Closed pdhirajkumarprasad closed 1 day ago
@pdhirajkumarprasad for such small examples it will also be helpful if you can attach the log with --mlir-print-ir-after-all --mlir-print-ir-before-all --mlir-disable-threading --mlir-elide-elementsattrs-if-larger=4
(the log can be large if it is a larger program, but for this size of program this is fine). I can redirect easily by just looking at the log.
@pashu123 please attach log when you get to this.
@MaheshRavishankar See the dump file: out_mlir.txt
This is related to #18297 as well.
Confirmed the issue no longer persists on main branch 56ecef7f .
Verified with latest build and don't see this issue anymore
What happened?
For the given IR
getting error as
model.torch_onnx.mlir:7:12: error: 'func.func' op exceeded stack allocation limit of 32768 bytes for function. Got 65664 bytes %262 = torch.operator "onnx.MatMul"(%261, %arg5) : (!torch.vtensor<[8,128,768],f32>, !torch.vtensor<[768,768],f32>) -> !torch.vtensor<[8,128,768],f32> ^ model.torch_onnx.mlir:7:12: note: see current operation:
IR after failure:
Steps to reproduce your issue
Command to reproduce:
IREE version: IREE compiler version 20240819.990 @ aeda14995f16ed1302db616adf0c03acf80f27ee LLVM version 20.0.0git
What component(s) does this issue relate to?
Compiler
Version information
No response
Additional context
No response