Merge branch 'develop' into wmma_op

test/convnd_bwd_weight/convnd_bwd_weight.cpp → test/grouped_convnd_bwd_weight/grouped_convnd_bwd_weight.cpp

@@ -4,14 +4,15 @@
 #include <cstdlib>
 #include <iostream>
 #include <initializer_list>
-#include <vector>
 #include <tuple>
+#include <vector>
+
 #include <gtest/gtest.h>
 
-#include "profiler/include/profile_conv_bwd_weight_impl.hpp"
+#include "profiler/include/profile_grouped_conv_bwd_weight_impl.hpp"
 
 template <typename Tuple>
-class TestConvndBwdWeight : public ::testing::Test
+class TestGroupedConvndBwdWeight : public ::testing::Test
 {
     protected:
     using DataType = std::tuple_element_t<0, Tuple>;
@@ -25,20 +26,20 @@ class TestConvndBwdWeight : public ::testing::Test
         {
             bool pass;
             EXPECT_FALSE(conv_params.empty());
-            pass = ck::profiler::profile_conv_bwd_weight_impl<
+            pass = ck::profiler::profile_grouped_conv_bwd_weight_impl<
                 NDimSpatial,
                 ck::tuple_element_t<NDimSpatial - 1,
-                                    ck::Tuple<ck::tensor_layout::convolution::NWC,
-                                              ck::tensor_layout::convolution::NHWC,
-                                              ck::tensor_layout::convolution::NDHWC>>,
+                                    ck::Tuple<ck::tensor_layout::convolution::GNWC,
+                                              ck::tensor_layout::convolution::GNHWC,
+                                              ck::tensor_layout::convolution::GNDHWC>>,
                 ck::tuple_element_t<NDimSpatial - 1,
-                                    ck::Tuple<ck::tensor_layout::convolution::KXC,
-                                              ck::tensor_layout::convolution::KYXC,
-                                              ck::tensor_layout::convolution::KZYXC>>,
+                                    ck::Tuple<ck::tensor_layout::convolution::GKXC,
+                                              ck::tensor_layout::convolution::GKYXC,
+                                              ck::tensor_layout::convolution::GKZYXC>>,
                 ck::tuple_element_t<NDimSpatial - 1,
-                                    ck::Tuple<ck::tensor_layout::convolution::NWK,
-                                              ck::tensor_layout::convolution::NHWK,
-                                              ck::tensor_layout::convolution::NDHWK>>,
+                                    ck::Tuple<ck::tensor_layout::convolution::GNWK,
+                                              ck::tensor_layout::convolution::GNHWK,
+                                              ck::tensor_layout::convolution::GNDHWK>>,
                 DataType,
                 DataType,
                 DataType>(true,  // do_verification
@@ -54,37 +55,37 @@ class TestConvndBwdWeight : public ::testing::Test
 
 using KernelTypes =
     ::testing::Types<std::tuple<float>, std::tuple<ck::half_t>, std::tuple<ck::bhalf_t>>;
-TYPED_TEST_SUITE(TestConvndBwdWeight, KernelTypes);
+TYPED_TEST_SUITE(TestGroupedConvndBwdWeight, KernelTypes);
 
-TYPED_TEST(TestConvndBwdWeight, Test1D)
+TYPED_TEST(TestGroupedConvndBwdWeight, Test1D)
 {
     this->conv_params.clear();
-    this->conv_params.push_back({1, 1, 128, 128, 256, {1}, {14}, {2}, {1}, {0}, {0}});
-    this->conv_params.push_back({1, 1, 128, 128, 256, {3}, {28}, {1}, {1}, {1}, {1}});
-    this->conv_params.push_back({1, 1, 128, 128, 256, {1}, {3}, {1}, {1}, {0}, {0}});
+    this->conv_params.push_back({1, 4, 128, 128, 256, {1}, {14}, {2}, {1}, {0}, {0}});
+    this->conv_params.push_back({1, 4, 128, 128, 256, {3}, {28}, {1}, {1}, {1}, {1}});
+    this->conv_params.push_back({1, 4, 128, 128, 256, {1}, {3}, {1}, {1}, {0}, {0}});
     this->template Run<1>();
 }
 
-TYPED_TEST(TestConvndBwdWeight, Test2D)
+TYPED_TEST(TestGroupedConvndBwdWeight, Test2D)
 {
     this->conv_params.clear();
     this->conv_params.push_back(
-        {2, 1, 128, 128, 256, {1, 1}, {7, 7}, {2, 2}, {1, 1}, {0, 0}, {0, 0}});
+        {2, 4, 128, 128, 256, {1, 1}, {7, 7}, {2, 2}, {1, 1}, {0, 0}, {0, 0}});
     this->conv_params.push_back(
-        {2, 1, 32, 128, 256, {3, 3}, {14, 14}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
+        {2, 4, 32, 128, 256, {3, 3}, {14, 14}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
     this->conv_params.push_back(
-        {2, 1, 128, 128, 256, {1, 1}, {3, 3}, {1, 1}, {1, 1}, {0, 0}, {0, 0}});
+        {2, 4, 128, 128, 256, {1, 1}, {3, 3}, {1, 1}, {1, 1}, {0, 0}, {0, 0}});
     this->template Run<2>();
 }
 
-TYPED_TEST(TestConvndBwdWeight, Test3D)
+TYPED_TEST(TestGroupedConvndBwdWeight, Test3D)
 {
     this->conv_params.clear();
     this->conv_params.push_back(
-        {3, 1, 128, 128, 256, {1, 1, 1}, {7, 7, 7}, {2, 2, 2}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}});
+        {3, 4, 128, 128, 256, {1, 1, 1}, {7, 7, 7}, {2, 2, 2}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}});
     this->conv_params.push_back(
-        {3, 1, 32, 128, 256, {3, 3, 3}, {14, 14, 3}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}});
+        {3, 4, 32, 128, 256, {3, 3, 3}, {14, 14, 3}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}});
     this->conv_params.push_back(
-        {3, 1, 128, 128, 256, {1, 1, 1}, {3, 3, 3}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}});
+        {3, 4, 128, 128, 256, {1, 1, 1}, {3, 3, 3}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}});
     this->template Run<3>();
 }

test/reference_conv_fwd/reference_conv_fwd.cpp

@@ -12,6 +12,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 
+#include "ck/library/utility/algorithm.hpp"
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/fill.hpp"
 #include "ck/library/utility/host_tensor.hpp"
@@ -54,7 +55,7 @@ run_reference_convolution_forward(const ck::utils::conv::ConvParam& conv_param,
 
     fill_input_op(input.begin(), input.end());
     fill_weights_op(weights.begin(), weights.end());
-    std::fill(host_output.begin(), host_output.end(), OutDataType(0.f));
+    ck::ranges::fill<OutDataType>(host_output, 0.f);
 
     auto ref_conv     = ck::tensor_operation::host::ReferenceConvFwd<NDimSpatial,
                                                                  InDataType,
@@ -122,7 +123,7 @@ TEST(ReferenceConvolutionFWD, Conv2DGNHWC)
                                 508.5};
     EXPECT_TRUE(ck::utils::check_err(
         out_tensor.mDesc.GetLengths(), ref_dims, "Error: wrong output tensor dimensions!"));
-    EXPECT_TRUE(ck::utils::check_err(out_tensor.mData, ref_data, "Error: incorrect results!"));
+    EXPECT_TRUE(ck::utils::check_err(out_tensor, ref_data, "Error: incorrect results!"));
 }
 
 TEST(ReferenceConvolutionFWD, Conv2DGNHWCStridesDilationsPadding)
@@ -149,7 +150,7 @@ TEST(ReferenceConvolutionFWD, Conv2DGNHWCStridesDilationsPadding)
         1323., 1323., 2002.5, 2002.5, 2038.5, 2038.5, 2074.5, 2074.5, 2110.5, 2110.5};
     EXPECT_TRUE(ck::utils::check_err(
         out_tensor.mDesc.GetLengths(), ref_dims, "Error: wrong output tensor dimensions!"));
-    EXPECT_TRUE(ck::utils::check_err(out_tensor.mData, ref_data, "Error: incorrect results!"));
+    EXPECT_TRUE(ck::utils::check_err(out_tensor, ref_data, "Error: incorrect results!"));
 }
 
 TEST(ReferenceConvolutionFWD, Conv1DGNWC)
@@ -178,7 +179,7 @@ TEST(ReferenceConvolutionFWD, Conv1DGNWC)
     std::vector<float> ref_data{7.5, 13.5, 19.5, 25.5};
     EXPECT_TRUE(ck::utils::check_err(
         out_tensor.mDesc.GetLengths(), ref_dims, "Error: wrong output tensor dimensions!"));
-    EXPECT_TRUE(ck::utils::check_err(out_tensor.mData, ref_data, "Error: incorrect results!"));
+    EXPECT_TRUE(ck::utils::check_err(out_tensor, ref_data, "Error: incorrect results!"));
 }
 
 TEST(ReferenceConvolutionFWD, Conv1DGNWCStridesDilationsPadding)
@@ -207,7 +208,7 @@ TEST(ReferenceConvolutionFWD, Conv1DGNWCStridesDilationsPadding)
     std::vector<float> ref_data{9., 9., 19.5, 19.5, 31.5, 31.5, 43.5, 43.5, 55.5, 55.5};
     EXPECT_TRUE(ck::utils::check_err(
         out_tensor.mDesc.GetLengths(), ref_dims, "Error: wrong output tensor dimensions!"));
-    EXPECT_TRUE(ck::utils::check_err(out_tensor.mData, ref_data, "Error: incorrect results!"));
+    EXPECT_TRUE(ck::utils::check_err(out_tensor, ref_data, "Error: incorrect results!"));
 }
 
 TEST(ReferenceConvolutionFWD, Conv1DGNWCSameOutputSize)
@@ -301,7 +302,7 @@ TEST(ReferenceConvolutionFWD, Conv1DGNWCSameOutputSize)
         49.4,      49.4,      49.4,      49.4,      49.4,      49.4,      49.4,      49.4};
     EXPECT_TRUE(ck::utils::check_err(
         out_tensor2.mDesc.GetLengths(), ref_dims, "Error: wrong output tensor dimensions!"));
-    EXPECT_TRUE(ck::utils::check_err(out_tensor2.mData, ref_data, "Error: incorrect results!"));
+    EXPECT_TRUE(ck::utils::check_err(out_tensor2, ref_data, "Error: incorrect results!"));
 }
 #endif
 
@@ -340,8 +341,7 @@ TEST(ReferenceConvolutionFWD, Conv3DGNCDHW)
     EXPECT_TRUE(ck::utils::check_err(out_tensor.mDesc.GetLengths(),
                                      ref_dims,
                                      "Error [case 1]: wrong output tensor dimensions!"));
-    EXPECT_TRUE(
-        ck::utils::check_err(out_tensor.mData, ref_data, "Error [case 1]: incorrect results!"));
+    EXPECT_TRUE(ck::utils::check_err(out_tensor, ref_data, "Error [case 1]: incorrect results!"));
 }
 
 TEST(ReferenceConvolutionFWD, Conv3DGNCDHWStridesDilations)
@@ -388,5 +388,5 @@ TEST(ReferenceConvolutionFWD, Conv3DGNCDHWStridesDilations)
                                      ref_dims,
                                      "Error [case 2]: wrong output tensor dimensions!"));
     EXPECT_TRUE(ck::utils::check_err(
-        out_tensor.mData, ref_data, "Error [case 2]: incorrect results!", 1e-4f, 1e-6f));
+        out_tensor, ref_data, "Error [case 2]: incorrect results!", 1e-4f, 1e-6f));
 }