Implement dynamic fusion softmax operator

 - Return aux tensorInfo by get_aux_tensors() at runtime to init the aux
   tensor with the right size.
 - Keep softmax unfusable for this commit
 - Hence, added Tensor3D to template writer arguments declaration, for sake of
   keeping dynamic fusion softmax componenets' kernels matching their cl
   counterparts.

Resolves: COMPMID-5523
Change-Id: I667f39545db925f667036ef448302c79a0330373
Signed-off-by: Ramy Elgammal <ramy.elgammal@arm.com>
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/483924
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: bsgcomp <bsgcomp@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8986
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 11 insertions, 5 deletions

diff --git a/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateWriter.cpp b/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateWriter.cpp
index 2ab6316947..eda15f1d95 100644
--- a/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateWriter.cpp
+++ b/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateWriter.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2022 Arm Limited.
+ * Copyright (c) 2022-2023 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -180,7 +180,8 @@ std::string ClTemplateWriter::write_code()
     }
 
     auto arguments = _components.get_argument_tensors();
-    std::sort(arguments.begin(), arguments.end(), [](const ITensorInfo *l, const ITensorInfo *r) {
+    std::sort(arguments.begin(), arguments.end(), [](const ITensorInfo * l, const ITensorInfo * r)
+    {
         return l->id() < r->id();
     });
     code += write_kernel_signature(_vtable.get_variable_list(arguments));
@@ -192,16 +193,16 @@ std::string ClTemplateWriter::write_code()
     code += "    //------------------ END KERNEL_BUILDER_COORDINATE ---------------------\n";
 
     {
-        const auto tiles = _components.get_tiles();
+        const auto        tiles = _components.get_tiles();
         std::stringstream tiles_ss;
 
         tiles_ss << "    //------------------ START TILE DECLARATION ---------------------\n";
 
         for(auto tile : tiles)
         {
-            const auto var = _vtable.get_variable(tile);
+            const auto var       = _vtable.get_variable(tile);
             const auto data_type = get_cl_type_from_data_type(tile->data_type());
-            const auto var_name = var.uniq_name;
+            const auto var_name  = var.uniq_name;
 
             tiles_ss << "    TILE(" << data_type << ", M0, N0, " << var_name << ");\n";
         }
@@ -276,6 +277,11 @@ std::string ClTemplateWriter::write_argument_declaration(const GpuKernelVariable
             code += "\n    TENSOR4D_T(" + var.uniq_name + ", IMAGE)";
             break;
         }
+        case GpuKernelArgumentInfo::Type::Tensor_3D:
+        {
+            code += "\n    TENSOR3D_DECLARATION(" + var.uniq_name + ")";
+            break;
+        }
         default:
         {
             ARM_COMPUTE_ERROR("Unsupported declaration generation for GpuKernelArgumentInfo::Type");