Issues

• [Transform][Arith] ConstrSet::Populate ordering is load-bearing in two incompatible directions (eager Analyzer::Bind snapshots)

  tile-ai/tilelang#3220 · Filter by author LeiWang1999 opened on Sep 13, 2026

• [Maint] layout_inference golden snapshots are stale on main; harness has no CI hook

  tile-ai/tilelang#3188 · Filter by author LeiWang1999 opened on Sep 9, 2026

• [Feature Request] [RFC] Add explicit CUDA cache policies for normal global-memory copies

  enhancement
  tile-ai/tilelang#3178 · Filter by author yanglf1121 opened on Sep 7, 2026
  

• [Question][Carver] TensorCorePolicy: tiles whose smem_cost × pipeline_stage exceeds smem_cap are pruned rather than downgraded, and the shared.dyn fallback in _assign_block_size looks unreachable

  question
  tile-ai/tilelang#3164 · Filter by author mocusez opened on Sep 3, 2026
  1

• [Perf] Lower T.any_of / T.all_of via warp vote intrinsics (__any_sync / __all_sync) where safe

  enhancementgood first issue
  tile-ai/tilelang#3120 · Filter by author LeiWang1999 opened on Aug 31, 2026
  2

  

• [BUG][Fuzzer][ice-on-valid-code] 8-bit-operand sparse T.gemm_sp with K > block_K crashes ThreadSync (Cannot match type int32 vs handle) instead of compiling

  tile-ai/tilelang#3037 · Filter by author Hughshine opened on Aug 17, 2026
  1

• [BUG][Fuzzer][ice-on-invalid-code] T.rng_rand/T.rng_rand_float with no preceding T.rng_init emit curand…(&) and crash nvcc, instead of a clean "init required" error

  tile-ai/tilelang#3034 · Filter by author Hughshine opened on Aug 17, 2026
  3

• [BUG][Fuzzer][ice-on-invalid-code] T.gemm on a narrow dtype (int8/fp8/bfloat16) with any sub-atom block_K crashes nvcc instead of cleanly rejecting

  tile-ai/tilelang#3032 · Filter by author Hughshine opened on Aug 17, 2026

• [BUG][Fuzzer][ice-on-valid-code] Binding T.rng_init()'s documented state handle to a variable emits void state = ; and nvcc rejects the kernel

  tile-ai/tilelang#3029 · Filter by author Hughshine opened on Aug 17, 2026
  2

• [BUG][Fuzzer][ice-on-valid-code] Region-form tile atomic with a higher-rank source than destination crashes at compile with a raw IndexError

  tile-ai/tilelang#3023 · Filter by author Hughshine opened on Aug 17, 2026
  1

• [BUG][Fuzzer][ice-on-valid-code] T.import_source prelude with no trailing newline fuses with the next #include, so nvcc rejects the CUDA instead of compiling

  tile-ai/tilelang#3020 · Filter by author Hughshine opened on Aug 17, 2026

• [BUG][Fuzzer][ice-on-valid-code] T.q_multiply_shift crashes on a runtime (non-constant) multiplier instead of computing the result

  tile-ai/tilelang#3014 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-valid-code] T.Parallel(coalesced_width=<int>) aborts compilation instead of accepting the documented int

  tile-ai/tilelang#3013 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-valid-code] T.cumsum/T.cummax shared src + fragment dst aborts with a message-less internal ICHECK(IsSharedBuffer(dst)) instead of scanning

  tile-ai/tilelang#3011 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-valid-code] a non-bool buffer element as a bare if/assert condition aborts with an internal is_bool() assert instead of testing != 0

  tile-ai/tilelang#3009 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-valid-code] Float operands to the integer-only div/mod ops (%, //, T.floormod/floordiv/ceildiv/truncdiv) abort with an internal assert

  tile-ai/tilelang#3008 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-valid-code] T.q_multiply_shift_per_axis aborts with an internal is_bool() check when is_lshift_required is a plain integer (its documented type)

  tile-ai/tilelang#3006 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-invalid-code] Reading a local.fragment at two different indices in T.Parallel aborts with an internal StructuralEqual ICHECK instead of a clear "unsupported" diagnostic

  tile-ai/tilelang#3005 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-valid-code] T.q_multiply_shift with s=1 and a power-of-2 multiplier aborts with an internal shift-range ICHECK instead of computing the result

  tile-ai/tilelang#3003 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-valid-code] T.gemm accepts bf16/int8/tf32 input with float16 accum (no MMA specialization) then aborts nvcc with a template static_assert

  tile-ai/tilelang#3004 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-valid-code] int32 expression whose constant terms fold past INT32_MAX aborts compilation instead of wrapping

  tile-ai/tilelang#3002 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-valid-code] Casting an out-of-range integer constant to a narrow int crashes the compiler instead of wrapping

  tile-ai/tilelang#3000 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-valid-code] A finite fp32 constant above a narrow float dtype's max (fp16/fp8/fp4) aborts in FloatImm instead of rounding like the runtime cast

  tile-ai/tilelang#2999 · Filter by author Hughshine opened on Aug 16, 2026

• [BUG][Fuzzer][ice-on-valid-code] T.reduce_max/reduce_min into a float8_e4m3/float4 fragment crashes nvcc with undefined inff instead of compiling

  tile-ai/tilelang#2998 · Filter by author Hughshine opened on Aug 16, 2026
  1

• [BUG][Fuzzer][ice-on-valid-code] int2/uint2 (any integer width ∉ {1,4,8,16,32,64}) aborts CUDA codegen with Cannot convert type int2 instead of compiling

  tile-ai/tilelang#2996 · Filter by author Hughshine opened on Aug 16, 2026