self review

build_tools/ci/cpu_comparison/test_files/matmul_f32_8_4_8.mlir

@@ -3,11 +3,11 @@
 //
 // This test is a part of three test files that developers can look at together.
 // 1. two_matmul_switching.mlir => switches calls between two matmuls M,N,K size (8,4,8) and (8,8,4) 
-      4 times each (8 calls total). The graph ((8,4,8) -> (8,8,4)) x 4
+//    4 times each (8 calls total). The graph ((8,4,8) -> (8,8,4)) x 4
 // 2. matmul_f32_8_4_8.mlir => calls the (8,4,8) matmul 4 times hence doesnt have a switching cost so we
-      have a baseline for it. The graph is (8,4,8) x 4
+//    have a baseline for it. The graph is (8,4,8) x 4
 // 3. matmul_f32_8_8_4.mlir => calls the (8,8,4) matmul 4 times hence doesnt have a switching cost so we
-      have a baseline for it. The graph is (8,8,4) x 4
+//    have a baseline for it. The graph is (8,8,4) x 4
 
 // These 2 lines are required by the script which generates input data:
 //

build_tools/ci/cpu_comparison/test_files/matmul_f32_8_8_4.mlir

@@ -2,11 +2,11 @@
 // and we successively call the same matmul
 // This test is a part of three test files that developers can look at together.
 // 1. two_matmul_switching.mlir => switches calls between two matmuls M,N,K size (8,4,8) and (8,8,4) 
-      4 times each (8 calls total). The graph ((8,4,8) -> (8,8,4)) x 4
+//    4 times each (8 calls total). The graph ((8,4,8) -> (8,8,4)) x 4
 // 2. matmul_f32_8_4_8.mlir => calls the (8,4,8) matmul 4 times hence doesnt have a switching cost so we
-      have a baseline for it. The graph is (8,4,8) x 4
+//    have a baseline for it. The graph is (8,4,8) x 4
 // 3. matmul_f32_8_8_4.mlir => calls the (8,8,4) matmul 4 times hence doesnt have a switching cost so we
-      have a baseline for it. The graph is (8,8,4) x 4
+//    have a baseline for it. The graph is (8,8,4) x 4
 
 // These 2 lines are required by the script which generates input data:
 //

build_tools/ci/cpu_comparison/test_files/two_matmul_switching.mlir

@@ -1,11 +1,11 @@
 // This test shows switching between two matmuls and is useful to model the switching cost.
 // This test is a part of three test files that developers can look at together.
 // 1. two_matmul_switching.mlir => switches calls between two matmuls M,N,K size (8,4,8) and (8,8,4) 
-      4 times each (8 calls total). The graph ((8,4,8) -> (8,8,4)) x 4
-// 2. matmul_f32_8_4_8 => calls the (8,4,8) matmul 4 times hence doesnt have a switching cost so we
-      have a baseline for it. The graph is (8,4,8) x 4
-// 3. matmul_f32_8_8_4 => calls the (8,8,4) matmul 4 times hence doesnt have a switching cost so we
-      have a baseline for it. The graph is (8,8,4) x 4
+//    4 times each (8 calls total). The graph ((8,4,8) -> (8,8,4)) x 4
+// 2. matmul_f32_8_4_8.mlir => calls the (8,4,8) matmul 4 times hence doesnt have a switching cost so we
+//    have a baseline for it. The graph is (8,4,8) x 4
+// 3. matmul_f32_8_8_4.mlir => calls the (8,8,4) matmul 4 times hence doesnt have a switching cost so we
+//    have a baseline for it. The graph is (8,8,4) x 4
 
 // These 2 lines are required by the script which generates input data:
 //

compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIELowerFuncArgs.cpp

@@ -16,8 +16,6 @@
 #include "iree-amd-aie/Transforms/AMDAIEUtils.h"
 #include "iree-amd-aie/Transforms/Passes.h"
 #include "iree-amd-aie/Transforms/Transforms.h"
-#include "mlir/Dialect/Linalg/IR/Linalg.h"
-#include "mlir/IR/IRMapping.h"
 #include "mlir/IR/Iterators.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 

compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.cpp

@@ -765,7 +765,6 @@ void addMLIRAIRLoweringPasses(OpPassManager &passManager) {
         xilinx::air::createAIRUnrollOuterPerfectlyNestedLoopsPass(options));
   }
   passManager.addPass(mlir::affine::createAffineExpandIndexOpsPass());
-
   passManager.addPass(createAMDAIELowerFuncArgsPass());
   passManager.addPass(xilinx::airrt::createAIRRtToNpuPass());
   passManager.addPass(createCanonicalizerPass());