Owner

ROCm

1 indexed repository · View on GitHub

TheRock
The HIP Environment and ROCm Kit - A lightweight open source build system for HIP and ROCm

Python · 1319 stars

Beginner friendly Everything

Open All

[Issue] actions expression `if: ${{ !cancelled() && !failure() }}` resolves to true even when needed steps time out Open

bug CICD Under Investigation

Difficulty 4/5 3-5 days Newbie friendliness 42/100

ROCm/TheRock#8372 · Opened Sep 19, 2026
[Issue][gfx90c][OpenCL] SVP workload causes GPU hang; ROCm remains broken after reset and clinfo triggers another hang Open

status: triage

ROCm/TheRock#8357 · 1 assignee · Opened Sep 18, 2026
Install a pinned rocm-bootstrap dependency for packaging tools Open

ROCm/TheRock#8354 · 2 assignees · Opened Sep 18, 2026
[Issue Linux PyTorch nightly: `grid_sample` 3D bicubic tests fail — `bicubic interpolation only supports 4D input` Open

ecosystem: PyTorch monitoring

Difficulty 4/5 3-5 days Newbie friendliness 38/100

ROCm/TheRock#8342 · Opened Sep 18, 2026
[Issue] Linux PyTorch nightly: `TestInputGradBuffersCUDA ` fails — grad_fn missing 'input_grad_buffers' attribute Open

ecosystem: PyTorch monitoring

Difficulty 4/5 3-5 days Newbie friendliness 35/100

ROCm/TheRock#8340 · Opened Sep 18, 2026
[Issue] Windows PyTorch nightly: gfx120X-all, gfx1151 — `test_rnn_packed_sequence_batch_sizes_must_be_cpu` crashes with access violation Open

ecosystem: PyTorch monitoring

ROCm/TheRock#8337 · 1 assignee · Opened Sep 18, 2026
[Issue] `test_hip_vector_add` sanity check fails across all GPU architectures — PR #8321 introduced a broken binary path `build/lib/llvm/bin/amdllvm` Open

monitoring platform: Linux release-nightly

ROCm/TheRock#8333 · 1 comment · 1 assignee · Opened Sep 18, 2026
[Documentation]: Full support for RDNA4 on Windows Open

documentation gfx120X-all platform: Windows

Difficulty 3/5 1-2 days Newbie friendliness 55/100

ROCm/TheRock#8330 · 1 comment · Opened Sep 18, 2026
[Issue] Linux gfx94X-dcgpu: rocjpeg tests segfault on linux-gfx942-1gpu-ccs-ossci-rocm runner pool Open

bump-pr CICD gfx94X-dcgpu infra-machine platform: Linux project: rocjpeg

Difficulty 4/5 3-5 days Newbie friendliness 35/100

ROCm/TheRock#8310 · Opened Sep 17, 2026
[Issue] gfx125X and gfx90a tests failing on submodule bumps, for now are non-blocking Open

bump-pr CICD gfx125x gfx90a

Difficulty 4/5 3-5 days Newbie friendliness 35/100

ROCm/TheRock#8309 · Opened Sep 17, 2026
[Issue] Linux gfx103X-all: runner-pool disruption during nightly run causes job cancellations and test timeouts (17 jobs) Open

monitoring release-nightly

Difficulty 4/5 3-5 days Newbie friendliness 42/100

ROCm/TheRock#8297 · 2 comments · Opened Sep 17, 2026
[Feature]: add back missing slim/bare docker images Open

Feature Request platform: Linux

Difficulty 4/5 3-5 days Newbie friendliness 55/100

ROCm/TheRock#8295 · Opened Sep 17, 2026
[Issue] Windows: Run PyTorch tests crashes at import — rocminfo FileNotFoundError in test_selections.py Open

ecosystem: PyTorch monitoring

ROCm/TheRock#8293 · 2 comments · 1 assignee · Opened Sep 17, 2026
[Issue] Windows gfx120X-all PyTorch release/2.12+2.13: `test_rnn_backward_to_input_but_not_parameters_cuda`,`test_graph_cudnn_dropout `fail with miopenStatusUnknownError Open

ecosystem: PyTorch monitoring

ROCm/TheRock#8292 · 1 assignee · Opened Sep 17, 2026
Enable gfx1250-strict builds and shared gfx1250 packaging Open

ROCm/TheRock#8278 · 1 assignee · Opened Sep 16, 2026
Resolve or suppress issues detected by security scanners Open

Difficulty 5/5 Over a week Newbie friendliness 30/100

ROCm/TheRock#8272 · 1 comment · Opened Sep 16, 2026
[Issue] Linux gfx94X-dcgpu: intermittent HIP managed-memory prefetch data-integrity failures Open

CICD gfx94X-dcgpu platform: Linux test-flaky test:hip-tests

Difficulty 4/5 3-5 days Newbie friendliness 48/100

ROCm/TheRock#8271 · 1 comment · Opened Sep 16, 2026
[Issue] Linux gfx94X-dcgpu: intermittent HIP VMM allocation and graph-memory failures on non-CSP pool Open

CICD gfx94X-dcgpu infra platform: Linux test:hip-tests

Difficulty 4/5 3-5 days Newbie friendliness 35/100

ROCm/TheRock#8270 · Opened Sep 16, 2026
[Issue] Linux gfx94X-dcgpu: rocdecode tests segfault on linux-gfx942-1gpu-ccs-ossci-rocm runner pool Open

ci-nightly CICD gfx94X-dcgpu infra-machine platform: Linux project: rocdecode

Difficulty 4/5 3-5 days Newbie friendliness 38/100

ROCm/TheRock#8265 · Opened Sep 16, 2026
feat: add skip_gpu_tests flag to Multi-Arch CI for build-only runs Open

ROCm/TheRock#8263 · 1 assignee · Opened Sep 16, 2026
[Issue] [rccl] `standard` suite fails with `[hip_fake] stub error` / `ncclCommSetAsyncError state 9` after switching from `test_rccl.py` to `test_runner.py` Open

bump-pr gfx94X-dcgpu gfx950-dcgpu platform: Linux project:rccl

ROCm/TheRock#8252 · 3 comments · 1 assignee · Opened Sep 16, 2026
Upstream LLVM about to raise minimum CMake version to 3.31 Open

Difficulty 3/5 1-2 days Newbie friendliness 58/100

ROCm/TheRock#8247 · Opened Sep 16, 2026
[Issue]: Torch profiling fails on TheRock 10 for subprocesses spawned after torch.cuda.init Open

Difficulty 4/5 3-5 days Newbie friendliness 50/100

ROCm/TheRock#8233 · Opened Sep 15, 2026
[Issue]: RX 7900 XTX over M.2 OCuLink: BACO resume failure leaves ROCm unusable Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

ROCm/TheRock#8232 · 1 comment · Opened Sep 15, 2026
Concurrent rocprofv3 profiling causes node reset on AMD Instinct MI300A GPUs Open

Verified Issue

ROCm/TheRock#8229 · 1 assignee · Opened Sep 15, 2026
[issue] enhancing our build and test selection strategy based on architecture Open

ROCm/TheRock#8222 · 1 comment · 1 assignee · Opened Sep 15, 2026
[Issue]: rocm_sdk test testCLIUsesDevelRootPath fails unless the venv is activated Open

packaging: python

Difficulty 2/5 1-3 hours Newbie friendliness 88/100

ROCm/TheRock#8219 · Opened Sep 15, 2026
[Issue] Windows PyTorch nightly: `test_mempool_limited_memory_with_allocator`, ` test_graph_capture_error_releases_reserved_segments` fail Open

ecosystem: PyTorch monitoring

ROCm/TheRock#8217 · 1 comment · 1 assignee · Opened Sep 15, 2026
[Issue] Windows PyTorch nightly/2.12/2.13: `test_compile_kernel_template`,` test_graph_memory_stats_`and_use_result_after_destroy_graph, test_graph_rng_after_failed_captur`e fail Open

ecosystem: PyTorch monitoring

ROCm/TheRock#8216 · 1 assignee · Opened Sep 15, 2026
[Documentation]: Ryzen AI MAX+ 395 / Radeon 8060S (gfx1151): ROCr AsyncEventsLoop consumes 100% of one CPU core after GPU workloads in ComfyUI Open

status: triage

ROCm/TheRock#8213 · 1 assignee · Opened Sep 15, 2026
[Feature]: Nightly (tip-vs-tip) job for JAX Framework (Phase 1) Open

ROCm/TheRock#8201 · 1 assignee · Opened Sep 14, 2026
[Feature]: Build `compiler/amd-llvm/amd/sqtt-marker` Open

project: compiler

ROCm/TheRock#8190 · 1 assignee · Opened Sep 14, 2026
ASan unknown-crash (wild pointer WRITE) deep in libamdhip64.so.7, hit early by rocSOLVER/hipSPARSE/rocSPARSE comprehensive suites (gfx950-dcgpu asan-debug) Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

ROCm/TheRock#8173 · 1 comment · Opened Sep 12, 2026
gfx1151 release index: its only JAX set (0.9.1+rocm7.13.0) cannot load — needs sysdeps libraries no 7.13.0 SDK wheel ships; PyPI 0.11.1 + the same SDK wheels works Open

ecosystem: JAX

Difficulty 4/5 3-5 days Newbie friendliness 48/100

ROCm/TheRock#8163 · 1 comment · Opened Sep 11, 2026
gfx950-dcgpu test jobs run on MI325X (gfx942) hardware, not gfx950 — widespread 'invalid device function' failures (rocWMMA, hipSOLVER, hipSPARSE) Open

gfx950-dcgpu

Difficulty 4/5 3-5 days Newbie friendliness 48/100

ROCm/TheRock#8151 · 8 comments · Opened Sep 11, 2026
[Issue] ASAN bad-free in libamdhip64.so.7 HSA memory pool allocator during static destructor teardown Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

ROCm/TheRock#8142 · 4 comments · Opened Sep 11, 2026
Ground rocm-pr-quality skill's testing guidance in TESTING.md + risk dimensions Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

ROCm/TheRock#8135 · Opened Sep 10, 2026
[Feature]: Integration with Torch auto-installers Open

Difficulty 5/5 Over a week Newbie friendliness 30/100

ROCm/TheRock#8115 · Opened Sep 9, 2026
[gfx1201][Windows] BF16 training loss diverges from native Linux on the same RX 9070 XT (ROCm 7.2.1; reproduced with a 7.13 nightly wheel) Open

status: triage

ROCm/TheRock#8095 · 1 assignee · Opened Sep 9, 2026
[Issue]: tracking of free memory does not work correctly when using stream ordered allocation Open

status: fix submitted

ROCm/TheRock#8092 · 1 comment · 1 reaction · 1 assignee · Opened Sep 9, 2026
Libraries TheRock pin bump PRs omit Multi-Arch workflow uses: pins Open

bug CICD

ROCm/TheRock#8068 · 1 assignee · Opened Sep 8, 2026
[Feature]: Add a workflow input for selecting `FLAGS.cmake` build flags in CI runs Open

ci-improvement CICD

ROCm/TheRock#8001 · 1 comment · 1 assignee · Opened Sep 8, 2026
[Issue] Linux nightly sanity test: `test_hip_printf (./hip_check) hangs and times out` Open

monitoring release-nightly

Difficulty 4/5 3-5 days Newbie friendliness 48/100

ROCm/TheRock#7989 · 5 comments · Opened Sep 8, 2026
[Issue]: hipBLASLt TensileLite runtime does not probe library/<arch>/ subdirectory — TensileLibrary_lazy_<arch>.dat not found Open

project: hipblaslt

Difficulty 3/5 1-2 days Newbie friendliness 74/100

ROCm/TheRock#7976 · 1 comment · Opened Sep 7, 2026
[Issue] Linux gfx950-dcgpu: `emulated mi350x rocjitsu daemon fails to start` — num_sdma_queues_per_engine must be nonzero Open

bump-pr code-regression gfx950-dcgpu monitoring platform: Linux

Difficulty 3/5 1-2 days Newbie friendliness 55/100

ROCm/TheRock#7964 · 4 comments · Opened Sep 7, 2026
[Issue] `dctools-core ASAN build fails `— protoc cannot find libclang_rt.asan.so after grpc unconditionally switched to amd-llvm toolchain Open

bump-pr code-regression gfx94X-dcgpu gfx950-dcgpu monitoring platform: Linux

ROCm/TheRock#7961 · 3 comments · 1 assignee · Opened Sep 7, 2026
MIOpen on Windows native ROCm (gfx1100): 'Invalid elapsed time detected in EvaluateInvokers' - find results never cached, full solver search re-runs on every inference Open

status: triage

ROCm/TheRock#7958 · 1 assignee · Opened Sep 6, 2026
[Issue]: gfx1151/Strix Halo: svm_range_restore_work busy-loops a kworker at 100% for days when an idle ROCm process has SVM pages swapped out Open

status: triage

ROCm/TheRock#7956 · 1 comment · 1 assignee · Opened Sep 6, 2026
rocminfo fails with HSA_STATUS_ERROR: A generic error has occurred on gfx1152 (Ryzen AI 5 340 / Radeon 840M, Strix Point) — reproducible on both ROCm 7.13.0 (pre-release) and ROCm 10.0.0 (stable) Open

status: triage

ROCm/TheRock#7952 · 2 comments · 1 assignee · Opened Sep 5, 2026
[Issue]: Backport offload-arch Windows LoadLibraryExW path fix Open

platform: Windows project: compiler status: triage

ROCm/TheRock#7950 · 1 comment · 1 assignee · Opened Sep 5, 2026

Showing the newest 100

This page lists what was indexed most recently. Advanced filter has the whole inventory, narrowed by language, difficulty and how long a task takes.

Open advanced filter

ROCm

Showing the newest 100

Get new issues in your inbox