huge refactor for high RewritePatterns usage and less ad-hoc cpp code

remove Spatial many ops in favor of tensor ops like in pim

src/PIM/Conversion/ONNXToSpatial/Common/ShapeTilingUtils.cpp

@@ -5,6 +5,8 @@
 
 #include "ShapeTilingUtils.hpp"
 #include "src/Accelerators/PIM/Compiler/PimCompilerOptions.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp"
 
 using namespace mlir;
 
@@ -30,10 +32,29 @@ SmallVector<Value> sliceTensor(
 
   for (int64_t i = 0; i < numSlices; i++) {
     offsets[axis] = rewriter.getIndexAttr(i * sliceSize);
-    if (i == numSlices - 1 && lastSliceSize != 0)
+    int64_t currentSliceSize = sliceSize;
+    if (i == numSlices - 1 && lastSliceSize != 0) {
+      currentSliceSize = lastSliceSize;
       sizes[axis] = rewriter.getIndexAttr(lastSliceSize);
+    }
 
-    Value slice = tensor::ExtractSliceOp::create(rewriter, loc, tensorToSlice, offsets, sizes, strides);
+    SmallVector<int64_t> sliceShape(shape.begin(), shape.end());
+    sliceShape[axis] = currentSliceSize;
+    auto sliceType =
+      RankedTensorType::get(sliceShape, cast<RankedTensorType>(tensorToSlice.getType()).getElementType());
+
+    Value slice;
+    if (isHostFoldableValue(tensorToSlice)) {
+      slice = tensor::ExtractSliceOp::create(rewriter, loc, tensorToSlice, offsets, sizes, strides);
+    }
+    else {
+      auto sliceCompute =
+        createSpatCompute<1>(rewriter, loc, TypeRange {sliceType}, {}, ValueRange {tensorToSlice}, [&](Value input) {
+          Value computedSlice = tensor::ExtractSliceOp::create(rewriter, loc, input, offsets, sizes, strides);
+          spatial::SpatYieldOp::create(rewriter, loc, computedSlice);
+        });
+      slice = sliceCompute.getResult(0);
+    }
     slices.push_back(slice);
   }