All issues

Search results

[BUG] [maca] [regression] No module named 'tilelang.intrinsics.maca_mma_macro_generator'

bugSomething isn't workingSomething isn't working

Status: Open.

#100 In tile-ai/tilelang-metax;

· ventijing opened on Jun 2, 2026

[BUG] [maca] [regression] no available layout found

bugSomething isn't workingSomething isn't working

Status: Open.

#99 In tile-ai/tilelang-metax;

· ventijing opened on Jun 2, 2026

[BUG] [maca] [regression]Unresolved call Op(tir.ptx_init_barrier_thread_count)

bugSomething isn't workingSomething isn't working

Status: Open.

#98 In tile-ai/tilelang-metax;

· ventijing opened on Jun 1, 2026

MACA backend: FP16 accumulation MMA compile error

enhancementNew feature or requestNew feature or request

Status: Open.

#69 In tile-ai/tilelang-metax;

· Dayuxiaoshui opened on May 8, 2026

Sparse tensorcore example has no executable sparse MMA path on MACA

Status: Open.

#32 In tile-ai/tilelang-metax;

· VitalyAnkh opened on Apr 9, 2026

The minference MACA path requires portable preprocessing and a synchronized fallback kernel

Status: Open.

#31 In tile-ai/tilelang-metax;

· VitalyAnkh opened on Apr 9, 2026

DeepSeek V32 MACA examples rely on CUDA-specific barriers, TMA paths, and vector atomics

Status: Open.

#30 In tile-ai/tilelang-metax;

· VitalyAnkh opened on Apr 9, 2026

Several MACA examples exceed platform resource limits or require contiguous inputs

Status: Open.

#29 In tile-ai/tilelang-metax;

· VitalyAnkh opened on Apr 9, 2026

Low-precision MACA examples assume CUDA/Hopper-specific emitters, launch shapes, and reference paths

Status: Open.

#28 In tile-ai/tilelang-metax;

· VitalyAnkh opened on Apr 9, 2026

MACA backend lacks dtype and code generation support required by FP8 and vectorised kernels

Status: Open.

#27 In tile-ai/tilelang-metax;

· VitalyAnkh opened on Apr 9, 2026