[ROCm][Windows] Fixing undefined symbol linker error after exposing MIOpen symbols#2415

Merged

jeffdaily merged 84 commits intorelease/2.8from

fix_torch_macros_for_miopen

Aug 25, 2025

Commits on Jul 18, 2025

[release/2.8] Enable wheels
jithunnair-amd
committed
Updates to build for Noble (Ubuntu 24.04) and py3.12
jithunnair-amd
committed
[release/2.8] Make triton build ROCm version agnostic

ethanwee1
and
jithunnair-amd
committed
[release/2.8] Replace upstream install_rocm_magma.sh with rocm fork version
jithunnair-amd
committed
[release/2.8] Upgrade numpy versions; Use different package versions for py3.9;

jithunnair-amd
and
ethanwee1
committed
[release/2.8] Removing --user flag from all pip install commands

ethanwee1
and
jithunnair-amd
committed
[ROCm] Remove use of warpsize on host-side compilation (pytorch#156979)
jithunnair-amd
committed
[release/2.8] Improve C10_WARP_SIZE compatibility

xinyazhang
and
jithunnair-amd
committed
Fix sha256 for aotriton ROCm7.0 tarball
jithunnair-amd
committed
Update third_party/composable_kernel submodule commit as per https://github.com/ROCm/pytorch/commit/80cca7006d94df97ee932fd5903ed20c08c2eb34 to enable PyTorch build on ROCm7.0
jithunnair-amd
committed
Use ROCm/triton and update triton.txt
jithunnair-amd
committed

Commits on Jul 22, 2025

Add related_commits file (#2396)
pragupta
authored
Add QA automation scripts for running PyTorch unit tests
jithunnair-amd
committed
[release/2.6] enable NHWC batchnorm with MIOpen (#2023)

dnikolaev-amd
and
jithunnair-amd
committed
test_decompose_mem_bound_mm.py tolerance increase for navi3x

iupaikov-amd
authored and
jithunnair-amd
committed
[release/2.7] enable NHWC batchnorm by default on ROCm7.0+ (#2180)

dnikolaev-amd
authored and
jithunnair-amd
committed
[release/2.7] import 'Dict' to fix common_utils.py (#2181)

dnikolaev-amd
authored and
jithunnair-amd
committed
[AUTOGENERATED] [release/2.7] [rocm6.4_internal_testing] Replaced ROCm specific skips to generalized conditions (#2126)

authored and
jithunnair-amd
committed
[release/2.7] [SWDEV-535259] enable miopen channels last 3d for conv and batchnorm (#2232)

authored and
jithunnair-amd
committed
[AUTOGENERATED] [release/2.7] Add 3D batchnorm tests (#2243)

authored and
jithunnair-amd
committed
[AUTOGENERATED] [release/2.5] [ROCm][layer_norm] Use __builtin_amdgcn_rcpf(x) instead of 1.f/x (#1800)

authored and
jithunnair-amd
committed
[release/2.6] remove xfail from 'batch_norm_with_update' (#2070)

dnikolaev-amd
authored and
jithunnair-amd
committed
[release/2.7] Enable mx fp8 support on ROCm (#2199)

jagadish-amd
authored and
jithunnair-amd
committed
Extend CK gemm/sdpa support to gfx950 (#45)

alugorey
authored and
jithunnair-amd
committed
[release/2.6] [SWDEV-529824] Fix Float16 CooperativeReduction Test Failure (#2204)

authored and
jithunnair-amd
committed
[ROCm] Set thread_work_size to 16 for vectorized elementwise kernels (#2259)

jerrymannil
authored and
jithunnair-amd
committed
[release/2.7] Fix SDPA skip logic (#2281)

AmdSampsa
authored and
jithunnair-amd
committed
[release/2.7] Update test_binary_ufuncs.py after numpy upgrade (#2289)

ethanwee1
authored and
jithunnair-amd
committed
[AUTOGENERATED] [release/2.7] fix jit_utils.cpp (#2320)

authored and
jithunnair-amd
committed
Clean up CUDA state between tests (#2335)

rraminen
authored and
jithunnair-amd
committed
cublaslt/hipblaslt persistent workspace (#156495)

jeffdaily
authored and
jithunnair-amd
committed
[AUTOGENERATED] [release/2.7] [release/2.6] Fix dtype before comparing torch and numpy tensors (#2362)

authored and
jithunnair-amd
committed

Commits on Jul 25, 2025

[ROCm][Windows] Fixing undefined symbol linker error after exposing MIOpen symbols (#156479)

tvukovic-amd
and
jeffdaily
committed

Commits on Aug 20, 2025

[MPS] Fix `index_kernel` for large tensors (#158239)

authored and
tvukovic-amd
committed
Add flag to fx.passes.split_module to normalize input names (#157793)

authored and
tvukovic-amd
committed
Add warning about removed sm50 and sm60 arches (#158478)

authored and
tvukovic-amd
committed
[cherry-pick][inductor][triton] Update HAS_WARP_SPEC to check triton.Config params. Update Triton Hash to top of release/3.4.x stack (#158646)

authored and
tvukovic-amd
committed
[CUDA] Use runtime driver API for cuStreamWriteValue32 (#158585)

authored and
tvukovic-amd
committed
Add stride check for attn_mask on non-cpu device (#158618)

CaoE
authored and
tvukovic-amd
committed
[cherry-pick] temporarily disabling generation of weblinks for torch v2.8 & removing string literals for weblink generation (#157951)

Sidharth123-cpu
authored and
tvukovic-amd
committed
[Reland] Add warning about removed sm50 and sm60 arches (#158744)

atalman
authored and
tvukovic-amd
committed
[cherry-pick][release 2.8] Update OpenBLAS commit (#151547) (#158243)

authored and
tvukovic-amd
committed
[cherry-pick][Docker builds] Move from Miniconda to Miniforge (#158370) (#158756)

authored and
tvukovic-amd
committed
[async-TP] Turn asserts back into silent skips (#158736)

authored and
tvukovic-amd
committed
[cherry-pick] Fix AArch64 segfaults by disabling strict-aliasing in GridSamplerKernel for GCC 12 and above (#158445)

robert-hardwick
authored and
tvukovic-amd
committed
Pull latest Sphinx theme (#158595) (#158673)

svekars
authored and
tvukovic-amd
committed
[Dynamo] Use proper sources for constructing dataclass defaults (#158689)

authored and
tvukovic-amd
committed
[cherry-pick] Unify torch.tensor and torch.ops.aten.scalar_tensor behavior (#158537) (#158655)

authored and
tvukovic-amd
committed
Cherry pick PR 158746 (#158801)

svekars
authored and
tvukovic-amd
committed
[MPS] Reimplement `tri[ul]` as Metal shaders (#158867)

authored and
tvukovic-amd
committed
[MPS] Switch Cholesky decomp to column wise (#158237)

authored and
tvukovic-amd
committed
Revert "[Dynamo] Allow inlining into AO quantization modules (#152934)" (#158677)

atalman
authored and
tvukovic-amd
committed
Move out super large one off foreach_copy test (#158880)

authored and
tvukovic-amd
committed
[Release Only] Remove nvshmem from list of preload libraries (#158925)

atalman
authored and
tvukovic-amd
committed
Use ROCm/triton and update triton.txt

jithunnair-amd
authored and
tvukovic-amd
committed
[release/2.8] [Bugfix][Inductor] Fix dependency list merged incorrectly for a custo… (#2419)

authored and
tvukovic-amd
committed
[release/2.8] enable py3.13 (#2366)

ethanwee1
authored and
tvukovic-amd
committed
[SWDEV-539076] Initial naive foreach autotune support (#2377)

jataylo
authored and
tvukovic-amd
committed
[release/2.7][ROCm][tunableop] UT tolerance increase for matmul_small_brute_force_… (#2397)

naromero77amd
authored and
tvukovic-amd
committed
[release/2.7] [SWDEV-543214] Reland #2416 Fix warps runtime (#2421)

authored and
tvukovic-amd
committed
[AUTOGENERATED] [release/2.8] [ROCm] Use opportunistic fastatomics based on heuristics (#2441)

okakarpa
authored and
tvukovic-amd
committed
Update triton pin for gfx950 improvements (#2443)

jataylo
authored and
tvukovic-amd
committed
[AUTOGENERATED] [release/2.8] [release/2.7] [SWDEV-543214] Reland #2416 Fix warps runtime part 2 (#2455)

authored and
tvukovic-amd
committed
[AUTOGENERATED] [release/2.8] [ROCm] Limit number of values per thread for reductions on three dimensions (#2469)

authored and
tvukovic-amd
committed
[release/2.8] Define datatypes when ROCM_VERSION >= 70000 (#2470)

rraminen
authored and
tvukovic-amd
committed
[release/2.8] Add mx fp4 support (#2472)

authored and
tvukovic-amd
committed
[AUTOGENERATED] [release/2.8] [rocm7.0_internal_testing] skip test_transformer_req_grad on Navi32/Navi4x (#2464)

authored and
tvukovic-amd
committed
Update version as 2.8.0

jithunnair-amd
authored and
tvukovic-amd
committed
[release/2.8] pin requirements.txt (#2481)

ethanwee1
authored and
tvukovic-amd
committed
[AUTOGENERATED] [release/2.8] [SWDEV-539215] - Autotune support for persistent reduction and no_x_dim removal (#2454)

authored and
tvukovic-amd
committed
[release/2.8] fp8: skip rowwise tests (#2477)

jagadish-amd
authored and
tvukovic-amd
committed
[release/2.8] update related_commit (#2490)

amd-sriram
authored and
tvukovic-amd
committed
[SWDEV-539119] [release/2.8] Add fast_tanh support (#2484)

jataylo
authored and
tvukovic-amd
committed
[AUTOGENERATED] [release/2.8] remove extra transposes in NHWC convolutions on MIOpen (#2410)

authored and
tvukovic-amd
committed
[release/2.8] [triton] Triton bump to fix ROCm 7.0 issues (#2498)

iupaikov-amd
authored and
tvukovic-amd
committed
[AUTOGENERATED] [release/2.8] [ROCm] Improve reduction sum performance (#2505)

authored and
tvukovic-amd
committed
[release/2.8] Using c10d.barrier() in test_extra_cuda_context test in test_c10d_nccl.py (#2522)

authored and
tvukovic-amd
committed
[AUTOGENERATED] [release/2.8] Change triton package name depending on rocm version (#2529)

authored and
tvukovic-amd
committed
[AUTOGENERATED] [release/2.8] NAVI32 specific fixes (#2467)

authored and
tvukovic-amd
committed
[release/2.8] Define uint32 t when ROCM_VERSION >= 70000 (#2513)

rraminen
authored and
tvukovic-amd
committed
[AUTOGENERATED] [release/2.8] Remove tb-nightly (#2538)

authored and
tvukovic-amd
committed
Use ROCm/triton and update triton.txt

jithunnair-amd
authored and
tvukovic-amd
committed
Merge branch 'release/2.8' into fix_torch_macros_for_miopen
tvukovic-amd
authored

Commits on Aug 25, 2025

revert triton change
tvukovic-amd
committed