Issues

• [runtime] TornadoLogger can't show warnings to users, so the runtime prints them with System.out

  beehive-lab/TornadoVM#1286 · Filter by author kotselidis opened on Oct 25, 2026
  

• [Metal] Host reads after MLX library tasks return stale data with -Dtornado.mlx.async=true

  beehive-lab/TornadoVM#1242 · Filter by author orionpapadakis opened on Oct 15, 2026

• [Metal] Asynchronous dispatch for MLX library tasks

  beehive-lab/TornadoVM#1236 · Filter by author kotselidis opened on Oct 13, 2026
  

• consumeFromDevice without a producer name is ignored in a graph that also consumes by name

  beehive-lab/TornadoVM#1212 · Filter by author kotselidis opened on Oct 8, 2026
  

• [cuda] new HalfFloat((short) bits).getFloat32() fails to compile: CUDAConvertHalfToFloat input cleared by a graph rewrite

  beehive-lab/TornadoVM#1199 · Filter by author kotselidis opened on Oct 7, 2026

• [api] Native arrays are int-indexed: one array cannot exceed Integer.MAX_VALUE elements (fromSegment truncates silently)

  beehive-lab/TornadoVM#1156 · Filter by author mikepapadim opened on Oct 1, 2026

• [cuda] cuVS library provider: k-means, brute-force kNN and CAGRA as library tasks

  CUDAfeature
  beehive-lab/TornadoVM#1154 · Filter by author mikepapadim opened on Sep 30, 2026
  

• [api] KernelContext.atomicAdd should return the previous value (fetch-and-add)

  beehive-lab/TornadoVM#1152 · Filter by author mikepapadim opened on Sep 30, 2026

• [launcher] Pass arguments after the main class to the application instead of parsing them

  beehive-lab/TornadoVM#1150 · Filter by author mikepapadim opened on Sep 30, 2026

• CUDA backend codegen audit (7.0.1): wrong results (shifts, offset reductions, float->long, min/max NaN, FMA), compile failures, and measured FP64/aliasing optimisations

  CUDA
  beehive-lab/TornadoVM#1149 · Filter by author mikepapadim opened on Sep 29, 2026
  1

• CUDA backend: first-class cuda-gdb / compute-sanitizer debugging support

  CUDA
  beehive-lab/TornadoVM#1148 · Filter by author mikepapadim opened on Sep 29, 2026

• CUDA dynamic parallelism (device-side kernel launch) for the CUDA backend

  APICUDA
  beehive-lab/TornadoVM#1135 · Filter by author mikepapadim opened on Sep 27, 2026
  

• [cuda][profiler] Dispatch times are structurally 0 and copy-out time is never recorded, failing three non-whitelisted TestProfiler tests

  beehive-lab/TornadoVM#1071 · Filter by author mikepapadim opened on Sep 4, 2026
  1

  

• TornadoParallelScheduler: incorrect thread offset computation in scheduling for non-zero loop start with non-unit stride

  beehive-lab/TornadoVM#1044 · Filter by author karan-kamat opened on Aug 25, 2026
  1

• [bug] A batched plan with more than one output fails with a null device buffer at LAUNCH

  bug
  beehive-lab/TornadoVM#1034 · Filter by author mikepapadim opened on Aug 6, 2026
  

• Auto-tuning of thread configuration via TornadoExecutionPlan.withAutoTune()

  beehive-lab/TornadoVM#979 · Filter by author mikepapadim opened on Jul 24, 2026
  1

• [roadmap] NVIDIA ecosystem alignment based on findings from #953

  beehive-lab/TornadoVM#954 · Filter by author mikepapadim opened on Jul 20, 2026
  

• Deprecate string-keyed consumeFromDevice/persistOnDevice in favour of residency-by-object-identity + async handles

  beehive-lab/TornadoVM#942 · Filter by author mikepapadim opened on Jul 15, 2026
  3

  

• Async execution ergonomics: non-blocking executeAsync() + composable execution handles

  CUDA
  beehive-lab/TornadoVM#941 · Filter by author mikepapadim opened on Jul 15, 2026
  3

  

• include RELEASE file in sdk distro

  beehive-lab/TornadoVM#922 · Filter by author maxandersen opened on Jul 13, 2026
  1

  

• Metal backend reports grossly inflated GPU/dispatch timing metrics (TOTAL_KERNEL_TIME, TOTAL_DISPATCH_KERNEL_TIME, COPY_IN_TIME, COPY_OUT_TIME) despite comparable or better end-to-end execution time

  beehive-lab/TornadoVM#905 · Filter by author karan-kamat opened on Jul 9, 2026
  2

  

• Proposal: Same variable in different class

  beehive-lab/TornadoVM#784 · Filter by author ToTo3105 opened on Jan 20, 2026

• Proposal: Implement Triton-style Autotuning Support in TornadoVM

  APIperformanceruntime
  Feature

  beehive-lab/TornadoVM#778 · Filter by author mikepapadim opened on Dec 22, 2025
  

• NullPointerException when using persistOnDevice + consumeFromDevice across multiple TaskGraphs in a single TornadoExecutionPlan

  beehive-lab/TornadoVM#768 · Filter by author ToTo3105 opened on Dec 13, 2025
  2

• Sealed types cannot be instantiated. in Kotlin

  beehive-lab/TornadoVM#766 · Filter by author Blacktoviche opened on Dec 13, 2025
  2