[logo] Constant folding skips TensorBroadcast (#7847)

* [logo] Constant folding skips TensorBroadcast

TensorBroadcast is a good hint for optimizaiton. Let's skip this for
constant folding.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Add TODO comment

diff --git a/compiler/logo/src/Passes/ConstantFoldingPass.cpp b/compiler/logo/src/Passes/ConstantFoldingPass.cpp
index c31b3c9..e038e71 100644 (file)
--- a/compiler/logo/src/Passes/ConstantFoldingPass.cpp
+++ b/compiler/logo/src/Passes/ConstantFoldingPass.cpp
@@ -61,6 +61,10 @@ bool skip(const loco::Node *node)
 
       // We don't perform constant folding for Push
       static_cast<uint32_t>(loco::CanonicalOpcode::Push),
+
+      // TensorBroadcast is a good hint for optimization
+      // TODO Let this option be controlled by driver using logo
+      static_cast<uint32_t>(loco::CanonicalOpcode::TensorBroadcast),
   };
 
   if (node->dialect() == loco::CanonicalDialect::get())