Merge remote-tracking branch 'origin/develop' into aosewski/ggemm_multi_d2

example/30_grouped_conv_fwd_multiple_d/grouped_conv_fwd_bias_relu_add_xdl_int4.cpp

 // SPDX-License-Identifier: MIT
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
-#ifndef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
-#error Should compile this file with ck::int4_t support
-#endif
+#ifdef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
 
 #include "common.hpp"
 
@@ -29,3 +27,4 @@ using OutElementOp = ck::tensor_operation::element_wise::AddReluAdd;
 #include "run_grouped_conv_fwd_bias_relu_add_example.inc"
 
 int main(int argc, char* argv[]) { return !run_grouped_conv_fwd_bias_relu_add_example(argc, argv); }
+#endif

example/31_batched_gemm_gemm/batched_gemm_gemm_xdl_int4.cpp

@@ -9,9 +9,7 @@ Gemm + Gemm fused operation. Computes C_m_o = A_m_k * B0_k_n * B1_n_o
                                                        Gemm1
 */
 
-#ifndef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
-#error Should compile this file with ck::int4_t support
-#endif
+#ifdef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
 
 #include <iostream>
 #include <numeric>
@@ -144,3 +142,4 @@ static_assert(sizeof(ck::int4_t) == sizeof(int8_t));
 #endif
 
 int main(int argc, char* argv[]) { return run_batched_gemm_gemm_example(argc, argv) ? 0 : 1; }
+#endif

example/35_splitK_gemm/run_splitK_gemm_example.inc

@@ -157,7 +157,7 @@ bool run_splitK_gemm(const ProblemSize& problem_size, const ExecutionConfig& con
 
     if(config.time_kernel)
     {
-        float ave_time = invoker.Run(argument, StreamConfig{nullptr, config.time_kernel});
+        float ave_time = invoker.Run(argument, StreamConfig{nullptr, config.time_kernel, 1});
 
         std::size_t flop = std::size_t(2) * M * N * K;
         std::size_t num_btype =

example/35_splitK_gemm/splitK_gemm_xdl_fp16.cpp

@@ -42,7 +42,7 @@ using AElementOp = PassThrough;
 using BElementOp = PassThrough;
 using CElementOp = PassThrough;
 
-static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
+static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::KPadding;
 
 using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmXdlSplitKCShuffle
     // clang-format off

example/41_grouped_conv_conv_fwd/grouped_conv_conv_fwd_xdl_int4.cpp

 // SPDX-License-Identifier: MIT
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
-#ifndef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
-#error Should compile this file with ck::int4_t support
-#endif
+#ifdef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
 
 #include <cstdlib>
 #include <iostream>
@@ -120,3 +118,4 @@ static_assert(sizeof(ck::int4_t) == sizeof(int8_t));
 #endif
 
 int main(int argc, char* argv[]) { return run_grouped_conv_conv_fwd_example(argc, argv) ? 0 : 1; }
+#endif

example/44_elementwise_permute/elementwise_permute.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
 #include <iostream>
 #include <cstdlib>
 

example/44_elementwise_permute/elementwise_permute_3d.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
 #include <iostream>
 #include <cstdlib>
 
@@ -14,8 +17,8 @@
 using F16 = ck::half_t;
 using F32 = float;
 
-using ADataType = F16;
-using BDataType = F16;
+using ADataType = F32;
+using BDataType = F32;
 
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 using DeviceElementwisePermuteInstance =
@@ -25,10 +28,10 @@ using DeviceElementwisePermuteInstance =
                                                           2,                    // NumDim_m, {N, C}
                                                           2,                    // NumDim_n, {H, W}
                                                           1,                    // NumDim_k, {D}
-                                                          8,                    // MPerThread
-                                                          8,                    // NPerThread
-                                                          8,                    // KPerThread
-                                                          ck::Sequence<8>,  // InScalarPerVectorSeq
+                                                          4,                    // MPerThread
+                                                          4,                    // NPerThread
+                                                          4,                    // KPerThread
+                                                          ck::Sequence<4>,  // InScalarPerVectorSeq
                                                           ck::Sequence<4>>; // OutScalarPerVectorSeq
 
 template <typename HostTensorA, typename HostTensorB, typename Functor>

example/44_elementwise_permute/elementwise_permute_4D_fp16.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
 #include <iostream>
 #include <cstdlib>
 

example/44_elementwise_permute/elementwise_permute_4D_fp16_2d.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
 #include <iostream>
 #include <cstdlib>
 

example/44_elementwise_permute/elementwise_permute_4D_fp16_col.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
 #include <iostream>
 #include <cstdlib>
 #include <random>

example/44_elementwise_permute/elementwise_permute_4D_fp16_row.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
 #include <iostream>
 #include <cstdlib>
 

example/44_elementwise_permute/elementwise_permute_4D_fp32_col.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
 #include <iostream>
 #include <cstdlib>
 

example/44_elementwise_permute/elementwise_permute_4D_fp32_row.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
 #include <iostream>
 #include <cstdlib>
 

example/48_pool3d_fwd/pool3d_fwd_common.hpp

@@ -32,6 +32,8 @@ std::vector<ck::index_t> f_tensor_strides_ncdhw(ck::index_t N_,
         return {C_ * D * H * W, D * H * W, H * W, W, 1_uz};
     else if constexpr(ck::is_same<decltype(layout), ck::tensor_layout::convolution::NDHWC>::value)
         return {D * C_ * H * W, 1_uz, C_ * H * W, W * C_, C_};
+    throw std::runtime_error("Pool3d_fwd: problem with layout. ");
+    return {0, 0, 0, 0, 0};
 };
 
 template <typename TensorLayout>
@@ -53,6 +55,8 @@ HostTensorDescriptor f_host_tensor_descriptor(std::size_t N_,
         return HostTensorDescriptor({N_, C_, D, H, W},
                                     {D * C_ * H * W, 1_uz, C_ * H * W, W * C_, C_});
     }
+    throw std::runtime_error("Pool3d_fwd: problem with layout. ");
+    return HostTensorDescriptor({0, 0, 0, 0, 0}, {0, 0, 0, 0, 0});
 };
 
 template <typename DevicePoolFwdInstance,

example/51_avgpool3d_bwd/avgpool3d_bwd_common.hpp

@@ -26,6 +26,8 @@ std::vector<ck::index_t> f_tensor_strides_ncdhw(ck::index_t N_,
         return {C_ * D * H * W, D * H * W, H * W, W, 1_uz};
     else if constexpr(ck::is_same<decltype(layout), ck::tensor_layout::convolution::NDHWC>::value)
         return {D * C_ * H * W, 1_uz, C_ * H * W, W * C_, C_};
+    throw std::runtime_error("Avgpool3d_bwd: problem with layout. ");
+    return {0, 0, 0, 0, 0};
 };
 
 template <typename TensorLayout>
@@ -47,6 +49,8 @@ HostTensorDescriptor f_host_tensor_descriptor(std::size_t N_,
         return HostTensorDescriptor({N_, C_, D, H, W},
                                     {D * C_ * H * W, 1_uz, C_ * H * W, W * C_, C_});
     }
+    throw std::runtime_error("Avgpool3d_bwd: problem with layout. ");
+    return HostTensorDescriptor({0, 0, 0, 0, 0}, {0, 0, 0, 0, 0});
 };
 
 template <typename DevicePoolBwdInstance,

include/ck/ck.hpp

@@ -213,12 +213,12 @@
 #define CK_WORKAROUND_SWDEV_388832 1
 
 // flag to enable (1) or disable (0) the debugging output in some kernels
-#define DEBUG_LOG 1
+#define DEBUG_LOG 0
 
 // denorm test fix, required to work around dissue
 #ifndef CK_WORKAROUND_DENORM_FIX
 #define CK_WORKAROUND_DENORM_FIX 0
-#elif
+#else
 // enable only on MI200
 #define CK_WORKAROUND_DENORM_FIX = CK_WORKAROUND_DENORM_FIX && defined(__gfx90a__)
 #endif // CK_WORKAROUND_DENORM_FIX

include/ck/host_utility/hip_check_error.hpp

@@ -12,21 +12,23 @@ inline void hip_check_error(hipError_t x)
     if(x != hipSuccess)
     {
         std::ostringstream ss;
-        ss << "HIP runtime error: " << hipGetErrorString(x) << ". " << __FILE__ << ": " << __LINE__
-           << "in function: " << __func__;
+        ss << "HIP runtime error: " << hipGetErrorString(x) << ". "
+           << "hip_check_error.hpp"
+           << ": " << __LINE__ << "in function: " << __func__;
         throw std::runtime_error(ss.str());
     }
 }
 
-#define HIP_CHECK_ERROR(retval_or_funcall)                                         \
-    do                                                                             \
-    {                                                                              \
-        hipError_t _tmpVal = retval_or_funcall;                                    \
-        if(_tmpVal != hipSuccess)                                                  \
-        {                                                                          \
-            std::ostringstream ostr;                                               \
-            ostr << "HIP Function Failed (" << __FILE__ << "," << __LINE__ << ") " \
-                 << hipGetErrorString(_tmpVal);                                    \
-            throw std::runtime_error(ostr.str());                                  \
-        }                                                                          \
+#define HIP_CHECK_ERROR(retval_or_funcall)                                 \
+    do                                                                     \
+    {                                                                      \
+        hipError_t _tmpVal = retval_or_funcall;                            \
+        if(_tmpVal != hipSuccess)                                          \
+        {                                                                  \
+            std::ostringstream ostr;                                       \
+            ostr << "HIP Function Failed ("                                \
+                 << "hip_check_error.hpp"                                  \
+                 << "," << __LINE__ << ") " << hipGetErrorString(_tmpVal); \
+            throw std::runtime_error(ostr.str());                          \
+        }                                                                  \
     } while(0)

include/ck/host_utility/kernel_launch.hpp

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -30,7 +30,7 @@ float launch_and_time_kernel(const StreamConfig& stream_config,
                block_dim.y,
                block_dim.z);
 
-        printf("Warm up 1 time\n");
+        printf("Warm up %d times\n", stream_config.cold_niters_);
 #endif
         // warm up
         for(int i = 0; i < stream_config.cold_niters_; ++i)

include/ck/stream_config.hpp

@@ -11,6 +11,6 @@ struct StreamConfig
     hipStream_t stream_id_ = nullptr;
     bool time_kernel_      = false;
     int log_level_         = 0;
-    int cold_niters_       = 1;
-    int nrepeat_           = 10;
+    int cold_niters_       = 5;
+    int nrepeat_           = 50;
 };