err(
-            r1_m.mData, r1_m_host.mData, "Error: Incorrect results d1", 1e-2, 1e-2);
+        pass = ck::utils::check_err(e_m_n, e_m_n_host, "Error: Incorrect results c", 1e-2, 1e-2);
+        pass &= ck::utils::check_err(r0_m, r0_m_host, "Error: Incorrect results d0", 1e-2, 1e-2);
+        pass &= ck::utils::check_err(r1_m, r1_m_host, "Error: Incorrect results d1", 1e-2, 1e-2);
     }
 
     bool time_kernel = true;

example/16_gemm_multi_d_multi_reduces/gemm_add_addsquare_xdl_int8.cpp

@@ -5,7 +5,7 @@
 
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
 #include "ck/library/utility/literals.hpp"
-#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 
 // DataType
@@ -160,14 +160,12 @@ bool run_gemm_reduce_add_addsquare_xdl(ck::index_t M,
     {
     case 0: break;
     case 1:
-        ck::utils::FillUniformDistributionIntegerValue<ADataType>{-5.f, 5.f}(a_m_k.begin(),
-                                                                             a_m_k.end());
-        ck::utils::FillUniformDistributionIntegerValue<BDataType>{-5.f, 5.f}(b_k_n.begin(),
-                                                                             b_k_n.end());
+        ck::utils::FillUniformDistributionIntegerValue<ADataType>{-5.f, 5.f}(a_m_k);
+        ck::utils::FillUniformDistributionIntegerValue<BDataType>{-5.f, 5.f}(b_k_n);
         break;
     default:
-        ck::utils::FillUniformDistribution<ADataType>{-1.f, 1.f}(a_m_k.begin(), a_m_k.end());
-        ck::utils::FillUniformDistribution<BDataType>{-1.f, 1.f}(b_k_n.begin(), b_k_n.end());
+        ck::utils::FillUniformDistribution<ADataType>{-1.f, 1.f}(a_m_k);
+        ck::utils::FillUniformDistribution<BDataType>{-1.f, 1.f}(b_k_n);
         break;
     }
 
@@ -264,15 +262,13 @@ bool run_gemm_reduce_add_addsquare_xdl(ck::index_t M,
         Tensor<EDataType> e_m_n_host_converted(e_m_n_host);
 
         pass = ck::utils::check_err(
-            e_m_n.mData, e_m_n_host_converted.mData, "Error: Incorrect results c", 1e-2, 1e-2);
+            e_m_n, e_m_n_host_converted, "Error: Incorrect results c", 1e-2, 1e-2);
 
         r0_device_buf.FromDevice(r0_m.mData.data());
         r1_device_buf.FromDevice(r1_m.mData.data());
 
-        pass &= ck::utils::check_err(
-            r0_m.mData, r0_m_host.mData, "Error: Incorrect results d0", 1e-2, 1e-2);
-        pass &= ck::utils::check_err(
-            r1_m.mData, r1_m_host.mData, "Error: Incorrect results d1", 1e-2, 1e-2);
+        pass &= ck::utils::check_err(r0_m, r0_m_host, "Error: Incorrect results d0", 1e-2, 1e-2);
+        pass &= ck::utils::check_err(r1_m, r1_m_host, "Error: Incorrect results d1", 1e-2, 1e-2);
 
         if(pass)
         {

example/16_gemm_multi_d_multi_reduces/gemm_max_xdl_bf16.cpp

@@ -4,7 +4,7 @@
 #include "gemm_reduce_xdl_common.hpp"
 
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
-#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 
 // DataType

example/16_gemm_multi_d_multi_reduces/gemm_max_xdl_fp16.cpp

@@ -4,7 +4,7 @@
 #include "gemm_reduce_xdl_common.hpp"
 
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
-#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 
 // DataType

example/16_gemm_multi_d_multi_reduces/gemm_max_xdl_fp32.cpp

@@ -4,7 +4,7 @@
 #include "gemm_reduce_xdl_common.hpp"
 
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
-#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 
 // DataType

example/16_gemm_multi_d_multi_reduces/gemm_max_xdl_int4.cpp

@@ -4,7 +4,7 @@
 #include "gemm_reduce_xdl_common.hpp"
 
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
-#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 
 using ADataType         = INT4;

example/16_gemm_multi_d_multi_reduces/gemm_max_xdl_int8.cpp

@@ -4,7 +4,7 @@
 #include "gemm_reduce_xdl_common.hpp"
 
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
-#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 
 using ADataType         = INT8;

example/16_gemm_multi_d_multi_reduces/gemm_mean_meansquare_xdl_bf16.cpp

@@ -4,7 +4,7 @@
 #include "gemm_reduce_xdl_common.hpp"
 
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
-#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 
 // DataType

example/16_gemm_multi_d_multi_reduces/gemm_mean_meansquare_xdl_fp16.cpp

@@ -4,7 +4,7 @@
 #include "gemm_reduce_xdl_common.hpp"
 
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
-#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 
 // DataType

example/16_gemm_multi_d_multi_reduces/gemm_mean_meansquare_xdl_fp32.cpp

@@ -4,7 +4,7 @@
 #include "gemm_reduce_xdl_common.hpp"
 
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
-#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 
 // DataType

example/16_gemm_multi_d_multi_reduces/gemm_reduce_xdl_common.hpp

@@ -134,14 +134,12 @@ auto run_gemm_reduce_max_xdl(ck::index_t M,
     {
     case 0: break;
     case 1:
-        ck::utils::FillUniformDistributionIntegerValue<ADataType>{-5.f, 5.f}(a_m_k.begin(),
-                                                                             a_m_k.end());
-        ck::utils::FillUniformDistributionIntegerValue<BDataType>{-5.f, 5.f}(b_k_n.begin(),
-                                                                             b_k_n.end());
+        ck::utils::FillUniformDistributionIntegerValue<ADataType>{-5.f, 5.f}(a_m_k);
+        ck::utils::FillUniformDistributionIntegerValue<BDataType>{-5.f, 5.f}(b_k_n);
         break;
     default:
-        ck::utils::FillUniformDistribution<ADataType>{-1.f, 1.f}(a_m_k.begin(), a_m_k.end());
-        ck::utils::FillUniformDistribution<BDataType>{-1.f, 1.f}(b_k_n.begin(), b_k_n.end());
+        ck::utils::FillUniformDistribution<ADataType>{-1.f, 1.f}(a_m_k);
+        ck::utils::FillUniformDistribution<BDataType>{-1.f, 1.f}(b_k_n);
         break;
     }
 
@@ -243,8 +241,8 @@ auto run_gemm_reduce_max_xdl(ck::index_t M,
         if constexpr(std::is_same_v<ADataType, ck::int4_t>)
         {
             Tensor<EDataType> e_m_n_device_converted(e_m_n);
-            pass = ck::utils::check_err(e_m_n_device_converted.mData,
-                                        e_m_n_host_converted.mData,
+            pass = ck::utils::check_err(e_m_n_device_converted,
+                                        e_m_n_host_converted,
                                         "Error: Incorrect results c",
                                         1e-2,
                                         1e-2);
@@ -253,12 +251,11 @@ auto run_gemm_reduce_max_xdl(ck::index_t M,
 #endif // CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
         {
             pass = ck::utils::check_err(
-                e_m_n.mData, e_m_n_host_converted.mData, "Error: Incorrect results c", 1e-2, 1e-2);
+                e_m_n, e_m_n_host_converted, "Error: Incorrect results c", 1e-2, 1e-2);
         }
 
         r0_device_buf.FromDevice(r0_m.mData.data());
-        pass &= ck::utils::check_err(
-            r0_m.mData, r0_m_host.mData, "Error: Incorrect results d0", 1e-2, 1e-2);
+        pass &= ck::utils::check_err(r0_m, r0_m_host, "Error: Incorrect results d0", 1e-2, 1e-2);
 
         if(pass)
         {
@@ -339,14 +336,12 @@ bool run_gemm_reduce_mean_meansquare_xdl(ck::index_t M,
     {
     case 0: break;
     case 1:
-        ck::utils::FillUniformDistributionIntegerValue<ADataType>{-5.f, 5.f}(a_m_k.begin(),
-                                                                             a_m_k.end());
-        ck::utils::FillUniformDistributionIntegerValue<BDataType>{-5.f, 5.f}(b_k_n.begin(),
-                                                                             b_k_n.end());
+        ck::utils::FillUniformDistributionIntegerValue<ADataType>{-5.f, 5.f}(a_m_k);
+        ck::utils::FillUniformDistributionIntegerValue<BDataType>{-5.f, 5.f}(b_k_n);
         break;
     default:
-        ck::utils::FillUniformDistribution<ADataType>{-1.f, 1.f}(a_m_k.begin(), a_m_k.end());
-        ck::utils::FillUniformDistribution<BDataType>{-1.f, 1.f}(b_k_n.begin(), b_k_n.end());
+        ck::utils::FillUniformDistribution<ADataType>{-1.f, 1.f}(a_m_k);
+        ck::utils::FillUniformDistribution<BDataType>{-1.f, 1.f}(b_k_n);
         break;
     }
 
@@ -460,8 +455,8 @@ bool run_gemm_reduce_mean_meansquare_xdl(ck::index_t M,
         if constexpr(std::is_same_v<ADataType, ck::int4_t>)
         {
             Tensor<EDataType> e_m_n_device_converted(e_m_n);
-            pass = ck::utils::check_err(e_m_n_device_converted.mData,
-                                        e_m_n_host_converted.mData,
+            pass = ck::utils::check_err(e_m_n_device_converted,
+                                        e_m_n_host_converted,
                                         "Error: Incorrect results c",
                                         1e-2,
                                         1e-2);
@@ -470,16 +465,14 @@ bool run_gemm_reduce_mean_meansquare_xdl(ck::index_t M,
 #endif // CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
         {
             pass = ck::utils::check_err(
-                e_m_n.mData, e_m_n_host_converted.mData, "Error: Incorrect results c", 1e-2, 1e-2);
+                e_m_n, e_m_n_host_converted, "Error: Incorrect results c", 1e-2, 1e-2);
         }
 
         r0_device_buf.FromDevice(r0_m.mData.data());
         r1_device_buf.FromDevice(r1_m.mData.data());
 
-        pass &= ck::utils::check_err(
-            r0_m.mData, r0_m_host.mData, "Error: Incorrect results d0", 1e-2, 1e-2);
-        pass &= ck::utils::check_err(
-            r1_m.mData, r1_m_host.mData, "Error: Incorrect results d1", 1e-2, 1e-2);
+        pass &= ck::utils::check_err(r0_m, r0_m_host, "Error: Incorrect results d0", 1e-2, 1e-2);
+        pass &= ck::utils::check_err(r1_m, r1_m_host, "Error: Incorrect results d1", 1e-2, 1e-2);
 
         if(pass)
         {

example/17_convnd_bwd_data/CMakeLists.txt

 add_example_executable(example_convnd_bwd_data_xdl_fp16 convnd_bwd_data_xdl_fp16.cpp)
 target_link_libraries(example_convnd_bwd_data_xdl_fp16 PRIVATE utility)
+
+add_example_executable(example_convnd_bwd_data_dl_fp16 convnd_bwd_data_dl_fp16.cpp)
+target_link_libraries(example_convnd_bwd_data_dl_fp16 PRIVATE utility)

example/17_convnd_bwd_data/convnd_bwd_data_common.hpp

@@ -61,9 +61,13 @@ int run_conv_bwd_data(bool do_verification,
         out.GenerateTensorValue(GeneratorTensor_2<OutDataType>{-5, 5});
         wei.GenerateTensorValue(GeneratorTensor_2<WeiDataType>{-5, 5});
         break;
-    default:
+    case 2:
         out.GenerateTensorValue(GeneratorTensor_3<OutDataType>{0.0, 1.0});
         wei.GenerateTensorValue(GeneratorTensor_3<WeiDataType>{-0.5, 0.5});
+        break;
+    default:
+        out.GenerateTensorValue(GeneratorTensor_1<OutDataType>{1});
+        wei.GenerateTensorValue(GeneratorTensor_1<WeiDataType>{1});
     }
 
     DeviceMem in_device_buf(sizeof(InDataType) * in_device.mDesc.GetElementSpaceSize());
@@ -98,9 +102,8 @@ int run_conv_bwd_data(bool do_verification,
 
     if(!conv.IsSupportedArgument(argument))
     {
-        throw std::runtime_error(
-            "wrong! device_conv with the specified compilation parameters does "
-            "not support this Conv problem");
+        std::cout << "Not support,please check parameters or device";
+        return 0;
     }
 
     float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
@@ -142,7 +145,7 @@ int run_conv_bwd_data(bool do_verification,
 
         in_device_buf.FromDevice(in_device.mData.data());
 
-        return ck::utils::check_err(in_device.mData, in_host.mData) ? 0 : 1;
+        return ck::utils::check_err(in_device, in_host) ? 0 : 1;
     }
 
     return 0;

example/20_convnd_bwd_weight/convnd_bwd_weight_xdl_fp16.cpp → example/17_convnd_bwd_data/convnd_bwd_data_dl_fp16.cpp

 // SPDX-License-Identifier: MIT
 // Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
 
-#include "convnd_bwd_weight_common.hpp"
+#include "convnd_bwd_data_common.hpp"
 
-#include "ck/tensor_operation/gpu/device/device_convnd_bwd_weight_nwc_kxc_nwk_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_convnd_bwd_data_nwc_kxc_nwk_dl.hpp"
 
 using InDataType  = ck::half_t;
 using WeiDataType = ck::half_t;
@@ -17,61 +17,31 @@ using InElementOp  = ck::tensor_operation::element_wise::PassThrough;
 using WeiElementOp = ck::tensor_operation::element_wise::PassThrough;
 using OutElementOp = ck::tensor_operation::element_wise::PassThrough;
 
-static constexpr auto ConvBwdWeightDefault =
-    ck::tensor_operation::device::ConvolutionBackwardWeightSpecialization::Default;
+static constexpr auto ConvBwdDefault =
+    ck::tensor_operation::device::ConvolutionBackwardDataSpecialization::Default;
 
 template <ck::index_t NDimSpatial>
-using DeviceConvndBwdWeightInstance =
-    ck::tensor_operation::device::DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle<
-        NDimSpatial,          // NDimSpatial
-        InDataType,           // InDataType
-        WeiDataType,          // WeiDataType
-        OutDataType,          // OutDataType
-        AccDataType,          // AccDataType
-        InElementOp,          // InElementwiseOperation
-        WeiElementOp,         // WeiElementwiseOperation
-        OutElementOp,         // OutElementwiseOperation
-        ConvBwdWeightDefault, // ConvolutionBackwardWeightSpecialization
-        256,                  // BlockSize
-        128,                  // MPerBlock
-        128,                  // NPerBlock
-        4,                    // K0PerBlock
-        8,                    // K1
-        32,                   // MPerXdl
-        32,                   // NPerXdl
-        2,                    // MXdlPerWave
-        2,                    // NXdlPerWave
-        S<1, 4, 16, 4>,       // ABlockTransferThreadClusterLengths_K0_M_K1
-        S<0, 3, 1, 2>,        // ABlockTransferThreadClusterArrangeOrder
-        S<0, 2, 1, 3>,        // ABlockTransferSrcAccessOrder
-        2,                    // ABlockTransferSrcVectorDim
-        8,                    // ABlockTransferSrcScalarPerVector
-        2,                    // ABlockTransferDstScalarPerVector_K1
-        true,                 // ABlockLdsAddExtraM
-        S<1, 4, 16, 4>,       // BBlockTransferThreadClusterLengths_K0_N_K1
-        S<0, 3, 1, 2>,        // BBlockTransferThreadClusterArrangeOrder
-        S<0, 2, 1, 3>,        // BBlockTransferSrcAccessOrder
-        2,                    // BBlockTransferSrcVectorDim
-        8,                    // BBlockTransferSrcScalarPerVector
-        2,                    // BBlockTransferDstScalarPerVector_K1
-        true,                 // BBlockLdsAddExtraN
-        1,                    // CShuffleMXdlPerWavePerShuffle
-        1,                    // CShuffleNXdlPerWavePerShuffle
-        S<1, 32, 1, 4>,       // CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
-        8>;                   // CBlockTransferScalarPerVector_NWaveNPerXdl
+// clang-format off
+using DeviceConvNdBwdDataInstance = ck::tensor_operation::device::DeviceConvNdBwdDataNwcKxcNwk_Dl<
+//        ######|       NDim|     InData|     WeiData|     OutData|     AccData|           In|           Wei|           Out|    Convolution| Block|  MPer|  NPer| K0Per| K1|      M1Per|      N1Per|   KPer|  M11N11Thread|  M11N11Thread|     ABlockTransfer|       ABlockTransfer| ABlockTransfer| ABlockTransfer|      ABlockTransfer|     ABlockTransfer|      ABlockTransfer|     BBlockTransfer|       BBlockTransfer| BBlockTransfer| BBlockTransfer|      BBlockTransfer|     BBlockTransfer|      BBlockTransfer|     CThreadTransfer| CThreadTransfer|    CThreadTransfer|
+//        ######|    Spatial|       Type|        Type|        Type|        Type|  Elementwise|   Elementwise|   Elementwise|        Forward|  Size| Block| Block| Block|   | ThreadM111| ThreadN111| Thread| ClusterM110Xs| ClusterN110Xs| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|     DstVectorTensor| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|     DstVectorTensor|        SrcDstAccess| SrcDstVectorDim| DstScalarPerVector|
+//        ######|           |           |            |            |            |    Operation|     Operation|     Operation| Specialization|      |      |      |      |   |           |           |       |              |              |        K0_M0_M1_K1|          K0_M0_M1_K1|   ArrangeOrder|          Order| Lengths_K0_M0_M1_K1| ContiguousDimOrder| Lengths_K0_M0_M1_K1|        K0_N0_N1_K1|          K0_N0_N1_K1|   ArrangeOrder|          Order| Lengths_K0_N0_N1_K1| ContiguousDimOrder| Lengths_K0_N0_N1_K1|               Order|                |                   |
+//        ######|           |           |            |            |            |             |              |              |               |      |      |      |      |   |           |           |       |              |              |                   |                     |               |               |                    |                   |                    |                   |                     |               |               |                    |                   |                    |                    |                |                   |
+                 NDimSpatial, InDataType, WeiDataType, OutDataType, AccDataType,  InElementOp,  WeiElementOp,  OutElementOp, ConvBwdDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<8, 2>,       S<8, 2>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,       S<1, 1, 1, 2>,      S<1, 1, 8, 2>,      S<16, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 8, 1>,      S<0, 3, 1, 2>,       S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,               5,                 4>;
+// clang-format on
 
 int main(int argc, char* argv[])
 {
     namespace ctc = ck::tensor_layout::convolution;
 
+    print_helper_msg();
+
     bool do_verification = true;
     int init_method      = 1;
     bool time_kernel     = false;
 
     ck::utils::conv::ConvParam conv_param{
-        2, 1, 32, 256, 1024, {3, 3}, {14, 14}, {2, 2}, {1, 1}, {1, 1}, {1, 1}};
-
-    ck::index_t split_k = 4;
+        2, 1, 128, 256, 256, {3, 3}, {71, 71}, {2, 2}, {1, 1}, {1, 1}, {1, 1}};
 
     if(argc == 1)
     {
@@ -91,9 +61,6 @@ int main(int argc, char* argv[])
         const ck::index_t num_dim_spatial = std::stoi(argv[4]);
 
         conv_param = ck::utils::conv::parse_conv_param(num_dim_spatial, 5, argv);
-
-        split_k = std::stoi(argv[5 + 3 + 6 * num_dim_spatial - 1]);
-        split_k = std::max(1, split_k);
     }
 
     const auto in_element_op  = InElementOp{};
@@ -118,14 +85,14 @@ int main(int argc, char* argv[])
             ck::utils::conv::make_output_host_tensor_descriptor_g_n_k_wos_packed<OutLayout>(
                 conv_param);
 
-        return run_conv_bwd_weight<1,
+        return run_conv_bwd_data<1,
                                  InDataType,
                                  WeiDataType,
                                  OutDataType,
                                  InElementOp,
                                  WeiElementOp,
                                  OutElementOp,
-                                   DeviceConvndBwdWeightInstance<1>>(do_verification,
+                                 DeviceConvNdBwdDataInstance<1>>(do_verification,
                                                                  init_method,
                                                                  time_kernel,
                                                                  conv_param,
@@ -134,8 +101,7 @@ int main(int argc, char* argv[])
                                                                  out_g_n_k_wos_desc,
                                                                  in_element_op,
                                                                  wei_element_op,
-                                                                     out_element_op,
-                                                                     split_k);
+                                                                 out_element_op);
     }
     else if(conv_param.num_dim_spatial_ == 2)
     {
@@ -155,14 +121,14 @@ int main(int argc, char* argv[])
             ck::utils::conv::make_output_host_tensor_descriptor_g_n_k_wos_packed<OutLayout>(
                 conv_param);
 
-        return run_conv_bwd_weight<2,
+        return run_conv_bwd_data<2,
                                  InDataType,
                                  WeiDataType,
                                  OutDataType,
                                  InElementOp,
                                  WeiElementOp,
                                  OutElementOp,
-                                   DeviceConvndBwdWeightInstance<2>>(do_verification,
+                                 DeviceConvNdBwdDataInstance<2>>(do_verification,
                                                                  init_method,
                                                                  time_kernel,
                                                                  conv_param,
@@ -171,8 +137,7 @@ int main(int argc, char* argv[])
                                                                  out_g_n_k_wos_desc,
                                                                  in_element_op,
                                                                  wei_element_op,
-                                                                     out_element_op,
-                                                                     split_k);
+                                                                 out_element_op);
     }
     else if(conv_param.num_dim_spatial_ == 3)
     {
@@ -192,14 +157,14 @@ int main(int argc, char* argv[])
             ck::utils::conv::make_output_host_tensor_descriptor_g_n_k_wos_packed<OutLayout>(
                 conv_param);
 
-        return run_conv_bwd_weight<3,
+        return run_conv_bwd_data<3,
                                  InDataType,
                                  WeiDataType,
                                  OutDataType,
                                  InElementOp,
                                  WeiElementOp,
                                  OutElementOp,
-                                   DeviceConvndBwdWeightInstance<3>>(do_verification,
+                                 DeviceConvNdBwdDataInstance<3>>(do_verification,
                                                                  init_method,
                                                                  time_kernel,
                                                                  conv_param,
@@ -208,8 +173,7 @@ int main(int argc, char* argv[])
                                                                  out_g_n_k_wos_desc,
                                                                  in_element_op,
                                                                  wei_element_op,
-                                                                     out_element_op,
-                                                                     split_k);
+                                                                 out_element_op);
     }
 
     return 0;