[Mlir-commits] [mlir] [mlir][sparse] cleanup conversion test (PR #70356)

[Aart Bik]

https://github.com/aartbik created https://github.com/llvm/llvm-project/pull/70356

Various TODOs had been added that actually removed the actual test.
This puts the CHECK test backs and removes the TODOs that have no immediate plans.

>From a9aa52d25673b42cc4d4313b9d2c975794823620 Mon Sep 17 00:00:00 2001
From: Aart Bik <ajcbik at google.com>
Date: Thu, 26 Oct 2023 10:18:06 -0700
Subject: [PATCH] [mlir][sparse] cleanup conversion test

Various TODOs had been added that actually removed the actual
test. This moves the CHECK tests back (since improving conversions
is an ongoing task anyway that does not need to be reflected with
TODO's everywhere)
---
 .../SparseTensor/convert_sparse2sparse.mlir       | 15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/mlir/test/Dialect/SparseTensor/convert_sparse2sparse.mlir b/mlir/test/Dialect/SparseTensor/convert_sparse2sparse.mlir
index 658e8aa40022eb2..0280e27b4e312a0 100644
--- a/mlir/test/Dialect/SparseTensor/convert_sparse2sparse.mlir
+++ b/mlir/test/Dialect/SparseTensor/convert_sparse2sparse.mlir
@@ -41,25 +41,24 @@ func.func @sparse_nop_convert(%arg0: tensor<64xf32, #SparseVector>) -> tensor<64
 }
 
 // CHECK-LABEL:   func.func @sparse_hidden_nop_cast
-// TODO: The following convert should be a cast instead.
-// CHECK:           sparse_tensor.convert
-// CHECK:           return
+// CHECK-NEXT:      sparse_tensor.convert
+// CHECK-NEXT:      return
 func.func @sparse_hidden_nop_cast(%arg0: tensor<32xf32, #SparseVector>) -> tensor<?xf32, #SparseVector> {
   %0 = sparse_tensor.convert %arg0 : tensor<32xf32, #SparseVector> to tensor<?xf32, #SparseVector>
   return %0 : tensor<?xf32, #SparseVector>
 }
 
 // CHECK-LABEL:   func.func @sparse_convert_1d_ss(
-// TODO: libgen path need to support efficient format conversion (e.g., 32 bit pos -> 64 bit pos).
-// Maybe we should use a different operator as well to be clear.
+// CHECK-NEXT:      sparse_tensor.convert
+// CHECK-NEXT:      return
 func.func @sparse_convert_1d_ss(%arg0: tensor<?xf32, #SparseVector64>) -> tensor<?xf32, #SparseVector32> {
   %0 = sparse_tensor.convert %arg0 : tensor<?xf32, #SparseVector64> to tensor<?xf32, #SparseVector32>
   return %0 : tensor<?xf32, #SparseVector32>
 }
 
 // CHECK-LABEL:   func.func @sparse_convert(
-// TODO: libgen path need to support efficient format conversion (e.g., 32 bit pos -> 64 bit pos).
-// Maybe we should use a different operator as well to be clear.
+// CHECK-NEXT:      sparse_tensor.convert
+// CHECK-NEXT:      return
 func.func @sparse_convert(%arg0: tensor<?xf32, #SparseVector64>) -> tensor<?xf32, #SparseVector32> {
   %0 = sparse_tensor.convert %arg0 : tensor<?xf32, #SparseVector64> to tensor<?xf32, #SparseVector32>
   return %0 : tensor<?xf32, #SparseVector32>
@@ -73,6 +72,7 @@ func.func @sparse_convert(%arg0: tensor<?xf32, #SparseVector64>) -> tensor<?xf32
 // CHECK:           sparse_tensor.foreach
 // CHECK:             sparse_tensor.insert
 // CHECK:           sparse_tensor.load
+// CHECK:           return
 func.func @sparse_convert_permuted(%arg0: tensor<?x?x?xf32, #SortedCOO3D>) -> tensor<?x?x?xf32, #TsssPermuted> {
   %0 = sparse_tensor.convert %arg0 : tensor<?x?x?xf32, #SortedCOO3D> to tensor<?x?x?xf32, #TsssPermuted>
   return %0 : tensor<?x?x?xf32, #TsssPermuted>
@@ -83,6 +83,7 @@ func.func @sparse_convert_permuted(%arg0: tensor<?x?x?xf32, #SortedCOO3D>) -> te
 // CHECK:             sparse_tensor.insert
 // CHECK:           sparse_tensor.load
 // CHECK-NOT:       sparse_tensor.reorder_coo
+// CHECK:           return
 func.func @sparse_convert_slice(%arg0: tensor<2x13xi32, #COOSlice>) -> (tensor<2x13xi32, #SortedCOO2D>)  {
   %0 = sparse_tensor.convert %arg0 : tensor<2x13xi32, #COOSlice> to tensor<2x13xi32, #SortedCOO2D>
   return %0 : tensor<2x13xi32, #SortedCOO2D>