Add multiple output support for dynamic fusion
* The dependency graph now can schedule any acyclic graph into
a sequential list of operators. This is needed as the output
operators now form branches in the graph.
* Fix the definition of input, output and intermediate tensors
in GpuKernelComponentGroup to support non-linear but sequential
list of operators.
* Add constraint on GpuOperatorGroup to enforce strictly linear
fusion style, but allow output operator as the only form of
branch.
Resolves: COMPMID-5771
Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Change-Id: I68de3a31a2456145081f0a397e4e61dd66327682
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8823
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
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 0afd0e7..eed481f 100644
--- a/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateWriter.cpp
+++ b/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateWriter.cpp
@@ -203,9 +203,7 @@
}
std::string ClTemplateWriter::write_global_section() const
{
- const auto dst_tensors = _components.get_dst_tensors();
- ARM_COMPUTE_ERROR_ON_MSG(dst_tensors.size() != 1, "Only one destination tensor per kernel is allowed");
- const auto dst_info = dst_tensors[0];
+ const auto dst_info = _components.get_any_dst_tensor();
const auto dst_w = dst_info->dimension(0);
const auto tile_w = std::max(1, get_window().x().step());
const auto tile_h = std::max(1, get_window().y().step());