Lunderberg commented on code in PR #16569:
URL: https://github.com/apache/tvm/pull/16569#discussion_r1492699232
##########
tests/python/arith/test_arith_domain_touched.py:
##########
@@ -71,7 +71,8 @@ def test_domain_touched():
def test_domain_touched_vector():
m = tvm.runtime.convert(128)
- @T.prim_func
+ # n is undefined
Review Comment:
Probably not. We could make it well formed by adding `n: T.int32` to the
arguments. It looks like this test is validating that a fixed integer extent
can be inferred, even when `n` is dynamic.
##########
tests/python/tir-base/test_tir_renew_defs.py:
##########
@@ -82,7 +82,8 @@ def _get_block(f):
def test_match_buffer():
- @T.prim_func
+ # well-formed checker complains about multiple definitions for a variable
A0_s1>?
Review Comment:
I'd say bug in the unit test. The `A0_s1` is a variable generated to
represent `A.strides[1]`. The `strides = [s, s]` should probably be `strides =
[s, 1]`.
##########
tests/python/tvmscript/test_tvmscript_parser_tir.py:
##########
@@ -272,7 +272,7 @@ def test_tir_starred_for_loop():
@T.prim_func(private=True)
def starred(a: T.handle, b: T.handle):
A = T.match_buffer(a, [*dims, 128], "int32")
- B = T.match_buffer(a, dims, "int32")
+ B = T.match_buffer(b, dims, "int32")
Review Comment:
Agreed, I think this was a bug in the unit test.
##########
tests/python/tir-base/test_tir_specialize.py:
##########
@@ -65,7 +65,7 @@ def matmul_m_128(a: T.handle, b: T.handle, c: T.handle) ->
None:
C[vi, vj] = C[vi, vj] + A[vi, vk] * B[vj, vk]
[email protected]_func
[email protected]_func(check_well_formed=False)
Review Comment:
I believe this is from `x` being undefined in the TIR. It appears as part
of a shape in `T.match_buffer`, but as `x * 8*` rather than on its own.
--
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]
