Yongqi ZhuoandClaude Fable 5 c2d847f497 [BugFix][CUDA] Advance tcgen05 ld/st segment pointers in b32 columns (#2952) ...

The multi-segment recursion in tcgen05_{ld,st}_core advanced the
register pointer in dst_t/src_t elements while segment lengths count
b32 columns, so a 16-bit copy whose column count is not a power of two
(e.g. 48 columns = one x32 + one x16 segment) accessed the tail
segments' registers at half the intended offset -- a silent
miscompilation. Advance the pointer in b32 units instead.

Add asymmetric whole-store/whole-load roundtrips at (128, 96) bf16: a
symmetric roundtrip cancels the unit error, so exactly one side is
split into power-of-two (single-segment) slices, catching each
direction independently.

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

2026-08-13 13:35:42 +08:00

..

test_tilelang_cast_rounding.py

[Testing] Gate CUDA-only tests so non-CUDA backends can run the suite (#2863)

2026-08-04 18:22:40 +08:00

test_tilelang_intimm.py

[Lint] Phaseout Yapf format and embrace ruff format (#1417)

2025-12-12 17:22:18 +08:00

test_tilelang_language_access_ptr_codegen.py

[CUDA] Support int4 T.gemm (#2063)

2026-04-20 16:35:43 +08:00

test_tilelang_language_access_ptr.py

[BugFix] Preserve buffer element offsets in access pointers (#2727)

2026-07-22 15:42:39 +08:00

test_tilelang_language_alias.py

[Bugfix] Open Rocm ci test and fix some bugs. (#1443)

2026-01-06 02:18:29 +08:00

test_tilelang_language_all_of.py

[Refactor][Pipeline] Run pipeline rewriting before layout inference and stabilize tiled WS (#2002)

2026-04-07 23:54:55 +08:00

test_tilelang_language_alloc_var_cmp.py

[Bugfix] Fix alloc_var re-bind warning when assigned with comparison ops (#1974)

2026-03-27 11:17:21 +08:00

test_tilelang_language_alloc.py

[Testing] Run portable regression tests on auto targets (#2914)

2026-08-07 15:07:26 +08:00

test_tilelang_language_annot.py

[Bugfix] Open Rocm ci test and fix some bugs. (#1443)

2026-01-06 02:18:29 +08:00

test_tilelang_language_annotate_safe_value.py

[Refactor][Pipeline] Run pipeline rewriting before layout inference and stabilize tiled WS (#2002)

2026-04-07 23:54:55 +08:00

test_tilelang_language_any_of.py

[Refactor][Pipeline] Run pipeline rewriting before layout inference and stabilize tiled WS (#2002)

2026-04-07 23:54:55 +08:00

test_tilelang_language_assume.py

[TIR][Runtime] Enforce host-evaluable assumptions at runtime (#2655)

2026-07-14 12:24:26 +08:00

test_tilelang_language_async_copy_gemm_sm80.py

[Refactor] Improve cp.async lowering and add async_copy op (#1887)

2026-03-06 17:35:16 +08:00

test_tilelang_language_atom_mma.py

[Refactor] Remove intrinsic compatibility facade (#2812)

2026-07-30 14:26:21 +08:00

test_tilelang_language_atomic.py

[CUDA] Add pre-SM80 fallback for bf16 atomic add (#2938)

2026-08-11 16:58:33 +08:00

test_tilelang_language_augassign.py

[Bugfix] Fix ast builder error for value -= 1 (#1825)

2026-02-09 17:30:22 +08:00

test_tilelang_language_ceildiv.py

[Feature] Expose multiple CUDA intrinsics (#2473)

2026-07-02 17:29:54 +08:00

test_tilelang_language_chain_equal.py

[Refactor][Pipeline] Run pipeline rewriting before layout inference and stabilize tiled WS (#2002)

2026-04-07 23:54:55 +08:00

test_tilelang_language_clamp.py

[Enhancement] Update examples and tests for improved type handling functionality (#1448)

2025-12-17 14:41:25 +08:00

test_tilelang_language_clear.py

[BugFix] Fix make_int sign-extending negative int8 lanes (#2438)

2026-06-23 14:40:11 +08:00

test_tilelang_language_cluster.py

[Transform] Materialize kernel launch per backend (#2387)

2026-06-12 17:14:02 +08:00

test_tilelang_language_composable_index.py

[Refactor][Pipeline] Run pipeline rewriting before layout inference and stabilize tiled WS (#2002)

2026-04-07 23:54:55 +08:00

test_tilelang_language_cooperative.py

[Fix] Add register to read A ptr in test_tilelang_language_cooperative.py (#1593)

2026-01-04 15:45:38 +08:00

test_tilelang_language_copy.py

[Feature] Warp specialization schedules and materialization (#2892)

2026-08-11 15:39:40 +08:00

test_tilelang_language_dialect.py

[TIR][Python] Split backend-specific op proxies (#2870)

2026-08-05 12:57:23 +08:00

test_tilelang_language_dtype_as_torch.py

[JIT][FFI] Allocate kernel outputs through the packed API (#2937)

2026-08-11 10:47:08 +08:00

test_tilelang_language_eager_ast_signature.py

Fix eager AST handling for *args and **kwargs (#2330)

2026-06-04 14:14:14 +08:00

test_tilelang_language_eager_jit.py

[Testing] Gate CUDA-only tests so non-CUDA backends can run the suite (#2863)

2026-08-04 18:22:40 +08:00

test_tilelang_language_extract_lane.py

[TIR][Language] Add typed vector lane extraction API (#2789)

2026-07-28 17:18:25 +08:00

test_tilelang_language_fastmath.py

[BugFix] Reject unsupported fast-math input dtypes (#2804)

2026-07-30 14:22:27 +08:00

test_tilelang_language_float4_e2m1_unpacked_dtype.py

[Testing] Gate CUDA-only tests so non-CUDA backends can run the suite (#2863)

2026-08-04 18:22:40 +08:00

test_tilelang_language_frontend_v2.py

[Testing] Run portable regression tests on auto targets (#2914)

2026-08-07 15:07:26 +08:00

test_tilelang_language_func_attrs.py

[JIT][FFI] Allocate kernel outputs through the packed API (#2937)

2026-08-11 10:47:08 +08:00

test_tilelang_language_get_warp_info.py

[Backend] Add target detector/normalizer registeration and merge ExecutionBackend into backend module (#2323)

2026-06-09 00:35:07 +08:00

test_tilelang_language_if_range.py

[Refactor] Remove GEMM v1 and promote gemm_py to be the canonical gemm op (#2033)

2026-04-13 12:42:44 +08:00

test_tilelang_language_infinity.py

[BugFix] Support T.infinity for float8_e5m2 (#2671)

2026-07-15 17:37:20 +08:00

test_tilelang_language_int64.py

[Feature] Add full Windows support and fix related cross-platform issues (#2093)

2026-05-06 23:08:24 +08:00

test_tilelang_language_intrinsics_codegen.py

[Enhancement] Update examples and tests for improved type handling functionality (#1448)

2025-12-17 14:41:25 +08:00

test_tilelang_language_isfinite_codegen.py

[Feature] Add CUDA intrinsic for isfinite operation (#1925)

2026-03-12 17:21:01 +08:00

test_tilelang_language_kernel_threads.py

[BugFix] Reject non-positive thread extents in T.Kernel (#2653)

2026-07-13 17:29:01 +08:00

test_tilelang_language_ldg.py

[Refactor] Unify the usage of cast-related operators (#1757)

2026-02-02 18:47:18 +08:00

test_tilelang_language_ldst.py

[Feature] Expose multiple CUDA intrinsics (#2473)

2026-07-02 17:29:54 +08:00

test_tilelang_language_let_layout.py

[Refactor][Pipeline] Run pipeline rewriting before layout inference and stabilize tiled WS (#2002)

2026-04-07 23:54:55 +08:00

test_tilelang_language_let.py

[Fix] Bind statement missing role marker (#2362)

2026-06-12 17:35:42 +08:00

test_tilelang_language_mask_op.py

[Refactor][Pipeline] Run pipeline rewriting before layout inference and stabilize tiled WS (#2002)

2026-04-07 23:54:55 +08:00

test_tilelang_language_min_blocks_per_sm.py

[Refactor] Remove GEMM v1 and promote gemm_py to be the canonical gemm op (#2033)

2026-04-13 12:42:44 +08:00

test_tilelang_language_negative_index.py

[Enhancement] Update examples and tests for improved type handling functionality (#1448)

2025-12-17 14:41:25 +08:00

test_tilelang_language_nvf4_mma_block_scale.py

[CUDA] Add SM120 NVF4 block-scale MMA support (#2364)

2026-07-30 18:10:10 +08:00

test_tilelang_language_parallel.py

test: reduce CI runtime for slow Python suites (#1932)

2026-03-18 13:08:30 +08:00

test_tilelang_language_pdl.py

[Backend] Add target detector/normalizer registeration and merge ExecutionBackend into backend module (#2323)

2026-06-09 00:35:07 +08:00

test_tilelang_language_pipeline.py

[Refactor][Pipeline] Run pipeline rewriting before layout inference and stabilize tiled WS (#2002)

2026-04-07 23:54:55 +08:00

test_tilelang_language_ptr.py

[BugFix] Skip descriptor TMA for device-bound copy bases (#2803)

2026-07-29 20:35:07 +08:00

test_tilelang_language_rand.py

[CUDA][Codegen] Keep RNG state in kernel scope (#2540)

2026-07-07 14:41:03 +08:00

test_tilelang_language_reduce_maxmin_nan.py

[BugFix] Honor nan_propagate in reduce max/min/absmax clear=False write-back (#2788)

2026-07-28 16:04:07 +08:00

test_tilelang_language_reduce.py

[Testing] Run portable regression tests on auto targets (#2914)

2026-08-07 15:07:26 +08:00

test_tilelang_language_reshape.py

[Refactor] Move backend-specific GEMM implementations and transforms into backend directories (#2165)

2026-05-08 17:11:34 +08:00

test_tilelang_language_scan.py

[Feature] Warp specialization schedules and materialization (#2892)

2026-08-11 15:39:40 +08:00

test_tilelang_language_select.py

[TIR][Analyzer] Fix vectorized Select constraint handling (#2731)

2026-07-22 17:12:20 +08:00

test_tilelang_language_source_kernel.py

[Feature] Introduce T.CUDASourceCodeKernel (#1970)

2026-04-17 16:48:28 +08:00

test_tilelang_language_span.py

[TIR] Inject source spans into tirx IR and surface source locations in compiler errors (#2751)

2026-07-28 13:46:14 +08:00

test_tilelang_language_subtype.py

[Bugfix][Subtype] Fix scalar fp4 store/load codegen for non-packed buffers (#2037)

2026-04-14 15:54:28 +08:00

test_tilelang_language_sync_threads.py

Allow variable barrier id in T.sync_threads() (#2197)

2026-05-18 14:50:43 +08:00

test_tilelang_language_tcgen05_cutedsl.py

[Layout] Support half-subpartition (M=64) TMEM tiles in tcgen05.ld/st (#2880)

2026-08-11 15:35:15 +08:00

test_tilelang_language_tcgen05_gemm.py

[Layout][CUDA] Support reinterpreting (dtype-changing) T.view aliases (#2953)

2026-08-13 13:01:51 +08:00

test_tilelang_language_tcgen05_sliced_operands.py

[CUDA] Support arbitrary TMEM layouts (#2785)

2026-07-29 13:07:28 +08:00

test_tilelang_language_ternary.py

[Enhancement] Update examples and tests for improved type handling functionality (#1448)

2025-12-17 14:41:25 +08:00

test_tilelang_language_tile_schedule.py

[Language][Scheduler] Expose scalar tile scheduler state (#2553)

2026-07-08 14:34:55 +08:00

test_tilelang_language_tma_1d.py

[Testing] Gate CUDA-only tests so non-CUDA backends can run the suite (#2863)

2026-08-04 18:22:40 +08:00

test_tilelang_language_tma_copy.py

[Enhancement] Expose cluster_mask on T.tma_copy (#2932)

2026-08-11 20:12:38 +08:00

test_tilelang_language_tma_gather_scatter.py

[Feature] Add read option to TMA store wait (#2300)

2026-05-31 03:51:36 +08:00

test_tilelang_language_tma_store.py

[Feature] Add read option to TMA store wait (#2300)

2026-05-31 03:51:36 +08:00

test_tilelang_language_tmem_copy.py

[BugFix][CUDA] Advance tcgen05 ld/st segment pointers in b32 columns (#2952)

2026-08-13 13:35:42 +08:00

test_tilelang_language_transpose.py

[Feature] Add full Windows support and fix related cross-platform issues (#2093)

2026-05-06 23:08:24 +08:00

test_tilelang_language_unroll.py

[Enhancement] Improve unroll loop functionality for dynamic extent and corresponding test case (#1654)

2026-01-12 22:33:17 +08:00

test_tilelang_language_var_init.py

[Testing] Run portable regression tests on auto targets (#2914)

2026-08-07 15:07:26 +08:00

test_tilelang_language_vectorize.py

[Refactor][BugFix] Refactor the loop vectorization plan with ConstraintKind (#2935)

2026-08-11 12:19:53 +08:00

test_tilelang_language_vectorized_cast.py

[Enhancement] Add vectorized fp8x2 <-> fp16/bf16 cast codegen (#2475)

2026-06-29 16:00:57 +08:00

test_tilelang_language_view.py

[Layout][CUDA] Support reinterpreting (dtype-changing) T.view aliases (#2953)

2026-08-13 13:01:51 +08:00

test_tilelang_language_warp_reduce.py

[BugFix] Fix warp_reduce truncating int64/uint64 to 32 bits on sm_80+ (#2782)

2026-07-29 16:39:26 +08:00

test_tilelang_language_warp_sync.py

[TIR][IR] Update to use tirx (#2216)

2026-05-20 15:38:13 +08:00

test_tilelang_language_warp_vote.py

[BugFix] Reject floating-point predicates in vote intrinsics (#2800)

2026-07-29 13:33:48 +08:00

test_tilelang_language_wgmma_gemm.py

[CUDA] Adopt multi-staged buffers in examples (#2836)

2026-08-04 11:26:00 +08:00

test_tilelang_memory_leak.py

[Cherry-Pick][BugFix] Fix get_parent_locals frame self-reference leak (#2934)

2026-08-10 16:08:38 +08:00