reduce spatial compile-times in convolutions using a scf.for instead of materializing a huge number of instructions

src/PIM/Conversion/SpatialToPim/SpatialToPimPass.cpp

@@ -1,5 +1,6 @@
 #include "mlir/Dialect/Arith/IR/Arith.h"
 #include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/Dialect/SCF/IR/SCF.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Dialect/Tosa/IR/TosaOps.h"
 #include "mlir/IR/BuiltinDialect.h"
@@ -134,7 +135,12 @@ void SpatialToPimPass::runOnOperation() {
   MLIRContext* ctx = moduleOp.getContext();
 
   ConversionTarget target(*ctx);
-  target.addLegalDialect<PimDialect, tensor::TensorDialect, arith::ArithDialect, func::FuncDialect, BuiltinDialect>();
+  target.addLegalDialect<PimDialect,
+                         tensor::TensorDialect,
+                         arith::ArithDialect,
+                         func::FuncDialect,
+                         scf::SCFDialect,
+                         BuiltinDialect>();
 
   RewritePatternSet patterns(ctx);
   populateWithGenerated(patterns);