Issues

• [Codegen] Dynamic matmul causes a segfault when scf.forall uses its own result as an initializer

  bug 🐞
  iree-org/iree#25156 · Filter by author quiet-maple-47 opened on Oct 29, 2026
  2

• [Codegen][CUDA] Dynamic batch_reduce_matmul produces nondeterministic wrong results (missing cross-thread reduction)

  bug 🐞
  iree-org/iree#25155 · Filter by author quiet-maple-47 opened on Oct 29, 2026
  2

• [Codegen][CUDA] linalg.batch_reduce_matmul silently drops a batch contribution

  bug 🐞
  iree-org/iree#25154 · Filter by author quiet-maple-47 opened on Oct 29, 2026
  2

• [Codegen][CUDA] Silent wrong results: incorrect identities for signed min/max and bitwise AND reductions

  bug 🐞
  iree-org/iree#25153 · Filter by author quiet-maple-47 opened on Oct 29, 2026
  2

• [Stream] Silent Wrong Code: EncodeHostTensors turns a single-element tensor.insert into a multi-element fill

  bug 🐞
  iree-org/iree#25152 · Filter by author quiet-maple-47 opened on Oct 29, 2026
  2

• [VM] Silent Wrong Code: i8/i16 comparisons use inconsistent 32-bit extensions

  bug 🐞
  iree-org/iree#25151 · Filter by author quiet-maple-47 opened on Oct 29, 2026
  2

• [Stream] ScheduleConcurrency moves an in-place update before an earlier read of the same resource (write before read)

  bug 🐞
  iree-org/iree#25150 · Filter by author quiet-maple-47 opened on Oct 29, 2026
  2

• IREE writes a destination-passing op's result into the caller's argument

  bug 🐞
  iree-org/iree#25139 · Filter by author quiet-maple-47 opened on Oct 28, 2026
  4

• [Codegen] HoistStaticallyBoundAllocations incorrectly inserts dealloc at entry block terminator in multi-block functions

  bug 🐞
  iree-org/iree#25135 · Filter by author Cerulime opened on Oct 28, 2026
  2

• Fix shift-count-overflow error in runtime alignment for 32-bit targets

  enhancement ➕
  iree-org/iree#25134 · Filter by author riestmo-nxp opened on Oct 28, 2026
  2

• How should named linalg ops be handled in global optimization and dispatch creation?

  discussion
  iree-org/iree#25127 · Filter by author ziereis opened on Oct 26, 2026
  1

  

• torch.aten.scatter.src lowering fails on llvmcpu (too big alloca)

  bug 🐞
  iree-org/iree#25126 · Filter by author sixcy opened on Oct 26, 2026
  3

• Generalizing per-layer executables to reduce generated code size

  iree-org/iree#25124 · Filter by author jonathansong opened on Oct 26, 2026

• [Bug] online_attention produces type error after tiling

  bug 🐞
  iree-org/iree#25121 · Filter by author ppppqp opened on Oct 24, 2026
  2

• [DataTiling] affine.delinearize_index op no basis element may be statically non-positive for a matmul that follows a matmul with dynamic sizes

  iree-org/iree#25118 · Filter by author FedericoBruzzone opened on Oct 23, 2026

• [VMVX] verifier error memref.alloca op incorrect number of dynamic sizes with two conv2d with bias and a dynamic batch

  iree-org/iree#25116 · Filter by author FedericoBruzzone opened on Oct 23, 2026

• Conversion rounding issues in base/internal/math.h

  bug 🐞
  iree-org/iree#25111 · Filter by author sw23 opened on Oct 23, 2026
  2

• [DispatchCreation] Overly permissive isBitExtendOrTruncateOp clones complex producers into dispatches, causing GPU shared memory exhaustion

  bug 🐞
  iree-org/iree#25109 · Filter by author Cerulime opened on Oct 22, 2026
  2

• Support PyTorch models through TorchAX and StableHLO, with extensible composite lowering and device placement

  enhancement ➕integrations/pytorchintegrations/stablehlo
  Enhancement

  iree-org/iree#25108 · Filter by author jschuhmacher opened on Oct 22, 2026
  

• [GlobalOptimization] bf16/f16 contraction-input demotion rounds integer position ids in the K=1 rotary-embedding outer product

  iree-org/iree#25106 · Filter by author tonykuttai opened on Oct 22, 2026
  3

• [Discussion] Using IREE for compiler-guided subgraph prototyping in vLLM

  iree-org/iree#25092 · Filter by author weimin023 opened on Oct 20, 2026
  6

• [EmitC] Internal vm.call passes ref arguments without a reference, but the callee releases them (use-after-free of the coarse-fences signal fence)

  iree-org/iree#25087 · Filter by author davidmcevadsp opened on Oct 19, 2026
  1

• [LLVM CPU] BF16-to-F32 conversion leaves resource-backed constants with mismatched types

  bug 🐞
  iree-org/iree#25084 · Filter by author Manewing opened on Oct 19, 2026
  

• [Runtime][Bindings] Make HAL result construction exception-safe

  iree-org/iree#25083 · Filter by author Manewing opened on Oct 19, 2026

• Release Tracker v3.14.0 - (2026-11-16)

  iree-org/iree#25081 · Filter by author devtbi opened on Oct 19, 2026