sync from public repo

example/ck_tile/05_reduce/reduce.cpp

+#include "ck_tile/host.hpp"
+#include "reduce.hpp"
+#include <cstring>
+
+auto create_args(int argc, char* argv[])
+{
+    ck_tile::ArgParser arg_parser;
+    arg_parser.insert("m", "3328", "m dimension")
+        .insert("n", "4096", "n dimension")
+        .insert("v", "1", "cpu validation or not")
+        .insert("prec", "fp16", "precision")
+        .insert("warmup", "5", "cold iter")
+        .insert("repeat", "20", "hot iter");
+
+    bool result = arg_parser.parse(argc, argv);
+    return std::make_tuple(result, arg_parser);
+}
+
+template <typename DataType>
+bool run(const ck_tile::ArgParser& arg_parser)
+{
+    using XDataType       = DataType;
+    using ComputeDataType = float;
+    using YDataType       = DataType;
+
+    ck_tile::index_t m = arg_parser.get_int("m");
+    ck_tile::index_t n = arg_parser.get_int("n");
+    int do_validation  = arg_parser.get_int("v");
+    int warmup         = arg_parser.get_int("warmup");
+    int repeat         = arg_parser.get_int("repeat");
+
+    ck_tile::HostTensor<XDataType> x_host({m, n});
+    ck_tile::HostTensor<YDataType> y_host_ref({m});
+    ck_tile::HostTensor<YDataType> y_host_dev({m});
+
+    ck_tile::FillUniformDistribution<XDataType>{-5.f, 5.f}(x_host);
+
+    ck_tile::DeviceMem x_buf(x_host.get_element_space_size_in_bytes());
+    ck_tile::DeviceMem y_buf(y_host_dev.get_element_space_size_in_bytes());
+
+    x_buf.ToDevice(x_host.data());
+
+    using ReduceOp   = ck_tile::ReduceOp::Add;
+    using BlockWarps = ck_tile::sequence<4, 1>;
+    using BlockTile  = ck_tile::sequence<128, 128>;
+    using WarpTile   = ck_tile::sequence<32, 128>;
+    using Vector     = ck_tile::sequence<8, 8>;
+
+    // cross warp-reduce
+    // using BlockWarps = ck_tile::sequence<2, 2>;
+    // using BlockTile  = ck_tile::sequence<2, 1024>;
+    // using WarpTile   = ck_tile::sequence<1, 512>;
+    // using Vector = ck_tile::sequence<1, 8>;
+
+    constexpr ck_tile::index_t kBlockSize  = 512;
+    constexpr ck_tile::index_t kBlockPerCu = 1;
+    ck_tile::index_t kGridSize             = (m / BlockTile::at(ck_tile::number<0>{}));
+    std::cout << "grid size " << kGridSize << std::endl;
+
+    using Shape = ck_tile::Reduce2dShape<BlockWarps, BlockTile, WarpTile, Vector>;
+    using Porblem =
+        ck_tile::Reduce2dProblem<XDataType, ComputeDataType, YDataType, Shape, ReduceOp>;
+
+    using Kernel = ck_tile::Reduce<Porblem>;
+
+    float ave_time = launch_kernel(ck_tile::stream_config{nullptr, true, 0, warmup, repeat},
+                                   ck_tile::make_kernel<kBlockSize, kBlockPerCu>(
+                                       Kernel{},
+                                       kGridSize,
+                                       kBlockSize,
+                                       0,
+                                       static_cast<XDataType*>(x_buf.GetDeviceBuffer()),
+                                       static_cast<YDataType*>(y_buf.GetDeviceBuffer()),
+                                       m,
+                                       n));
+
+    std::size_t num_btype = sizeof(XDataType) * m * n + sizeof(YDataType) * m;
+
+    float gb_per_sec = num_btype / 1.E6 / ave_time;
+
+    std::cout << "Perf: " << ave_time << " ms, " << gb_per_sec << " GB/s" << std::endl;
+
+    bool pass = true;
+
+    if(do_validation)
+    {
+        // reference
+        ck_tile::reference_reduce<XDataType, ComputeDataType, YDataType>(
+            x_host, y_host_ref, ReduceOp{});
+        y_buf.FromDevice(y_host_dev.mData.data());
+        pass = ck_tile::check_err(y_host_dev, y_host_ref);
+
+        std::cout << "valid:" << (pass ? "y" : "n") << std::flush << std::endl;
+    }
+
+    return pass;
+}
+
+int main(int argc, char* argv[])
+{
+    auto [result, arg_parser] = create_args(argc, argv);
+    if(!result)
+        return -1;
+
+    const std::string data_type = arg_parser.get_str("prec");
+
+    if(data_type == "fp16")
+    {
+        return run<ck_tile::half_t>(arg_parser) ? 0 : -2;
+    }
+    // else if(data_type == "bf16")
+    // {
+    //     return run<ck_tile::bf16_t>(arg_parser) ? 0 : -2;
+    // }
+}

example/ck_tile/06_permute/CMakeLists.txt

+# not using add_example_executable() to add this target, since we don't want this to have
+# to be included in "make all/install/check"
+add_executable(tile_example_permute EXCLUDE_FROM_ALL permute.cpp)
+
+if(NOT DEFINED PERMUTE_USE_ALTERNATIVE_IMPL)
+# set(PERMUTE_USE_ALTERNATIVE_IMPL false)
+set(PERMUTE_USE_ALTERNATIVE_IMPL true)
+endif()
+if(PERMUTE_USE_ALTERNATIVE_IMPL)
+target_compile_options(tile_example_permute PRIVATE -DPERMUTE_USE_ALTERNATIVE_IMPL)
+target_sources(tile_example_permute PRIVATE alternative_impl/matrix_core_swizzle.cpp)
+endif()
+# target_compile_options(tile_example_permute PRIVATE -v --save-temps -Wno-gnu-line-marker)

example/ck_tile/06_permute/README.md

+# permute
+
+This folder contains example for permute kernel, which is similiar to [torch.permute](https://pytorch.org/docs/stable/generated/torch.permute.html) (combined with [torch.contiguous](https://pytorch.org/docs/stable/generated/torch.Tensor.contiguous.html)). Currently we implement a generic permute kernel that support up to rank 8 arbitrary permutation with a single kernel instance. Performance is not the first consideration, we prefer a simple and general kernel implementation using `ck_tile` in this example.
+
+
+```
+args:
+          -v    weather do CPU validation or not (default:1)
+       -prec    data type. fp16/bf16/fp32 (default:fp16)
+      -shape    the shape of the input tensor (default:2,3,4)
+       -perm    permute perm (default:2,1,0)
+```
+
+## build
+```
+# in the root of ck_tile
+mkdir build && cd build
+sh ../script/cmake-ck-dev.sh  ../ <arch>  # you can replace this <arch> to gfx90a, gfx942...
+make tile_example_permute -j
+```
+This will result in an executable `build/bin/tile_example_permute`
+
+
+## some examples
+```
+# torch
+x=torch.randn(2,3,4,6)
+y=x.permute(0,3,2,1).contiguous()
+
+# ck_tile
+./build/bin/tile_example_permute -shape=2,3,4,6 -perm=0,3,2,1
+```
+
+or you can try the smoke_test
+```
+# in the root of ck_tile, after you build this example
+sh example/ck_tile/06_permute/script/smoke_test.sh
+```
+
+### alternative implementation
+we have an alternative implementation under `alternative_impl/` folder, that can swizzle the tensor to be more friendly for data loading for matrix core layout. This can be enabled when dealing with a `rank-7` tensor, with a fixed pattern of either `0,1,4,2,5,3,6` or `0,1,2,4,5,3,6`. There are other shape limitation of this implementation, check the source code of `permute.cpp` for detail.
+```
+# example
+./build/bin/tile_example_permute -shape=3,6,4,32,16,2,8 -perm=0,1,4,2,5,3,6 # b_n0_k0_n1_k1_n2_k2
+./build/bin/tile_example_permute -shape=3,8,4,16,16,4,8 -perm=0,1,2,4,5,3,6 # b_n0_n1_k0_k1_n2_k2
+```

example/ck_tile/06_permute/alternative_impl/matrix_core_swizzle.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+#include "matrix_core_swizzle_kernel.hpp"
+#include <string>
+
+struct matrix_core_swizzle_traits
+{
+    std::string data_type; // fp16 only
+    std::string inst;      // 32x32x8, 16x16x16
+    std::string permute;   //
+};
+
+using matrix_core_swizzle_args = matrix_core_swizzle_host_args;
+
+// host API
+float matrix_core_swizzle(matrix_core_swizzle_traits,
+                          matrix_core_swizzle_args,
+                          const ck_tile::stream_config&);