tqchen opened a new pull request, #19722:
URL: https://github.com/apache/tvm/pull/19722
Replace TVM's `Diagnostic` / `DiagnosticContext` machinery with the tvm-ffi
`visit_error_context` mechanism. Validators throw an `ffi::Error` seeded
with the
offending node; leaf pass executors (`ModulePass` / relax `Function` /
`DataflowBlock`) catch and rethrow `EnrichPassErrorWithContext`, which
appends the
failing pass name and a TVMScript-rendered, underlined source location.
`relax.analysis.well_formed` now throws on the first violation; a new
`check_well_formed` returns a bool, and all C++/Python/test callers are
routed
accordingly. `include/tvm/ir/diagnostic.h` and `src/ir/diagnostic.cc` are
deleted.
The enrichment renders with `num_context_lines=10` so a small function shows
in
full with no "(... N lines skipped ...)" marker, while a large module stays
bounded.
Example (a TIR module pass throws on a body store):
```
synthetic TIR validation failure on the function body store
Error in pass: TIRxErrorEnrichmentDemo
Location (TVMScript):
# from tvm.script import ir as I
# from tvm.script import tirx as T
# from tvm.tirx.layout import Axis
@I.ir_module
class Module:
@T.prim_func(private=True)
def main(a: T.handle):
A = T.match_buffer(a, (16,), layout=None)
i = T.int32()
A[i] = A[i] + T.float32(1.0)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
```
The generic `ModulePass` catch renders a pure-TIR (`T.*`) location, and
positive
well-formedness assertions are routed through the throwing `well_formed(X)`
for
descriptive failures. `tvm.error.DiagnosticError` (used by the TVMScript
parser)
is retained.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
