[Bugfix][TE] Fix Ascend wildcard location probe fallback to host memory

[gygdh-001]

Description

ResolveAscendMemType() previously used CHECK_ACL, treating an
aclrtPointerGetAttributes() failure as a hard error (returning -1). As a result, registering
host memory (hugepage / shm / malloc) with the wildcard location "*" failed on Ascend. CUDA
already falls back when the wildcard probe cannot identify the memory; this change aligns Ascend
by treating the probe failure as host memory (adxl::MEM_HOST). This matches the existing fix
for the heterogeneous RDMA transport (#1657).

Fixes #4054

Module

• Transfer Engine (mooncake-transfer-engine)
• Mooncake Store (mooncake-store)
• Reshard (mooncake-reshard)
• Mooncake EP (mooncake-ep)
• Mooncake PG (mooncake-pg)
• Integration (mooncake-integration)
• P2P Store (mooncake-p2p-store)
• Python Wheel (mooncake-wheel)
• Common (mooncake-common)
• Mooncake RL (mooncake-rl)
• CI/CD
• Docs
• Other

Type of Change

• Bug fix
• New feature
• Refactor
• Breaking change
• Documentation update
• Performance improvement
• Other

How Has This Been Tested?

Manual verification on Ascend 910B (bare mooncake.engine TE path). The wildcard fallback fix
follows the same pattern already merged for the heterogeneous-RDMA transport (#1657).

Test results:

• Unit tests pass
• Integration tests pass (if applicable)
• Manual testing done (describe below)

Checklist

• I have performed a self-review of my own code
• I have formatted my code using ./scripts/code_format.sh
• I have run pre-commit on the files changed in this PR and all hooks pass
• I have updated the documentation (if applicable)
• I have added tests to prove my changes are effective
• For changes >500 LOC: I have filed an RFC issue

AI Assistance Disclosure

• No AI tools were used
• AI tools were used (specify below)

[ascend-direct-dev]

Thanks for the report. We probed aclrtPointerGetAttributes() on Ascend 910 / CANN 9.2.0 for the host allocation paths mentioned here (hugepage / shm / malloc). The ACL call does not fail for these buffers, so the CHECK_ACL change is not needed.

Current ResolveAscendMemType() already covers this: a successful probe with location.type that is neither HOST nor DEVICE falls through to the existing else and is treated as host memory.

Allocation	ACL ret	location.type	Current code path
malloc / new[] / posix_memalign	0	UNREGISTERED(2)	else → HOST
anonymous mmap / THP	0	UNREGISTERED(2)	else → HOST
shm_open + mmap / memfd	0	UNREGISTERED(2)	else → HOST
MAP_HUGETLB	0	UNREGISTERED(2)	else → HOST
aclrtMallocHost	0	HOST(0)	HOST branch
aclrtMalloc (device)	0	DEVICE(1)	DEVICE branch
nullptr	100000	—	CHECK_ACL fails

ACL is not CUDA here: CUDA may fail the probe on unregistered host pointers, but ACL returns ACL_MEM_LOCATION_TYPE_UNREGISTERED successfully. Swallowing ACL errors would also turn a real invalid pointer (nullptr) into host memory.

Please keep the existing else HOST fallback and drop this change.