Repository Issue Activity (beta)

#20388 [VOTE] Release Apache TVM v0.27.0.rc0

#20197 [Bug][Relax] CombineParallelMatmul crashes on 1-D matmul boundary cases

#19578 [Bug] Matmul and reductions with zero-size dimension return uninitialized memory

#20263 [Bug][Relax] ONNX Slice with runtime negative starts gives wrong results (zeros) — dynamic_strided_slice mishandles negative begin

#20269 [Bug] Allocation/shape arithmetic wraps in storage planning and workspace lowering

#20361 [Bug] [Relax][cuDNN] CoreDump When cuDNN ConvEntry CleanWorkspace

#19986 [CI Problem] test_solution_consistency fails intentionally-nondeterministic check in solve_linear_equations

#20342 [Bug][Performance Regression][Relax][DLight] FP16 Exp producer inlining across broadcast causes 30-60% slowdown on RTX A6000

#20315 [Bug][TIRx][FP8] DTypeConversion mis-encodes float8_e5m2 zero/subnormal values when converting to/from float32

#20319 [Relax][Frontend][Torch] `_one_hot` does not validate `num_classes`: non-positive values crash `relax.op.one_hot` with an opaque `InternalError` ("depth must be positive")

#20352 [Bug] Hexagon codegen is broken on main and the only test covering it never runs

#20343 [Bug] Unsigned conditions crash block access-region analysis with an internally generated negative constant

#20340 [Bug] Fixed-point multiply legalization rejects runtime operands, zero-exponent scaling, and integer flags

#20335 [Bug][Relax] BundleModelParams output (tuple param with symbolic-shape fields) segfaults VMShapeLower

#20290 [Bug] keep_params_in_input strip names incorrectly when using tvm.relax.frontend.onnx.from_onnx

#20281 [Bug][Relax] Constant folding of an int64 Div goes through float64 and is off by one above 2^53

#19969 [Bug][Dlight]DLight GPU GEMV fails with undefined TX in broadcast epilogue path

#20047 [Bug] `relax.build` crashes with an `AttributeError` during the dlight GPU scheduling pass

#20307 [Bug][Relax] split with out-of-range indices: type layer clamps but topi computes a negative extent ? bad_alloc / uint error / silent oversized (OOB-read) output depending on config

#20328 [TIRx][CUDA] tcgen05.mma.ws has no collector_b modifier on sm_100a

#20321 [Relax][Frontend][Torch] `_squeeze` does not validate `dim`: out-of-bounds tuple dims are silently converted to `squeeze(None)` via `from_fx` (native PyTorch raises `IndexError`)

#20297 Bug][Relax] Unchecked indices in gather/scatter ops: silent OOB reads and memory-corrupting OOB writes (reachable through the ONNX importer)

#20314 [Bug] MetaSchedule tuning exits successfully with zero candidates measured on Windows

#20268 [Bug] TIR allocation-size analysis silently overflows in `ConstantAllocationSize` and `CalculateAllocatedBytes`

#19848 # [Performance] `nn.pad` reflect/replicate ~1.5x slower than needed since v0.21 (nested `if_then_else` boundary index)