Merge branch 'develop' of https://github.com/ROCmSoftwarePlatform/composable_kernel into wmma_gemm

test/grouped_convnd_fwd/grouped_convnd_fwd.cpp

@@ -7,7 +7,7 @@
 #include <vector>
 #include <gtest/gtest.h>
 
-#include "profiler/include/profile_grouped_conv_fwd_impl.hpp"
+#include "profiler/profile_grouped_conv_fwd_impl.hpp"
 
 class TestGroupedConvNdFwd : public ::testing::Test
 {

test/grouped_gemm/grouped_gemm_fp16.cpp

@@ -3,7 +3,7 @@
 
 #include <iostream>
 
-#include "profiler/include/profile_grouped_gemm_impl.hpp"
+#include "profiler/profile_grouped_gemm_impl.hpp"
 
 namespace {
 

test/normalization/test_groupnorm_fp16.cpp

@@ -2,7 +2,7 @@
 // Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "gtest/gtest.h"
-#include "profiler/include/profile_groupnorm_impl.hpp"
+#include "profiler/profile_groupnorm_impl.hpp"
 
 using F16 = ck::half_t;
 using F32 = float;

test/normalization/test_groupnorm_fp32.cpp

@@ -2,7 +2,7 @@
 // Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "gtest/gtest.h"
-#include "profiler/include/profile_groupnorm_impl.hpp"
+#include "profiler/profile_groupnorm_impl.hpp"
 
 using F16 = ck::half_t;
 using F32 = float;

test/normalization/test_layernorm2d_fp16.cpp

@@ -2,7 +2,7 @@
 // Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "gtest/gtest.h"
-#include "profiler/include/profile_layernorm_impl.hpp"
+#include "profiler/profile_layernorm_impl.hpp"
 
 using F16 = ck::half_t;
 using F32 = float;

test/normalization/test_layernorm2d_fp32.cpp

@@ -2,7 +2,7 @@
 // Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "gtest/gtest.h"
-#include "profiler/include/profile_layernorm_impl.hpp"
+#include "profiler/profile_layernorm_impl.hpp"
 
 using F16 = ck::half_t;
 using F32 = float;

test/reduce/reduce_no_index.cpp

@@ -4,7 +4,7 @@
 #include <getopt.h>
 
 #include "ck/library/utility/host_common_util.hpp"
-#include "profiler/include/profile_reduce_impl.hpp"
+#include "profiler/profile_reduce_impl.hpp"
 
 using namespace ck;
 

test/reduce/reduce_with_index.cpp

@@ -4,7 +4,7 @@
 #include <getopt.h>
 
 #include "ck/library/utility/host_common_util.hpp"
-#include "profiler/include/profile_reduce_impl.hpp"
+#include "profiler/profile_reduce_impl.hpp"
 
 using namespace ck;
 

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));
 }

test/softmax/test_softmax_util.hpp

@@ -13,7 +13,7 @@
 #include "ck/tensor_operation/gpu/device/impl/device_softmax_impl.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 #include "include/ck/utility/data_type.hpp"
-#include "profiler/include/profile_softmax_impl.hpp"
+#include "profiler/profile_softmax_impl.hpp"
 
 namespace ck {
 

test/wmma_op/wmma_op.cpp

@@ -60,9 +60,6 @@ int main(int, char*[])
     pass &= run_test<ck::half_t,  ck::half_t,  ck::half_t,  ck::half_t, 16    >();
     pass &= run_test<ck::bhalf_t, ck::bhalf_t, ck::bhalf_t, float,      16    >();
     pass &= run_test<int8_t,      int8_t,      int32_t,     int32_t,    8     >();
-#ifdef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
-    pass &= run_test<int4_t,      int4_t,      int32_t,     int32_t,    8     >();
-#endif
     // clang-format on
 
     std::cout << "TestGemm ..... " << (pass ? "SUCCESS" : "FAILURE") << std::endl;