| Issue |
131283
|
| Summary |
[MLIR] crashes with "only integers and floats have a bitwidth"
|
| Labels |
mlir
|
| Assignees |
|
| Reporter |
thaddywu
|
To reproduce: https://godbolt.org/z/T3PfGcdvT
This bug appears when turning any of such optimizations:
`mlir-opt --test-vector-linearize` or
`mlir-opt --arith-emulate-wide-int` or
`mlir-opt --async-to-async-runtime` or
`mlir-opt --expand-realloc`
Code snippet:
```
"builtin.module"() ({
"spirv.func"() <{
function_control = #spirv.function_control<None>,
function_type = (!spirv.coopmatrix<2x2xi32, Subgroup, MatrixA>, !spirv.coopmatrix<2x2xi32, Subgroup, MatrixB>) -> (),
sym_name = ""
}>({
^bb0(%arg0: !spirv.coopmatrix<2x2xi32, Subgroup, MatrixA>, %arg1: !spirv.coopmatrix<2x2xi32, Subgroup, MatrixB>):
%0 = "spirv.ISub"(%arg0, %arg0) : (!spirv.coopmatrix<2x2xi32, Subgroup, MatrixA>, !spirv.coopmatrix<2x2xi32, Subgroup, MatrixA>) -> !spirv.coopmatrix<2x2xi32, Subgroup, MatrixA>
"spirv.Return"() : () -> ()
}) : () -> ()
}) : () -> ()
```
![]()
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
