[mlir][SparseTensor] Fix invalid IR in ForallRewriter pattern

[matthias-springer]

The ForallRewriter pattern used to generate invalid IR:

mlir/test/Dialect/SparseTensor/GPU/gpu_combi.mlir:0:0: error: 'scf.for' op expects region #0 to have 0 or 1 blocks
mlir/test/Dialect/SparseTensor/GPU/gpu_combi.mlir:0:0: note: see current operation:
"scf.for"(%8, %2, %9) ({
^bb0(%arg5: index):
  // ...
  "scf.yield"() : () -> ()
^bb1(%10: index):  // no predecessors
  "scf.yield"() : () -> ()
}) : (index, index, index) -> ()

This commit fixes tests such as mlir/test/Dialect/SparseTensor/GPU/gpu_combi.mlir when verifying the IR after each pattern application (#74270).

[llvmbot]

@llvm/pr-subscribers-mlir

@llvm/pr-subscribers-mlir-sparse

Author: Matthias Springer (matthias-springer)

Changes

The ForallRewriter pattern used to generate invalid IR:

mlir/test/Dialect/SparseTensor/GPU/gpu_combi.mlir:0:0: error: 'scf.for' op expects region #<!-- -->0 to have 0 or 1 blocks
mlir/test/Dialect/SparseTensor/GPU/gpu_combi.mlir:0:0: note: see current operation:
"scf.for"(%8, %2, %9) ({
^bb0(%arg5: index):
  // ...
  "scf.yield"() : () -&gt; ()
^bb1(%10: index):  // no predecessors
  "scf.yield"() : () -&gt; ()
}) : (index, index, index) -&gt; ()

This commit fixes tests such as mlir/test/Dialect/SparseTensor/GPU/gpu_combi.mlir when verifying the IR after each pattern application (#74270).

---

Full diff: https://github.com/llvm/llvm-project/pull/74547.diff

1 Files Affected:

• (modified) mlir/lib/Dialect/SparseTensor/Transforms/SparseGPUCodegen.cpp (+4)

diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/SparseGPUCodegen.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/SparseGPUCodegen.cpp
index 927fc71f84a85..5155cab772d44 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/SparseGPUCodegen.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/SparseGPUCodegen.cpp
@@ -309,6 +309,10 @@ static void genGPUCode(PatternRewriter &rewriter, gpu::GPUFuncOp gpuFunc,
   //   }
   Value upper = irMap.lookup(forallOp.getUpperBound()[0]);
   scf::ForOp forOp = rewriter.create<scf::ForOp>(loc, row, upper, inc);
+  // The scf.for builder creates an empty block. scf.for does not allow multiple
+  // blocks in its region, so delete the block before `cloneRegionBefore` adds
+  // an additional block.
+  rewriter.eraseBlock(forOp.getBody());
   rewriter.cloneRegionBefore(forallOp.getRegion(), forOp.getRegion(),
                              forOp.getRegion().begin(), irMap);