Improve external interface for GEMM and GEMM+add+add+fastgelu (#311)

* interface for GEMM and GEMM+add+add+fastgelu * rename namespace * instance factory * fix build * fix build; add GEMM client example * clean
2026-04-20 06:49:15 +00:00 · 2022-06-30 22:11:00 -05:00
parent fa9a0a5cfb
commit 0dcb3496cf
259 changed files with 2915 additions and 2969 deletions
--- a/test/gemm/gemm_bf16.cpp
+++ b/test/gemm/gemm_bf16.cpp
@@ -0,0 +1,79 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <algorithm>
+#include <cstdlib>
+#include <iostream>
+#include <numeric>
+#include <tuple>
+#include <vector>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/device_gemm.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "ck/library/tensor_operation_instance/gpu/gemm.hpp"
+
+#include "ck/library/utility/check_err.hpp"
+#include "ck/library/host_tensor/device_memory.hpp"
+#include "ck/library/host_tensor/host_tensor.hpp"
+#include "ck/library/host_tensor/host_tensor_generator.hpp"
+#include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
+
+#include "test/gemm/gemm_util.hpp"
+
+int main()
+{
+    using ADataType   = ck::bhalf_t;
+    using BDataType   = ck::bhalf_t;
+    using CDataType   = ck::bhalf_t;
+    using AccDataType = float;
+
+    using Row = ck::tensor_layout::gemm::RowMajor;
+    using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+    using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+    auto test = [&](auto a_layout, auto b_layout, auto c_layout) {
+        bool pass = true;
+
+        using DeviceOp = ck::tensor_operation::device::DeviceGemm<decltype(a_layout),
+                                                                  decltype(b_layout),
+                                                                  decltype(c_layout),
+                                                                  ADataType,
+                                                                  BDataType,
+                                                                  CDataType,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>;
+
+        const auto gemmPtrs =
+            ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
+                DeviceOp>::GetInstances();
+
+        for(auto& gemmPtr : gemmPtrs)
+        {
+            pass &= ck::gemm_util::TestGemm<std::unique_ptr<DeviceOp>,
+                                            ADataType,
+                                            BDataType,
+                                            CDataType,
+                                            AccDataType,
+                                            decltype(a_layout),
+                                            decltype(b_layout),
+                                            decltype(c_layout),
+                                            PassThrough,
+                                            PassThrough,
+                                            PassThrough>{}(gemmPtr);
+        }
+
+        return pass;
+    };
+
+    bool pass = test(Row{}, Row{}, Row{}) && test(Row{}, Col{}, Row{}) &&
+                test(Col{}, Row{}, Row{}) && test(Col{}, Col{}, Row{});
+
+    std::cout << "TestGemm ..... " << (pass ? "SUCCESS" : "FAILURE") << std::endl;
+    return pass ? 0 : 1;
+}