shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity

[mlir][sme] Use signed comparison in ArmSMEToSCF (#156834)

Browse Source 
This change switches from `ult` to `slt` in the comparison generated by
`TileLoadOpWithMaskAndPadNonZeroConversion`.

From the updated test:
```
%[[ROW_IS_ACTIVE:.*]] = arith.cmpi slt, %[[TILE_SLICE_INDEX]], %[[NUM_ROWS]] : index
```

Here:
- `%[[TILE_SLICE_INDEX]]` is always non-negative.
- `%[[NUM_ROWS]]` represents the number of remaining rows.

`%[[NUM_ROWS]]` is computed as:

```mlir
    %c-4 = arith.constant -4 : index
    %c-4_vscale = arith.muli %c-4, %vscale_11 : index
    %num_rows_remaining = arith.addi %num_rows_init, %c-4_vscale : index
```

(inserted by the "arm-sme-vector-legalization" pass, see
`VectorLegalization.cpp`).

Because of the subtraction, `%num_rows_remaining` can be negative.
Therefore, the comparison must be signed (`slt`) rather than unsigned
(`ult`).

Fixes: https://github.com/iree-org/iree/issues/21714
This commit is contained in:
Andrzej Warzyński 2025-09-05 08:58:52 +01:00 committed by GitHub
parent e71ec0f7a4
commit e59414f9bf
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
mlir/lib/Conversion/ArmSMEToSCF/ArmSMEToSCF.cpp
View File

@ -309,7 +309,7 @@ struct TileLoadOpWithMaskAndPadNonZeroConversion
// Combine masks.
auto rowIsActive = arith::CmpIOp::create(
rewriter, loc, arith::CmpIPredicate::ult, tileSliceIndex, numRows);
rewriter, loc, arith::CmpIPredicate::slt, tileSliceIndex, numRows);
auto rowIsActiveI32 = arith::ExtSIOp::create(
rewriter, loc, rewriter.getI32Type(), rowIsActive);
auto mask =

2
mlir/test/Conversion/ArmSMEToSCF/arm-sme-to-scf.mlir
View File

@ -81,7 +81,7 @@ func.func @arm_sme_tile_load_hor_with_mask_and_pad_zero(%src : memref<?x?xi32>)
// CHECK-DAG: %[[VSCALE:.*]] = vector.vscale
// CHECK-NEXT: %[[NUM_TILE_SLICES:.*]] = arith.muli %[[C4]], %[[VSCALE]] : index
// CHECK-NEXT: scf.for %[[TILE_SLICE_INDEX:.*]] = %[[C0]] to %[[NUM_TILE_SLICES]] step %[[C1]] iter_args(%[[CURRENT_TILE:.*]] = %[[TILE]]) -> (vector<[4]x[4]xi32>) {
// CHECK-NEXT: %[[ROW_IS_ACTIVE:.*]] = arith.cmpi ult, %[[TILE_SLICE_INDEX]], %[[NUM_ROWS]] : index
// CHECK-NEXT: %[[ROW_IS_ACTIVE:.*]] = arith.cmpi slt, %[[TILE_SLICE_INDEX]], %[[NUM_ROWS]] : index
// CHECK-NEXT: %[[ROW_IS_ACTIVE_SEXT_I32:.*]] = arith.extsi %[[ROW_IS_ACTIVE]] : i1 to i32
// CHECK-NEXT: %[[MASK:.*]] = arith.andi %[[ROW_IS_ACTIVE_SEXT_I32]], %[[NUM_COLS_I32]] : i32
// CHECK-NEXT: %[[MASK_INDEX:.*]] = arith.index_cast %[[MASK]] : i32 to index