Integrate Dynamic Fusion patches

* Add public interfaces:
    * OperatorGraph: Describe a workload that could contain fused kernels
    * IWorkload: Generic interface for workloads built from OperatorGraph
    * ClWorkload: OpenCL workloads built from OperatorGraph
    * ClCompositeOperator: Runtime async operator to execute a ClWorkload
    * DependencyGraph (will likely be deprecated in later iterations)

* Add example
    * cl_fused_conv2d_elementwise_add.cpp to explain how to use the new
      interfaces

* Add internal translation layer

* Refactor ClKernelBuildingAPI
    * Remove non-tile based gemm native kernel component
    * Minor interface changes

* Add integration tests

Resolves COMPMID-5161

Signed-off-by: SiCong Li <sicong.li@arm.com>
Change-Id: Ib987ed79289ab0bcbd3130d54f5793408d9f1240
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/7510
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 73 insertions, 0 deletions

diff --git a/src/core/experimental/dynamic_fusion/WorkloadImpl/ClWorkload.cpp b/src/core/experimental/dynamic_fusion/WorkloadImpl/ClWorkload.cpp
new file mode 100644
index 0000000000..e97cf88b79
--- /dev/null
+++ b/src/core/experimental/dynamic_fusion/WorkloadImpl/ClWorkload.cpp
@@ -0,0 +1,73 @@
+/*
+ * Copyright (c) 2022 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ENABLE_EXPERIMENTAL_DYNAMIC_FUSION
+#error "This experimental feature must be enabled with -DENABLE_EXPERIMENTAL_DYNAMIC_FUSION"
+#endif /* ENABLE_EXPERIMENTAL_DYNAMIC_FUSION */
+#include "arm_compute/core/experimental/ClWorkload.h"
+#include "src/core/experimental/dynamic_fusion/WorkloadImpl/ClFusedKernelGraph.h"
+#include "src/core/experimental/dynamic_fusion/WorkloadImpl/ClKernelGraph.h"
+#include "src/core/experimental/dynamic_fusion/WorkloadImpl/OperatorGraphImpl.h"
+
+namespace arm_compute
+{
+namespace experimental
+{
+namespace dynamic_fusion
+{
+Status build(ClWorkload &workload, const OperatorGraph &op_graph, const ClWorkloadContext &ctx)
+{
+    workload.context = ctx;
+    ClKernelGraph kernel_graph;
+    workload.status = validate(op_graph);
+    ARM_COMPUTE_RETURN_ON_ERROR(workload.status);
+    workload.status = translate(kernel_graph, *op_graph.impl());
+    ARM_COMPUTE_RETURN_ON_ERROR(workload.status);
+    ClFusedKernelGraph fused_k_graph;
+    std::tie(workload.status, fused_k_graph) = init_fusion_graph(kernel_graph);
+    ARM_COMPUTE_RETURN_ON_ERROR(workload.status);
+    workload.status = fuse(fused_k_graph);
+    ARM_COMPUTE_RETURN_ON_ERROR(workload.status);
+    workload.status = generate(workload, ctx, fused_k_graph);
+    ARM_COMPUTE_RETURN_ON_ERROR(workload.status);
+
+    // Get operator tensor id to workload tensor id map
+    const auto op_tensor_to_kernel_tensor       = fused_k_graph.original_graph->graph.get_merge_points();
+    const auto kernel_tensor_to_workload_tensor = workload.graph.get_merge_points();
+    for(const auto op_t : op_graph.impl()->graph.src_tensors())
+    {
+        const auto kernel_t                   = op_tensor_to_kernel_tensor.at(op_t);
+        const auto workload_t                 = kernel_tensor_to_workload_tensor.at(kernel_t);
+        workload.op_tensor_id_lut[workload_t] = op_t;
+    }
+    for(const auto op_t : op_graph.impl()->graph.dst_tensors())
+    {
+        const auto kernel_t                   = op_tensor_to_kernel_tensor.at(op_t);
+        const auto workload_t                 = kernel_tensor_to_workload_tensor.at(kernel_t);
+        workload.op_tensor_id_lut[workload_t] = op_t;
+    }
+    return workload.status;
+}
+} // namespace dynamic_fusion
+} // namespace experimental
+} // namespace arm_compute
\ No newline at end of file