TileOPs-Metax/tileops/kernels
RMLYC 250e9357e9 [Fix] Replace deprecated T.gemm(..., wg_wait=...) with T.wgmma_gemm() + T.wait_wgmma()
TileLang 0.1.9 removed the wg_wait parameter from T.gemm() public API.
This change migrates all affected attention kernels to the explicit
T.wgmma_gemm() + T.wait_wgmma() pattern.

- gqa_bwd.py: 10 gemm calls migrated; block_n config adjusted to >=64
  to satisfy WGMMA M>=64 constraint in transpose_A gemms
- deepseek_mla_decode.py: 12 gemm calls migrated
- deepseek_dsa_decode.py: 6 gemm calls migrated
- deepseek_dsa.py: add is_hopper() guard in default_kernel_map

Closes #1071
2026-05-12 15:56:57 +08:00
..
attention [Fix] Replace deprecated T.gemm(..., wg_wait=...) with T.wgmma_gemm() + T.wait_wgmma() 2026-05-12 15:56:57 +08:00
deltanet [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
engram [Fix][Ops] Complete GqaSlidingWindow rename and enable RUF022 (#967) 2026-04-14 21:45:16 +08:00
gated_deltanet [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
gla [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
mamba [BugFix][Mamba] Fix da_cumsum kernel to support dt_bias, softplus, and clamp (#1118) 2026-05-09 18:00:38 +08:00
mhc [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
moe [Fix][Ops] Complete GqaSlidingWindow rename and enable RUF022 (#967) 2026-04-14 21:45:16 +08:00
norm [Chore][Release] merge testbed into main: op-family alignment + new ops (#1380) 2026-05-09 00:03:37 +08:00
pool [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
reduction [Refactor][Hygiene] strip development-process metadata from shipped source (#1290) 2026-05-08 13:26:53 +08:00
__init__.py [Feat][Attention] support GQA prefill fused RoPE (#1234) 2026-05-06 20:30:34 +08:00
convolution.py [Refactor][Hygiene] strip development-process metadata from shipped source (#1290) 2026-05-08 13:26:53 +08:00
deltanet_recurrence.py [Fix][Ops] Complete GqaSlidingWindow rename and enable RUF022 (#967) 2026-04-14 21:45:16 +08:00
dropout.py [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
elementwise.py [Feat][Kernels] int/bool kernel coverage for MaskedFillScalar (#1411) 2026-05-12 09:57:31 +08:00
fft.py [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
fp8_lighting_indexer.py [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
fp8_quant.py [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
gated_deltanet_recurrence.py [Fix][Ops] Complete GqaSlidingWindow rename and enable RUF022 (#967) 2026-04-14 21:45:16 +08:00
gemm.py [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
gla_recurrence.py [Fix][Ops] Complete GqaSlidingWindow rename and enable RUF022 (#967) 2026-04-14 21:45:16 +08:00
grouped_gemm.py [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
kernel_base.py [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00
online_softmax.py [Feat][Attention] support GQA prefill fused RoPE (#1234) 2026-05-06 20:30:34 +08:00
rope.py [BugFix][ROPE] fix non-neox 2d compile failure (#1270) 2026-05-07 22:03:48 +08:00
topk_selector.py [Refactor] Rename base modules to <module>_base.py convention (#950) 2026-04-13 18:13:32 +08:00