Revert "[origami] Major refactoring of codebase (#2718)"

[jayhawk-commits]

It was determined that this commit introduces memory corruption issues with TheRock python packaging build. See ROCm/TheRock#2522 and need this commit reverted to move the rocm-libraries submodule pointer in TheRock forward.

Error

corrupted size vs. prev_size in fastbins

Affected Versions

• python 3.11, 3.12, 3.13
• pytorch 2.7, 2.8, 2.9, main branches
• TheRock with rocm-libraries submodule pointed to this commit or newer

Ways to Reproduce

1. Build ROCm via TheRock.
2. Build pytorch wheels via TheRock.
3. Run rocm-sdk test.
4. Run python ./external-builds/pytorch/run_pytorch_smoke_tests.py.

Error happens during steps 2-4. I have not seen an instance of a workflow getting past step 4.

Investigation Test Results A

• pytorch build workflow with rocm-libraries submodule at 283f438
• Run: https://github.com/ROCm/TheRock/actions/runs/20254886491/job/58154625946

Investigation Test Results B

• pytorch build workflow with rocm-libraries submodule at above SHA-1: e0fa9ee
• Run: https://github.com/ROCm/TheRock/actions/runs/20254276739/job/58154885187

Test Results A show corrupted size vs. prev_size in fastbins throughout the different python 3.11-3.13 runs.
Test Results B do not show the issue.

Revert Sequence

1. Cloned latest rocm-libraries.
2. Used CLI to execute git revert 283f438877a0982f170f76d5441dcca365741aa6
3. Manually resolved merge conflict in projects/hipblaslt/library/src/amd_detail/rocblaslt/src/rocroller/solution_selection.cpp, referring to the diffs in PR [origami] Major refactoring of codebase #2718

Verifying This Pull Request

1. Creating a branch on TheRock that points to this PR's branch as the rocm-libraries submodule. See https://github.com/ROCm/TheRock/tree/users/jayhawk-commits/revert-origami-refactor-test
2. Run pytorch build workflow with that branch to verify the error does not occur.

Verification Results

• Release portable Linux packages run: https://github.com/ROCm/TheRock/actions/runs/20260025013
• Release portable Linux PyTorch Wheels run will follow. Will post link when I wake up.

[marbre]

The verification build (https://github.com/ROCm/TheRock/actions/runs/20260025013) fails building hipBLASLt:

[hipBLASLt] /therock/src/rocm-libraries/projects/hipblaslt/library/src/amd_detail/rocblaslt/src/rocroller/solution_selection.cpp:202:20: error: no member named 'scaleABlockRowSize' in 'KernelType'
[hipBLASLt]   202 |         kernelType.scaleABlockRowSize * kernelType.scaleABlockColSize, //Handle A vs B block size.
[hipBLASLt]       |         ~~~~~~~~~~ ^
[hipBLASLt] /therock/src/rocm-libraries/projects/hipblaslt/library/src/amd_detail/rocblaslt/src/rocroller/solution_selection.cpp:202:52: error: no member named 'scaleABlockColSize' in 'KernelType'
[hipBLASLt]   202 |         kernelType.scaleABlockRowSize * kernelType.scaleABlockColSize, //Handle A vs B block size.
[hipBLASLt]       |                                         ~~~~~~~~~~ ^
[hipBLASLt] /therock/src/rocm-libraries/projects/hipblaslt/library/src/amd_detail/rocblaslt/src/rocroller/solution_selection.cpp:245:27: error: use of undeclared identifier 'analytical_hardware'; did you mean 'analaytical_hardware'?
[hipBLASLt]   245 |             if(numTiles < analytical_hardware.N_CU && !isF6)
[hipBLASLt]       |                           ^~~~~~~~~~~~~~~~~~~
[hipBLASLt]       |                           analaytical_hardware
[hipBLASLt] /therock/src/rocm-libraries/projects/hipblaslt/library/src/amd_detail/rocblaslt/src/rocroller/solution_selection.cpp:185:31: note: 'analaytical_hardware' declared here
[hipBLASLt]   185 |     const origami::hardware_t analaytical_hardware = origami::hardware_t::get_hardware_for_device(0);
[hipBLASLt]       |                               ^
[hipBLASLt] 3 errors generated.

[jayhawk-commits]

It looks like #2732 and #3177 touch many of the same files. Will need hipblaslt developers to help with reverting these changes.

[ScottTodd]

It looks like #2732 and #3177 touch many of the same files. Will need hipblaslt developers to help with reverting these changes.

Revert all of them. See https://github.com/ROCm/TheRock/blob/main/docs/rfcs/RFC0002-MonoRepo-Gardener-Rotations.md#reverting-vs-fixing-forward

If multiple commits are implicated in broken workflows, cleanly reverting them all sequentially is less risky than reverting manually.

[jayhawk-commits]

Replaced by #3416