feat: optimize concrete mul with constant values

compiler/include/concretelang/Dialect/Concrete/CMakeLists.txt

@@ -1 +1,2 @@
 add_subdirectory(IR)
+add_subdirectory(Transforms)

compiler/include/concretelang/Dialect/Concrete/Transforms/CMakeLists.txt

@@ -0,0 +1,4 @@
+set(LLVM_TARGET_DEFINITIONS Optimization.td)
+mlir_tablegen(Optimization.h.inc -gen-pass-decls -name Transforms)
+add_public_tablegen_target(ConcretelangConcreteOptimizationPassIncGen)
+add_dependencies(mlir-headers ConcretelangConcreteOptimizationPassIncGen)

compiler/include/concretelang/Dialect/Concrete/Transforms/Optimization.h

@@ -0,0 +1,21 @@
+// Part of the Concrete Compiler Project, under the BSD3 License with Zama
+// Exceptions. See
+// https://github.com/zama-ai/concrete-compiler-internal/blob/master/LICENSE.txt
+// for license information.
+
+#ifndef CONCRETELANG_CONCRETE_OPTIMIZATION_PASS_H
+#define CONCRETELANG_CONCRETE_OPTIMIZATION_PASS_H
+
+#include <concretelang/Dialect/Concrete/IR/ConcreteDialect.h>
+#include <mlir/Pass/Pass.h>
+
+#define GEN_PASS_CLASSES
+#include <concretelang/Dialect/Concrete/Transforms/Optimization.h.inc>
+
+namespace mlir {
+namespace concretelang {
+std::unique_ptr<mlir::OperationPass<>> createConcreteOptimizationPass();
+} // namespace concretelang
+} // namespace mlir
+
+#endif

compiler/include/concretelang/Dialect/Concrete/Transforms/Optimization.td

@@ -0,0 +1,13 @@
+#ifndef CONCRETELANG_CONCRETE_OPTIMIZATION_PASS
+#define CONCRETELANG_CONCRETE_OPTIMIZATION_PASS
+
+include "mlir/Pass/PassBase.td"
+
+def ConcreteOptimization : Pass<"concrete-optimization"> {
+  let summary = "Optimize Concrete operations";
+  let constructor = "mlir::concretelang::createConcreteOptimizationPass()";
+  let options = [];
+  let dependentDialects = [ "mlir::concretelang::Concrete::ConcreteDialect" ];
+}
+
+#endif

compiler/include/concretelang/Support/Pipeline.h

@@ -47,6 +47,10 @@ lowerConcreteToBConcrete(mlir::MLIRContext &context, mlir::ModuleOp &module,
                          std::function<bool(mlir::Pass *)> enablePass,
                          bool parallelizeLoops);
 
+mlir::LogicalResult
+optimizeConcrete(mlir::MLIRContext &context, mlir::ModuleOp &module,
+                 std::function<bool(mlir::Pass *)> enablePass);
+
 mlir::LogicalResult
 lowerBConcreteToStd(mlir::MLIRContext &context, mlir::ModuleOp &module,
                     std::function<bool(mlir::Pass *)> enablePass);