Merge branch 'develop' into bnorm_bwd_pr

example/30_grouped_conv_fwd_multiple_d/grouped_conv_fwd_xdl_fp16.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "common.hpp"
+
+// kernel data types
+using InKernelDataType  = FP16;
+using WeiKernelDataType = FP16;
+using AccDataType       = FP32;
+using CShuffleDataType  = FP16;
+using OutKernelDataType = FP16;
+
+// tensor data types
+using InUserDataType  = InKernelDataType;
+using WeiUserDataType = WeiKernelDataType;
+using OutUserDataType = OutKernelDataType;
+
+using InElementOp  = PassThrough;
+using WeiElementOp = PassThrough;
+using OutElementOp = PassThrough;
+
+#include "run_grouped_conv_fwd_example.inc"
+
+int main(int argc, char* argv[]) { return !run_grouped_conv_fwd_example(argc, argv); }

example/30_grouped_convnd_fwd_bias_relu_add/CMakeLists.txt

-add_example_executable(example_grouped_convnd_fwd_bias_relu_add_xdl_fp16 grouped_convnd_fwd_bias_relu_add_xdl_fp16.cpp)
-
-add_example_executable(example_grouped_convnd_fwd_bias_relu_add_xdl_fp32 grouped_convnd_fwd_bias_relu_add_xdl_fp32.cpp)
-
-add_example_executable(example_grouped_convnd_fwd_bias_relu_add_xdl_bf16 grouped_convnd_fwd_bias_relu_add_xdl_bf16.cpp)
-
-add_example_executable(example_grouped_convnd_fwd_bias_relu_add_xdl_int8 grouped_convnd_fwd_bias_relu_add_xdl_int8.cpp)
-
-if(USE_BITINT_EXTENSION_INT4)
-  add_example_executable(example_grouped_convnd_fwd_bias_relu_add_xdl_int4 grouped_convnd_fwd_bias_relu_add_xdl_int4.cpp)
-endif() # USE_BITINT_EXTENSION_INT4

example/30_grouped_convnd_fwd_bias_relu_add/README.md

-```bash
-#arg1: verification (0=no, 1=yes)
-#arg2: initialization (0=no init, 1=integer value, 2=decimal value)
-#arg3: time kernel (0=no, 1=yes)
-#Following arguments (depending on number of spatial dims):
-# N spatial dimensions
-# G, N, K, C,
-# <filter spatial dimensions>, (ie Y, X for 2D)
-# <input image spatial dimensions>, (ie Hi, Wi for 2D)
-# <strides>, (ie Sy, Sx for 2D)
-# <dilations>, (ie Dy, Dx for 2D)
-# <left padding>, (ie LeftPy, LeftPx for 2D)
-# <right padding>, (ie RightPy, RightPx for 2D)
-
-bin/example_grouped_convnd_fwd_bias_relu_add_xdl_fp16 1 1 1
-```
-
-Result (MI100)
-```
-in: dim 5, lengths {2, 128, 192, 71, 71}, strides {192, 1935744, 1, 27264, 384}
-wei: dim 5, lengths {2, 256, 192, 3, 3}, strides {442368, 1728, 1, 576, 192}
-bias: dim 5, lengths {2, 128, 256, 36, 36}, strides {256, 0, 1, 0, 0}
-residual: dim 5, lengths {2, 128, 256, 36, 36}, strides {256, 0, 1, 0, 0}
-out: dim 5, lengths {2, 128, 256, 36, 36}, strides {256, 663552, 1, 18432, 512}
-A[M, K]: {165888, 1728}
-B[N, K]: {256, 1728}
-Ds[M, N]: {165888, 256}
-Ds[M, N]: {165888, 256}
-E[M, N]: {165888, 256}
-launch_and_time_kernel: grid_dim {2592, 1, 1}, block_dim {256, 1, 1}
-Warm up 1 time
-Start running 10 times...
-Perf: 2.48075 ms, 118.325 TFlops, 268.946 GB/s, DeviceGroupedConvFwdMultipleD_Xdl_CShuffle<256, 128, 256, 32, Default>
-```
\ No newline at end of file

example/31_batched_gemm_gemm/batched_gemm_gemm_xdl_bf16.cpp

@@ -23,6 +23,7 @@ Gemm + Gemm fused operation. Computes C_m_o = A_m_k * B0_k_n * B1_n_o
 #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_batched_gemm.hpp"
 
 template <ck::index_t... Is>

example/31_batched_gemm_gemm/batched_gemm_gemm_xdl_fp16.cpp

@@ -23,6 +23,7 @@ Gemm + Gemm fused operation. Computes C_m_o = A_m_k * B0_k_n * B1_n_o
 #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_batched_gemm.hpp"
 
 template <ck::index_t... Is>

example/31_batched_gemm_gemm/batched_gemm_gemm_xdl_fp32.cpp

@@ -23,6 +23,7 @@ Gemm + Gemm fused operation. Computes C_m_o = A_m_k * B0_k_n * B1_n_o
 #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_batched_gemm.hpp"
 
 template <ck::index_t... Is>

example/31_batched_gemm_gemm/batched_gemm_gemm_xdl_int4.cpp

@@ -27,6 +27,7 @@ Gemm + Gemm fused operation. Computes C_m_o = A_m_k * B0_k_n * B1_n_o
 #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_batched_gemm.hpp"
 
 template <ck::index_t... Is>

example/31_batched_gemm_gemm/batched_gemm_gemm_xdl_int8.cpp

@@ -23,6 +23,7 @@ Gemm + Gemm fused operation. Computes C_m_o = A_m_k * B0_k_n * B1_n_o
 #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_batched_gemm.hpp"
 
 template <ck::index_t... Is>

example/31_batched_gemm_gemm/run_batched_gemm_gemm_example.inc

@@ -106,15 +106,15 @@ bool run_batched_gemm_gemm_example(int argc, char* argv[])
                                        std::size_t stride,
                                        std::size_t batch_stride,
                                        auto layout) {
+        using namespace ck::literals;
+
         if(std::is_same<decltype(layout), Row>::value)
         {
-            return HostTensorDescriptor(std::vector<std::size_t>({batch_count, row, col}),
-                                        std::vector<std::size_t>({batch_stride, stride, 1}));
+            return HostTensorDescriptor({batch_count, row, col}, {batch_stride, stride, 1_uz});
         }
         else
         {
-            return HostTensorDescriptor(std::vector<std::size_t>({batch_count, row, col}),
-                                        std::vector<std::size_t>({batch_stride, 1, stride}));
+            return HostTensorDescriptor({batch_count, row, col}, {batch_stride, 1_uz, stride});
         }
     };
 
@@ -270,7 +270,7 @@ bool run_batched_gemm_gemm_example(int argc, char* argv[])
         c_g_m_o_device_buf.FromDevice(c_g_m_o_device_result.mData.data());
 #endif
 
-        return ck::utils::check_err(c_g_m_o_device_result.mData, c_g_m_o_host_result.mData);
+        return ck::utils::check_err(c_g_m_o_device_result, c_g_m_o_host_result);
     }
 
     return true;

example/32_batched_gemm_scale_softmax_gemm/CMakeLists.txt

 add_example_executable(example_batched_gemm_scale_softmax_gemm_xdl_fp16 batched_gemm_scale_softmax_gemm_xdl_fp16.cpp)
+add_example_executable(example_batched_gemm_scale_softmax_gemm_xdl_bf16 batched_gemm_scale_softmax_gemm_xdl_bf16.cpp)
 add_example_executable(example_batched_gemm_scale_softmax_gemm_permute_xdl_fp16 batched_gemm_scale_softmax_gemm_permute_xdl_fp16.cpp)
+add_example_executable(example_batched_gemm_scale_softmax_gemm_permute_xdl_bf16 batched_gemm_scale_softmax_gemm_permute_xdl_bf16.cpp)
 add_example_executable(example_grouped_gemm_scale_softmax_gemm_permute_xdl_fp16 grouped_gemm_scale_softmax_gemm_permute_xdl_fp16.cpp)
 add_example_executable(example_batched_gemm_lower_triangle_scale_softmax_gemm_permute_xdl_fp16 batched_gemm_lower_triangle_scale_softmax_gemm_permute_xdl_fp16.cpp)
 add_example_executable(example_grouped_gemm_lower_triangle_scale_softmax_gemm_permute_xdl_fp16 grouped_gemm_lower_triangle_scale_softmax_gemm_permute_xdl_fp16.cpp)
 
 add_custom_target(example_gemm_scale_softmax_gemm)
 add_dependencies(example_gemm_scale_softmax_gemm example_batched_gemm_scale_softmax_gemm_xdl_fp16)
+add_dependencies(example_gemm_scale_softmax_gemm example_batched_gemm_scale_softmax_gemm_xdl_bf16)
 add_dependencies(example_gemm_scale_softmax_gemm example_batched_gemm_scale_softmax_gemm_permute_xdl_fp16)
+add_dependencies(example_gemm_scale_softmax_gemm example_batched_gemm_scale_softmax_gemm_permute_xdl_bf16)
 add_dependencies(example_gemm_scale_softmax_gemm example_grouped_gemm_scale_softmax_gemm_permute_xdl_fp16)
 add_dependencies(example_gemm_scale_softmax_gemm example_batched_gemm_lower_triangle_scale_softmax_gemm_permute_xdl_fp16)
 add_dependencies(example_gemm_scale_softmax_gemm example_grouped_gemm_lower_triangle_scale_softmax_gemm_permute_xdl_fp16)

example/32_batched_gemm_scale_softmax_gemm/batched_gemm_lower_triangle_scale_softmax_gemm_permute_xdl_fp16.cpp

@@ -24,6 +24,7 @@ Gemm + Softmax + Gemm fused operation. Computes C_g_m_o = Softmax(A_g_m_k * B0_g
 #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_batched_gemm.hpp"
 #include "ck/library/reference_tensor_operation/cpu/reference_softmax.hpp"
 

example/32_batched_gemm_scale_softmax_gemm/batched_gemm_scale_softmax_gemm_permute_xdl_fp16.cpp

@@ -24,6 +24,7 @@ Gemm + Softmax + Gemm fused operation. Computes C_g_m_o = Softmax(A_g_m_k * B0_g
 #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_batched_gemm.hpp"
 #include "ck/library/reference_tensor_operation/cpu/reference_softmax.hpp"