From 04f4620cf999846a44089c81720aa920edec6993 Mon Sep 17 00:00:00 2001 From: Viet-Hoa Do Date: Wed, 14 Dec 2022 14:49:56 +0000 Subject: 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 Change-Id: I68de3a31a2456145081f0a397e4e61dd66327682 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8823 Reviewed-by: Gunes Bayir Tested-by: Arm Jenkins Comments-Addressed: Arm Jenkins --- src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateCast.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateCast.cpp') diff --git a/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateCast.cpp b/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateCast.cpp index 1ac49406a8..6ab3a68bb0 100644 --- a/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateCast.cpp +++ b/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateCast.cpp @@ -137,7 +137,7 @@ TagLUT ClTemplateCast::get_tag_lut(const GpuKernelVariableTable &vtable, const C lut["src"] = vtable.get_variable(_src); lut["dst"] = vtable.get_variable(_dst); - const auto dst_argument = vtable.get_variable(comp_group.get_dst_tensors()[0]); + const auto dst_argument = vtable.get_variable(comp_group.get_any_dst_tensor()); lut["arg_dst"] = dst_argument.uniq_name; // Local build options -- cgit v1.2.1