[GitHub Actions] python-based workflows based on files changed in a pull request by jayhawk-commits · Pull Request #8 · ROCm/rocm-libraries

jayhawk-commits · 2025-05-01T19:22:16Z

Key Workflows

Pull Request Fanout

When a pull request is opened or modified on the monorepo, determine which projects were changed. Fanout and create pull requests on those project's repos.
Those fanout pull requests will get the changed files for those individual projects and trigger the existing CICD pipelines on those repos.
Copying the results of the fanned out pull requests back to this monorepo pull request will be done in a separate change, as this change is already large to review and test.
For interfacing between the monorepo and subrepos, the pull request branches follow the naming convention: monorepo-pr-<monorepo-pr-number>-<subrepo-name>

Auto-Label

When a pull request is opened or modified on the monorepo, apply category labels based on files affected.
When a label is added to a monorepo pull request, apply the label to the fanned out pull requests in the subrepos if they have the label.
This latter mirroring is only one way from monorepo to individual repos. When a label is removed from the monorepo pull request, nothing happens.
As some teams use labels to determine what CICD pipelines are run, this mirroring is helpful for migration phase.
No new labels are created on the repo-level with this job. Creation of the expected labels are handled elsewhere.
Need to learn more about team practices to see if we need to mirror label removal on the monorepo pull request.

Pull Request Close Fanout

When a pull request is closed (including merges), close all the pull requests that were fanned out and delete those branches.
This is done through the naming convention of the fanned out branches.

Key Considerations

Sparse-Checkouts

Sparse-checkouts are leveraged as much as possible to clone as little code as possible.
The git subtree operations require fetch-depth of 0.

gh command-line interface and Authentication

The gh executable tool is used as much as possible for read/write operations.
'git subtree' is one exception where the raw command is executed.
This tool does not have support for writing checks, so the REST API will be used for that.
However, that also has token implications: https://docs.github.com/en/rest/checks/runs?apiVersion=2022-11-28
Authentication follows the convention in: https://docs.github.com/en/actions/writing-workflows/choosing-what-your-workflow-does/using-github-cli-in-workflows
The default GITHUB_TOKEN is used when operations in the workflow only involve the monorepo.
For operations involve the subrepos, a separate personal access token was created and defined as a secret.
The permissions for this access token is limited to specific repos and operations.
Based on the limitation on check-runs linked above, a GitHub App should probably be made.

python structure

Common code is stripped out into separate files.
The json defining the projects in the monorepo is now defined by a model.
All gh usage is done through a client class.
python script file names use underscores to follow PEP.
All python scripts are in one folder for now. We can figure out how we want to organize scripts as more get added.

Test Sequences

Environment:

A sandbox monorepo with forks of hipcub and rocthrust was created.
Simple randomized-result workflow was added to these project forks.
Prerequisite labels were added for testing.
Follow the sequence linked pull request events and comments describing them.
You can view action links here: https://github.com/amd-jmacaran/libs-mono-test/actions
Since my personal access token was used, my github account is attached to automated actions.
Assume all the operations done on the subrepos were automatically performed.

Test Sequence One:

Only one project in the monorepo is changed in a pull request.
Wait for workflows after pull request creation. See if a pull request is created on the subrepo.
Add label to the monorepo pull request. Monitor the subrepo pull request.
Add another label to the monorepo pull request. Monitor the subrepo pull request.
Close the monorepo pull request. Monitor the subrepo pull request.

Test Sequence Two:

Two projects in the monorepo are changed in a pull request.
Have a label added before hitting the create pull request button.
Wait for workflows after pull request creation. See if pull requests are created on the subrepo with corresponding labels.
Add label specific to one project to the monorepo pull request. Monitor the subrepo pull requests.
Remove changes for one project. Monitor the monorepo pull request's category labels. This subrepo is now ignored, so nothing happens on that pull request for now.
Close the monorepo pull request. Monitor the two subrepo pull requests.

Test Sequence Three:

Only one project in the monorepo is changed in a pull request.
Wait for workflows after pull request creation. See if a pull request is created on the subrepo.
Change additional files in the project in the monorepo pull request. Monitor the subrepo pull request.
Undo some of the changes in the project in the monorepo pull request. Monitor the subrepo pull request.

Two workflows and corresponding python scripts that take a look at files changed in a pull request on the monorepo. Then based on that output, perform tasks tied to the old individual repos. One workflow is applying labels. The other workflow is to fan out and create pull requests. There is some placeholder remaining on the second workflow.

geomin12

some housekeeping / organizational items! looks good and will re-review upon more work done on the second part

- Refactored script code into smaller functions. - Converted GitHub rest api calls to use gh cli. - Moved out all gh cli usage to a separate class. - Try to make the scripts as similar to each other, when applicable. - Leverage dependabot to handle GitHub Actions versioning.

jayhawk-commits · 2025-05-05T16:30:23Z

Applied most of the provided feedback, but still in the middle of fixing things. I am pushing to remote as I plan on moving between work spaces today (home versus office). I will share some test logs when I am ready for further feedback.

- Use pydantic to define the entries in the json file and for validation when the json file is consumed. - reponame field no longer used by any existing scripts.

- gh cli doesn't support writing GitHub checks, so REST API created to support this. - Implement the functions that synchronize the custom labels and GitHub Checks between the monorepo PR and the fanned out PRs.

…#8) * fix clang-format commands to use find correctly. Original issue with find was cmake escaping things oddly. * clang-tidy works with this setup * clang tidy now work as i expect. The headers are included using a positive filter, negtive filters dont work. * updated clang-tidy and copilot instructions * fix all repo violations, update clang-tidy for things were not going to bother dealing with. * add option to turn on/off clangtidy. default is off * un nesting namespaces, running full format on repo * ensure all files are checked by clang tidy. Add one more exclusion for the generated export header that I cant figure out how to exclude. * add more copilot instructions * fix incorrect name

…#8) * fix clang-format commands to use find correctly. Original issue with find was cmake escaping things oddly. * clang-tidy works with this setup * clang tidy now work as i expect. The headers are included using a positive filter, negtive filters dont work. * updated clang-tidy and copilot instructions * fix all repo violations, update clang-tidy for things were not going to bother dealing with. * add option to turn on/off clangtidy. default is off * un nesting namespaces, running full format on repo * ensure all files are checked by clang tidy. Add one more exclusion for the generated export header that I cant figure out how to exclude. * add more copilot instructions * fix incorrect name [ROCm/hipDNN commit: f8dce7f]

## Motivation Enable gfx1152 and gfx1153. ## Technical Details 1. combine arrays into tables and use local macros to reduce repetition (for maintainability) 2. monkey-see-monkey-do wherever `gfx11...` was found ## Test Plan Build existing ctests for, and run them on, gfx1152 and gfx1153. ## Test Result ### <details> <summary>gfx1152 passed (click to see log)</summary> ``` INFO:root:++ Exec [/tmp/eble]$ ctest --test-dir /tmp/eble/rocm/bin/rocprim --output-o n-failure --parallel 2 --exclude-regex 'rocprim.lookback_reproducibility|rocprim.link ing|rocprim.device_merge_inplace|rocprim.device_merge_sort|rocprim.device_partition|r ocprim.device_radix_sort|rocprim.device_scan|rocprim.device_select|rocprim.device_fin d_first_of|rocprim.device_reduce_by_key' --timeout 60 Test project /tmp/eble/rocm/bin/rocprim Start 1: hip.device_api Start 2: hip.async_copy 1/73 Test #2: hip.async_copy .............................. Passed 0.01 sec Start 3: hip.ordered_block_id 2/73 Test #1: hip.device_api .............................. Passed 0.02 sec Start 4: rocprim.internal_merge_path 3/73 Test #4: rocprim.internal_merge_path ................. Passed 0.01 sec Start 5: rocprim.basic_test 4/73 Test #3: hip.ordered_block_id ........................ Passed 0.01 sec Start 6: rocprim.arg_index_iterator 5/73 Test #5: rocprim.basic_test .......................... Passed 0.01 sec Start 7: rocprim.temporary_storage_partitioning 6/73 Test #6: rocprim.arg_index_iterator .................. Passed 0.01 sec Start 8: rocprim.block_adjacent_difference 7/73 Test #7: rocprim.temporary_storage_partitioning ...... Passed 0.01 sec Start 9: rocprim.block_discontinuity 8/73 Test #8: rocprim.block_adjacent_difference ........... Passed 2.34 sec Start 10: rocprim.bit_cast 9/73 Test #10: rocprim.bit_cast ............................ Passed 0.02 sec Start 11: rocprim.block_exchange 10/73 Test #11: rocprim.block_exchange ...................... Passed 0.73 sec Start 12: rocprim.block_histogram 11/73 Test #12: rocprim.block_histogram ..................... Passed 0.54 sec Start 13: rocprim.block_load_store 12/73 Test #13: rocprim.block_load_store .................... Passed 0.44 sec Start 14: rocprim.block_sort_merge 13/73 Test #14: rocprim.block_sort_merge .................... Passed 0.02 sec Start 15: rocprim.block_sort_merge_stable 14/73 Test #15: rocprim.block_sort_merge_stable ............. Passed 0.02 sec Start 16: rocprim.block_radix_rank 15/73 Test #16: rocprim.block_radix_rank .................... Passed 0.03 sec Start 17: rocprim.block_radix_sort 16/73 Test #17: rocprim.block_radix_sort .................... Passed 4.79 sec Start 18: rocprim.block_reduce 17/73 Test #18: rocprim.block_reduce ........................ Passed 0.26 sec Start 19: rocprim.block_run_length_decode 18/73 Test #19: rocprim.block_run_length_decode ............. Passed 0.54 sec Start 20: rocprim.block_scan 19/73 Test #20: rocprim.block_scan .......................... Passed 0.04 sec Start 21: rocprim.block_shuffle 20/73 Test #21: rocprim.block_shuffle ....................... Passed 2.70 sec Start 22: rocprim.block_sort_bitonic 21/73 Test #9: rocprim.block_discontinuity ................. Passed 17.36 sec Start 23: rocprim.config_dispatch 22/73 Test #23: rocprim.config_dispatch ..................... Passed 0.09 sec Start 24: rocprim.constant_iterator 23/73 Test #24: rocprim.constant_iterator ................... Passed 0.07 sec Start 25: rocprim.counting_iterator 24/73 Test #25: rocprim.counting_iterator ................... Passed 0.07 sec Start 26: rocprim.device_batch_memcpy 25/73 Test #26: rocprim.device_batch_memcpy ................. Passed 1.20 sec Start 27: rocprim.device_binary_search 26/73 Test #27: rocprim.device_binary_search ................ Passed 0.02 sec Start 28: rocprim.device_adjacent_difference 27/73 Test #28: rocprim.device_adjacent_difference .......... Passed 0.01 sec Start 29: rocprim.device_adjacent_find 28/73 Test #29: rocprim.device_adjacent_find ................ Passed 0.01 sec Start 30: rocprim.device_find_end 29/73 Test #30: rocprim.device_find_end ..................... Passed 0.01 sec Start 31: rocprim.device_histogram 30/73 Test #22: rocprim.block_sort_bitonic .................. Passed 13.23 sec Start 32: rocprim.device_merge 31/73 Test #31: rocprim.device_histogram .................... Passed 7.85 sec Start 33: rocprim.nth_element 32/73 Test #33: rocprim.nth_element ......................... Passed 0.03 sec Start 34: rocprim.device_partial_sort 33/73 Test #34: rocprim.device_partial_sort ................. Passed 0.02 sec Start 35: rocprim.device_reduce 34/73 Test #35: rocprim.device_reduce ....................... Passed 8.94 sec Start 36: rocprim.device_run_length_encode 35/73 Test #32: rocprim.device_merge ........................ Passed 14.05 sec Start 37: rocprim.device_search 36/73 Test #37: rocprim.device_search ....................... Passed 0.02 sec Start 38: rocprim.device_segmented_radix_sort 37/73 Test #36: rocprim.device_run_length_encode ............ Passed 13.92 sec Start 39: rocprim.device_search_n 38/73 Test #39: rocprim.device_search_n ..................... Passed 0.02 sec Start 40: rocprim.device_segmented_reduce 39/73 Test #40: rocprim.device_segmented_reduce ............. Passed 5.21 sec Start 41: rocprim.device_segmented_scan 40/73 Test #41: rocprim.device_segmented_scan ............... Passed 0.02 sec Start 42: rocprim.device_transform 41/73 Test #42: rocprim.device_transform .................... Passed 13.90 sec Start 43: rocprim.discard_iterator 42/73 Test #43: rocprim.discard_iterator .................... Passed 0.07 sec Start 44: rocprim.radix_key_codec 43/73 Test #44: rocprim.radix_key_codec ..................... Pas09:54:11 [55/1943] Start 45: rocprim.predicate_iterator 44/73 Test #45: rocprim.predicate_iterator .................. Passed 0.07 sec Start 46: rocprim.reverse_iterator 45/73 Test #46: rocprim.reverse_iterator .................... Passed 0.09 sec Start 47: rocprim.rocprim_tuple 46/73 Test #47: rocprim.rocprim_tuple ....................... Passed 0.01 sec Start 48: rocprim.rocprim_types 47/73 Test #48: rocprim.rocprim_types ....................... Passed 0.01 sec Start 49: rocprim.texture_cache_iterator 48/73 Test #49: rocprim.texture_cache_iterator .............. Passed 0.01 sec Start 50: rocprim.thread 49/73 Test #50: rocprim.thread .............................. Passed 0.07 sec Start 51: rocprim.thread_algos 50/73 Test #51: rocprim.thread_algos ........................ Passed 0.35 sec Start 52: rocprim.tuple 51/73 Test #52: rocprim.tuple ............................... Passed 0.02 sec Start 53: rocprim.utils_sort_checker 52/73 Test #53: rocprim.utils_sort_checker .................. Passed 0.01 sec Start 54: rocprim.transform_iterator 53/73 Test #54: rocprim.transform_iterator .................. Passed 0.11 sec Start 55: rocprim.type_traits_interface_cpp17 54/73 Test #55: rocprim.type_traits_interface_cpp17 ......... Passed 0.01 sec Start 56: rocprim.type_traits_interface_gnupp17 55/73 Test #56: rocprim.type_traits_interface_gnupp17 ....... Passed 0.01 sec Start 57: rocprim.type_traits_interface_cpp20 56/73 Test #57: rocprim.type_traits_interface_cpp20 ......... Passed 0.01 sec Start 58: rocprim.type_traits_interface_gnupp20 57/73 Test #58: rocprim.type_traits_interface_gnupp20 ....... Passed 0.01 sec Start 59: rocprim.no_half_operators 58/73 Test #59: rocprim.no_half_operators ................... Passed 0.01 sec Start 60: rocprim.intrinsics 59/73 Test #60: rocprim.intrinsics .......................... Passed 0.21 sec Start 61: rocprim.intrinsics_atomic 60/73 Test #61: rocprim.intrinsics_atomic ................... Passed 0.02 sec Start 62: rocprim.invoke_result 61/73 Test #62: rocprim.invoke_result ....................... Passed 0.01 sec Start 63: rocprim.warp_exchange 62/73 Test #63: rocprim.warp_exchange ....................... Passed 0.08 sec Start 64: rocprim.warp_load 63/73 Test #64: rocprim.warp_load ........................... Passed 0.08 sec Start 65: rocprim.warp_reduce 64/73 Test #65: rocprim.warp_reduce ......................... Passed 0.14 sec Start 66: rocprim.warp_scan 65/73 Test #66: rocprim.warp_scan ........................... Passed 0.20 sec Start 67: rocprim.warp_scan_disable_dpp_disable_dpp 66/73 Test #67: rocprim.warp_scan_disable_dpp_disable_dpp ... Passed 0.21 sec Start 68: rocprim.warp_sort 67/73 Test #68: rocprim.warp_sort ........................... Passed 0.09 sec Start 69: rocprim.warp_store 68/73 Test #69: rocprim.warp_store .......................... Passed 0.02 sec Start 70: rocprim.zip_iterator 69/73 Test #70: rocprim.zip_iterator ........................ Passed 0.02 sec Start 71: rocprim.accumulator_t 70/73 Test #71: rocprim.accumulator_t ....................... Passed 0.02 sec Start 72: hipgraph.basic 71/73 Test #72: hipgraph.basic .............................. Passed 0.02 sec Start 73: hipgraph.algs 72/73 Test #73: hipgraph.algs ............................... Passed 0.01 sec 73/73 Test #38: rocprim.device_segmented_radix_sort ......... Passed 31.97 sec 100% tests passed, 0 tests failed out of 73 Total Test time (real) = 71.80 sec ✅ test_rocprim.py PASSED ``` </details> <details> <summary>gfx1153 passed (click to see log)</summary> ``` INFO:root:++ Exec [/tmp/eble]$ ctest --test-dir /tmp/eble/rocm/bin/rocprim --output-o n-failure --parallel 1 --exclude-regex 'rocprim.lookback_reproducibility|rocprim.link ing|rocprim.device_merge_inplace|rocprim.device_merge_sort|rocprim.device_partition|r ocprim.device_radix_sort|rocprim.device_scan|rocprim.device_select|rocprim.device_fin d_first_of|rocprim.device_reduce_by_key' --timeout 60 Test project /tmp/eble/rocm/bin/rocprim Start 1: hip.device_api 1/73 Test #1: hip.device_api .............................. Passed 0.01 sec Start 2: hip.async_copy 2/73 Test #2: hip.async_copy .............................. Passed 0.01 sec Start 3: hip.ordered_block_id 3/73 Test #3: hip.ordered_block_id ........................ Passed 0.01 sec Start 4: rocprim.internal_merge_path 4/73 Test #4: rocprim.internal_merge_path ................. Passed 0.01 sec Start 5: rocprim.basic_test 5/73 Test #5: rocprim.basic_test .......................... Passed 0.01 sec Start 6: rocprim.arg_index_iterator 6/73 Test #6: rocprim.arg_index_iterator .................. Passed 0.01 sec Start 7: rocprim.temporary_storage_partitioning 7/73 Test #7: rocprim.temporary_storage_partitioning ...... Passed 0.01 sec Start 8: rocprim.block_adjacent_difference 8/73 Test #8: rocprim.block_adjacent_difference ........... Passed 2.94 sec Start 9: rocprim.block_discontinuity 9/73 Test #9: rocprim.block_discontinuity ................. Passed 21.10 sec Start 10: rocprim.bit_cast 10/73 Test #10: rocprim.bit_cast ............................ Passed 0.01 sec Start 11: rocprim.block_exchange 11/73 Test #11: rocprim.block_exchange ...................... Passed 2.20 sec Start 12: rocprim.block_histogram 12/73 Test #12: rocprim.block_histogram ..................... Passed 0.71 sec Start 13: rocprim.block_load_store 13/73 Test #13: rocprim.block_load_store .................... Passed 0.48 sec Start 14: rocprim.block_sort_merge 14/73 Test #14: rocprim.block_sort_merge .................... Passed 0.02 sec Start 15: rocprim.block_sort_merge_stable 15/73 Test #15: rocprim.block_sort_merge_stable ............. Passed 0.02 sec Start 16: rocprim.block_radix_rank 16/73 Test #16: rocprim.block_radix_rank .................... Passed 0.02 sec Start 17: rocprim.block_radix_sort 17/73 Test #17: rocprim.block_radix_sort .................... Passed 6.12 sec Start 18: rocprim.block_reduce 18/73 Test #18: rocprim.block_reduce ........................ Passed 0.31 sec Start 19: rocprim.block_run_length_decode 19/73 Test #19: rocprim.block_run_length_decode ............. Passed 0.68 sec Start 20: rocprim.block_scan 20/73 Test #20: rocprim.block_scan .......................... Passed 0.03 sec Start 21: rocprim.block_shuffle 21/73 Test #21: rocprim.block_shuffle ....................... Passed 3.63 sec Start 22: rocprim.block_sort_bitonic 22/73 Test #22: rocprim.block_sort_bitonic .................. Passed 19.34 sec Start 23: rocprim.config_dispatch 23/73 Test #23: rocprim.config_dispatch ..................... Passed 0.10 sec Start 24: rocprim.constant_iterator 24/73 Test #24: rocprim.constant_iterator ................... Passed 0.09 sec Start 25: rocprim.counting_iterator 25/73 Test #25: rocprim.counting_iterator ................... Passed 0.09 sec Start 26: rocprim.device_batch_memcpy 26/73 Test #26: rocprim.device_batch_memcpy ................. Passed 1.42 sec Start 27: rocprim.device_binary_search 27/73 Test #27: rocprim.device_binary_search ................ Passed 0.01 sec Start 28: rocprim.device_adjacent_difference 28/73 Test #28: rocprim.device_adjacent_difference .......... Passed 0.01 sec Start 29: rocprim.device_adjacent_find 29/73 Test #29: rocprim.device_adjacent_find ................ Passed 0.01 sec Start 30: rocprim.device_find_end 30/73 Test #30: rocprim.device_find_end ..................... Passed 0.01 sec Start 31: rocprim.device_histogram 31/73 Test #31: rocprim.device_histogram .................... Passed 8.77 sec Start 32: rocprim.device_merge 32/73 Test #32: rocprim.device_merge ........................ Passed 16.23 sec Start 33: rocprim.nth_element 33/73 Test #33: rocprim.nth_element ......................... Passed 0.01 sec Start 34: rocprim.device_partial_sort 34/73 Test #34: rocprim.device_partial_sort ................. Passed 0.02 sec Start 35: rocprim.device_reduce 35/73 Test #35: rocprim.device_reduce ....................... Passed 10.92 sec Start 36: rocprim.device_run_length_encode 36/73 Test #36: rocprim.device_run_length_encode ............ Passed 14.34 sec Start 37: rocprim.device_search 37/73 Test #37: rocprim.device_search ....................... Passed 0.01 sec Start 38: rocprim.device_segmented_radix_sort 38/73 Test #38: rocprim.device_segmented_radix_sort ......... Passed 38.28 sec Start 39: rocprim.device_search_n 39/73 Test #39: rocprim.device_search_n ..................... Passed 0.02 sec Start 40: rocprim.device_segmented_reduce 40/73 Test #40: rocprim.device_segmented_reduce ............. Passed 7.19 sec Start 41: rocprim.device_segmented_scan 41/73 Test #41: rocprim.device_segmented_scan ............... Passed 0.02 sec Start 42: rocprim.device_transform 42/73 Test #42: rocprim.device_transform .................... Passed 17.64 sec Start 43: rocprim.discard_iterator 43/73 Test #43: rocprim.discard_iterator .................... Passed 0.12 sec Start 44: rocprim.radix_key_codec 44/73 Test #44: rocprim.radix_key_codec ..................... Passed 0.01 sec Start 45: rocprim.predicate_iterator 45/73 Test #45: rocprim.predicate_iterator .................. Passed 0.08 sec Start 46: rocprim.reverse_iterator 46/73 Test #46: rocprim.reverse_iterator .................... Pas10:13:26 [46/1844] Start 47: rocprim.rocprim_tuple 47/73 Test #47: rocprim.rocprim_tuple ....................... Passed 0.01 sec Start 48: rocprim.rocprim_types 48/73 Test #48: rocprim.rocprim_types ....................... Passed 0.01 sec Start 49: rocprim.texture_cache_iterator 49/73 Test #49: rocprim.texture_cache_iterator .............. Passed 0.01 sec Start 50: rocprim.thread 50/73 Test #50: rocprim.thread .............................. Passed 0.08 sec Start 51: rocprim.thread_algos 51/73 Test #51: rocprim.thread_algos ........................ Passed 0.43 sec Start 52: rocprim.tuple 52/73 Test #52: rocprim.tuple ............................... Passed 0.01 sec Start 53: rocprim.utils_sort_checker 53/73 Test #53: rocprim.utils_sort_checker .................. Passed 0.01 sec Start 54: rocprim.transform_iterator 54/73 Test #54: rocprim.transform_iterator .................. Passed 0.12 sec Start 55: rocprim.type_traits_interface_cpp17 55/73 Test #55: rocprim.type_traits_interface_cpp17 ......... Passed 0.01 sec Start 56: rocprim.type_traits_interface_gnupp17 56/73 Test #56: rocprim.type_traits_interface_gnupp17 ....... Passed 0.01 sec Start 57: rocprim.type_traits_interface_cpp20 57/73 Test #57: rocprim.type_traits_interface_cpp20 ......... Passed 0.01 sec Start 58: rocprim.type_traits_interface_gnupp20 58/73 Test #58: rocprim.type_traits_interface_gnupp20 ....... Passed 0.01 sec Start 59: rocprim.no_half_operators 59/73 Test #59: rocprim.no_half_operators ................... Passed 0.01 sec Start 60: rocprim.intrinsics 60/73 Test #60: rocprim.intrinsics .......................... Passed 0.29 sec Start 61: rocprim.intrinsics_atomic 61/73 Test #61: rocprim.intrinsics_atomic ................... Pas10:13:27 [16/1844] Start 62: rocprim.invoke_result 62/73 Test #62: rocprim.invoke_result ....................... Passed 0.01 sec Start 63: rocprim.warp_exchange 63/73 Test #63: rocprim.warp_exchange ....................... Passed 0.09 sec Start 64: rocprim.warp_load 64/73 Test #64: rocprim.warp_load ........................... Passed 0.09 sec Start 65: rocprim.warp_reduce 65/73 Test #65: rocprim.warp_reduce ......................... Passed 0.17 sec Start 66: rocprim.warp_scan 66/73 Test #66: rocprim.warp_scan ........................... Passed 0.26 sec Start 67: rocprim.warp_scan_disable_dpp_disable_dpp 67/73 Test #67: rocprim.warp_scan_disable_dpp_disable_dpp ... Passed 0.26 sec Start 68: rocprim.warp_sort 68/73 Test #68: rocprim.warp_sort ........................... Passed 0.10 sec Start 69: rocprim.warp_store 69/73 Test #69: rocprim.warp_store .......................... Passed 0.01 sec Start 70: rocprim.zip_iterator 70/73 Test #70: rocprim.zip_iterator ........................ Passed 0.01 sec Start 71: rocprim.accumulator_t 71/73 Test #71: rocprim.accumulator_t ....................... Passed 0.01 sec Start 72: hipgraph.basic 72/73 Test #72: hipgraph.basic .............................. Passed 0.01 sec Start 73: hipgraph.algs 73/73 Test #73: hipgraph.algs ............................... Passed 0.01 sec 100% tests passed, 0 tests failed out of 73 Total Test time (real) = 175.34 sec ✅ test_rocprim.py PASSED ``` <detail> ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

…EV-536565 SWDEV-536565 Add missing includes to fix build

…being used after being freed. (#5220) ## Motivation A `heap-use-after-free` error was triggered by AddressSanitizer on test `CPU_Dump_NAN_FP32.testDump`. ## Technical Details Root Cause Analysis: The AddressSanitizer error occurred because the HIPOCProgramImpl constructor was not storing the binary data passed to it. When LoadProgram called LoadBinary and created a HIPOCProgram with the returned vector, the temporary vector would go out of scope, but COMGR still needed to access the binary data later, causing a use-after-free. - The fix ensures that the HIPOCProgramImpl object owns the binary data for its entire lifetime - Both constructors now consistently store the binary data in the `binary` member variable (std::vector) - The uint8_t constructor converts the data to char format using iterator range construction - This prevents the use-after-free that occurred when COMGR tried to access freed memory ## Test Plan Test output before change: ``` HSA_XNACK=1 ASAN_OPTIONS=symbolize=1 ./build/ml-libs/MIOpen/build/bin/miopen_gtest --gtest_filter="*CPU_Dump_NAN_FP32*" PRNG seed: 12345678 Note: Google Test filter = *CPU_Dump_NAN_FP32* [==========] Running 1 test from 1 test suite. [----------] Global test environment set-up. [----------] 1 test from CPU_Dump_NAN_FP32 [ RUN ] CPU_Dump_NAN_FP32.testDump ================================================================= ==3639==ERROR: AddressSanitizer: heap-use-after-free on address 0x7e0f08c50200 at pc 0x7f5f8d7a6554 bp 0x7ffcb7c4a730 sp 0x7ffcb7c49ee8 READ of size 26088 at 0x7e0f08c50200 thread T0 #0 0x7f5f8d7a6553 in memcpy /data/nhanna/repos/TheRock/compiler/amd-llvm/compiler-rt/lib/asan/../sanitizer_common/sanitizer_common_interceptors_memintrinsics.inc:117:5 #1 0x7f5f23d61d78 in COMGR::setCStr(char*&, llvm::StringRef, unsigned long*) /data/nhanna/repos/TheRock/compiler/amd-llvm/amd/comgr/src/comgr.cpp:216:9 #2 0x7f5f23d61d78 in COMGR::DataObject::setData(llvm::StringRef) /data/nhanna/repos/TheRock/compiler/amd-llvm/amd/comgr/src/comgr.cpp:334:17 #3 0x7f5f23d61d78 in amd_comgr_set_data /data/nhanna/repos/TheRock/compiler/amd-llvm/amd/comgr/src/comgr.cpp:606:24 #4 0x7f5f221dc1d3 in amd::Comgr::set_data(amd_comgr_data_s, unsigned long, char const*) /data/nhanna/repos/TheRock/rocm-systems/projects/clr/rocclr/device/comgrctx.hpp:252:12 #5 0x7f5f221dc1d3 in amd::device::Program::getSymbolsFromCodeObj(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>*, amd_comgr_symbol_type_s) const /data/nhanna/repos/TheRock/rocm-systems/projects/clr/rocclr/device/devprogram.cpp:2061:14 #6 0x7f5f219e6f7c in hip::DynCO::populateDynGlobalVars() /data/nhanna/repos/TheRock/rocm-systems/projects/clr/hipamd/src/hip_code_object.cpp:216:22 #7 0x7f5f219e8e6a in hip::DynCO::getDynFunc(ihipModuleSymbol_t**, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>) /data/nhanna/repos/TheRock/rocm-systems/projects/clr/hipamd/src/hip_code_object.cpp:125:22 #8 0x7f5f21f842ba in hip::PlatformState::GetDynFunc(ihipModuleSymbol_t**, ihipModule_t*, char const*) /data/nhanna/repos/TheRock/rocm-systems/projects/clr/hipamd/src/hip_platform.cpp:884:22 #9 0x7f5f21ec2d71 in hip::hipModuleGetFunction(ihipModuleSymbol_t**, ihipModule_t*, char const*) /data/nhanna/repos/TheRock/rocm-systems/projects/clr/hipamd/src/hip_module.cpp:89:47 #10 0x7f5f2212c588 in hipModuleGetFunction /data/nhanna/repos/TheRock/rocm-systems/projects/clr/hipamd/src/hip_table_interface.cpp:1926:10 #11 0x7f5f7478d806 in miopen::HIPOCKernel::HIPOCKernel(miopen::HIPOCProgram, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::vector<unsigned long, std::allocator<unsigned long>>, std::vector<unsigned long, std::allocator<unsigned long>>) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/include/miopen/hipoc_kernel.hpp:225:25 #12 0x7f5f766febb7 in miopen::KernelCache::AddKernel(miopen::Handle const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, miopen::HIPOCProgram*) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/kernel_cache.cpp:161:18 #13 0x7f5f76b6f0e4 in miopen::Handle::AddKernel(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/hip/handlehip.cpp:450:34 #14 0x7f5f7411b52f in miopen::checkNumericsImpl(miopen::Handle const&, int, miopen::TensorDescriptor const&, void const*, bool) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/check_numerics.cpp:107:12 #15 0x55e87c72ebee in void testDumpWithNan<float>(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:130:8 #16 0x55e87c72d4e8 in CPU_Dump_NAN_FP32_testDump_Test::TestBody() /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:157:37 #17 0x55e87ef19d5e in void testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2653:27 #18 0x55e87ef19d5e in void testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2689:52 #19 0x55e87ef04cdd in testing::Test::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2728:50 #20 0x55e87ef04cdd in testing::Test::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2718:6 #21 0x55e87ef04e64 in testing::TestInfo::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2874:14 #22 0x55e87ef0500e in testing::TestSuite::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:3052:33 #23 0x55e87ef0500e in testing::TestSuite::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:3006:6 #24 0x55e87ef0d20b in testing::internal::UnitTestImpl::RunAllTests() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:6004:47 #25 0x55e87ef1a1de in bool testing::internal::HandleSehExceptionsInMethodIfSupported<testing::internal::UnitTestImpl, bool>(testing::internal::UnitTestImpl*, bool (testing::internal::UnitTestImpl::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2653:27 #26 0x55e87ef1a1de in bool testing::internal::HandleExceptionsInMethodIfSupported<testing::internal::UnitTestImpl, bool>(testing::internal::UnitTestImpl*, bool (testing::internal::UnitTestImpl::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2689:52 #27 0x55e87ef051b5 in testing::UnitTest::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:5583:55 #28 0x55e87eee3f9b in RUN_ALL_TESTS() /data/nhanna/repos/TheRock/build/third-party/googletest/dist/include/gtest/gtest.h:2334:73 #29 0x55e87eee3f9b in main /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/main_hip.cpp:34:12 #30 0x7f5f20a587e4 in __libc_start_main (/lib64/libc.so.6+0x3a7e4) (BuildId: 889235a2805b8308b2d0274921bbe1890e9a1986) #31 0x55e87b0bcf2d in _start (/data/nhanna/repos/TheRock/build/ml-libs/MIOpen/build/bin/miopen_gtest+0x126bf2d) 0x7e0f08c50200 is located 0 bytes inside of 26088-byte region [0x7e0f08c50200,0x7e0f08c567e8) freed by thread T0 here: #0 0x7f5f8d7b8ba2 in operator delete(void*, unsigned long) /data/nhanna/repos/TheRock/compiler/amd-llvm/compiler-rt/lib/asan/asan_new_delete.cpp:190:3 #1 0x7f5f76b7317d in std::__new_allocator<char>::deallocate(char*, unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/new_allocator.h:172:2 #2 0x7f5f76b7317d in std::allocator<char>::deallocate(char*, unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/allocator.h:210:25 #3 0x7f5f76b7317d in std::allocator_traits<std::allocator<char>>::deallocate(std::allocator<char>&, char*, unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/alloc_traits.h:517:13 #4 0x7f5f76b7317d in std::_Vector_base<char, std::allocator<char>>::_M_deallocate(char*, unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:390:4 #5 0x7f5f76b7317d in std::_Vector_base<char, std::allocator<char>>::~_Vector_base() /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:369:2 #6 0x7f5f76b7317d in std::vector<char, std::allocator<char>>::~vector() /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:738:7 #7 0x7f5f76b7317d in miopen::Handle::LoadProgram(std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, bool) const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/hip/handlehip.cpp:633:5 #8 0x7f5f766fda82 in miopen::KernelCache::AddKernel(miopen::Handle const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, miopen::HIPOCProgram*)::'lambda'()::operator()() const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/kernel_cache.cpp:143:30 #9 0x7f5f766fda82 in miopen::KernelCache::AddKernel(miopen::Handle const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, miopen::HIPOCProgram*) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/kernel_cache.cpp:124:26 #10 0x7f5f76b6f0e4 in miopen::Handle::AddKernel(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/hip/handlehip.cpp:450:34 #11 0x7f5f7411b52f in miopen::checkNumericsImpl(miopen::Handle const&, int, miopen::TensorDescriptor const&, void const*, bool) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/check_numerics.cpp:107:12 #12 0x55e87c72ebee in void testDumpWithNan<float>(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:130:8 #13 0x55e87c72d4e8 in CPU_Dump_NAN_FP32_testDump_Test::TestBody() /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:157:37 #14 0x55e87ef19d5e in void testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2653:27 #15 0x55e87ef19d5e in void testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2689:52 previously allocated by thread T0 here: #0 0x7f5f8d7b7f9d in operator new(unsigned long) /data/nhanna/repos/TheRock/compiler/amd-llvm/compiler-rt/lib/asan/asan_new_delete.cpp:109:35 #1 0x7f5f76b720a5 in std::__new_allocator<char>::allocate(unsigned long, void const*) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/new_allocator.h:151:27 #2 0x7f5f76b720a5 in std::allocator<char>::allocate(unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/allocator.h:198:32 #3 0x7f5f76b720a5 in std::allocator_traits<std::allocator<char>>::allocate(std::allocator<char>&, unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/alloc_traits.h:482:20 #4 0x7f5f76b720a5 in std::_Vector_base<char, std::allocator<char>>::_M_allocate(unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:381:20 #5 0x7f5f76b720a5 in std::_Vector_base<char, std::allocator<char>>::_M_create_storage(unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:398:33 #6 0x7f5f76b720a5 in std::_Vector_base<char, std::allocator<char>>::_Vector_base(unsigned long, std::allocator<char> const&) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:335:9 #7 0x7f5f76b720a5 in std::vector<char, std::allocator<char>>::vector(std::vector<char, std::allocator<char>> const&) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:602:9 #8 0x7f5f76b720a5 in miopen::Handle::LoadProgram(std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, bool) const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/hip/handlehip.cpp:623:27 #9 0x7f5f766fda82 in miopen::KernelCache::AddKernel(miopen::Handle const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, miopen::HIPOCProgram*)::'lambda'()::operator()() const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/kernel_cache.cpp:143:30 #10 0x7f5f766fda82 in miopen::KernelCache::AddKernel(miopen::Handle const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, miopen::HIPOCProgram*) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/kernel_cache.cpp:124:26 #11 0x7f5f76b6f0e4 in miopen::Handle::AddKernel(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/hip/handlehip.cpp:450:34 #12 0x7f5f7411b52f in miopen::checkNumericsImpl(miopen::Handle const&, int, miopen::TensorDescriptor const&, void const*, bool) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/check_numerics.cpp:107:12 #13 0x55e87c72ebee in void testDumpWithNan<float>(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:130:8 #14 0x55e87c72d4e8 in CPU_Dump_NAN_FP32_testDump_Test::TestBody() /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:157:37 #15 0x55e87ef19d5e in void testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2653:27 #16 0x55e87ef19d5e in void testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2689:52 SUMMARY: AddressSanitizer: heap-use-after-free /data/nhanna/repos/TheRock/compiler/amd-llvm/amd/comgr/src/comgr.cpp:216:9 in COMGR::setCStr(char*&, llvm::StringRef, unsigned long*) Shadow bytes around the buggy address: 0x7e0f08c4ff80: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa 0x7e0f08c50000: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa 0x7e0f08c50080: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa 0x7e0f08c50100: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa 0x7e0f08c50180: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa =>0x7e0f08c50200:[fd]fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd 0x7e0f08c50280: fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd 0x7e0f08c50300: fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd 0x7e0f08c50380: fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd 0x7e0f08c50400: fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd 0x7e0f08c50480: fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd Shadow byte legend (one shadow byte represents 8 application bytes): Addressable: 00 Partially addressable: 01 02 03 04 05 06 07 Heap left redzone: fa Freed heap region: fd Stack left redzone: f1 Stack mid redzone: f2 Stack right redzone: f3 Stack after return: f5 Stack use after scope: f8 Global redzone: f9 Global init order: f6 Poisoned by user: f7 Container overflow: fc Array cookie: ac Intra object redzone: bb ASan internal: fe Left alloca redzone: ca Right alloca redzone: cb ==3639==ABORTING ``` ## Test Result Test output after change: ``` HSA_XNACK=1 ASAN_OPTIONS=symbolize=1 ./build/ml-libs/MIOpen/build/bin/miopen_gtest --gtest_filter="*CPU_Dump_NAN_FP32*" PRNG seed: 12345678 Note: Google Test filter = *CPU_Dump_NAN_FP32* [==========] Running 1 test from 1 test suite. [----------] Global test environment set-up. [----------] 1 test from CPU_Dump_NAN_FP32 [ RUN ] CPU_Dump_NAN_FP32.testDump [ OK ] CPU_Dump_NAN_FP32.testDump (51 ms) [----------] 1 test from CPU_Dump_NAN_FP32 (51 ms total) [----------] Global test environment tear-down [==========] 1 test from 1 test suite ran. (52 ms total) [ PASSED ] 1 test. ``` ## Cline Analysis ### Test Coverage Analysis: __1. LoadProgram Code Path (std::vector constructor):__ - __Primary Test__: `rocm-libraries/projects/miopen/test/gtest/db_sync.cpp` - __Function__: `BuildKernel()` calls `handle.LoadProgram(program_file, program_args, "")` - __Coverage__: This test extensively exercises the LoadProgram → LoadBinary → HIPOCProgramImpl constructor path - __Scope__: Tests multiple GPU architectures (gfx908, gfx90a, gfx942, gfx1030) with different CU counts - __Frequency__: Runs on thousands of kernel configurations in the database sync tests __2. Solution Binary Serialization (std::vector usage):__ - __Primary Test__: `rocm-libraries/projects/miopen/test/gtest/find_2_conv.cpp` - __Function__: `miopenSaveSolution()` and `miopenLoadSolution()` with `std::vector<char> solution_binary` - __Coverage__: Tests the save/load cycle of solution binaries - __Scope__: Tests all convolution directions (Forward, BackwardData, BackwardWeights) __3. Additional Coverage:__ - __Cache Tests__: `rocm-libraries/projects/miopen/test/gtest/cache.cpp` tests compression/decompression with `std::vector<char>` - __Dropout Tests__: Uses `std::vector<unsigned char>` for reserve space (related pattern) __Test Quality Assessment:__ ✅ __Both constructors are well-tested__: - The `std::vector<char>` constructor is heavily exercised through database sync tests - The `std::vector<uint8_t>` constructor would be tested through any code paths that use uint8_t binary data ✅ __Real-world scenarios covered__: - Database synchronization (production kernel loading) - Solution serialization (runtime binary handling) - Multi-threaded execution (db_sync uses up to 32 threads) ✅ __Comprehensive architecture coverage__: - Tests run on multiple GPU architectures - Different compute unit configurations tested __Confidence Level__: Very High ### Performance Analysis: Regarding the performance impact of this fix, it's actually quite minimal and represents good engineering practice: __Memory Impact:__ - __Additional Memory Usage__: Each HIPOCProgramImpl object now stores a copy of the binary data in its `binary` member variable - __Typical Size__: GPU code objects are usually relatively small (typically a few KB to a few MB depending on kernel complexity) - __Lifetime__: The memory is only held for the lifetime of the HIPOCProgram object, which is typically short-lived during kernel loading __Performance Characteristics:__ - __One-time Copy Cost__: There's a single memory copy operation during construction (std::vector copy or iterator range construction) - __No Runtime Overhead__: Once constructed, there's no additional performance cost during kernel execution - __Memory Safety Benefit__: Eliminates potential crashes and undefined behavior, which far outweighs the small memory cost __Context in MIOpen:__ - This occurs during the kernel loading phase, not during actual ML inference/training - Kernel loading is already an expensive operation involving compilation, module creation, etc. - The additional memory copy is negligible compared to the overall kernel loading time __Trade-off Analysis:__ - __Cost__: Small increase in memory usage during kernel loading - __Benefit__: Eliminates memory safety bugs that could cause crashes or data corruption - __Net Result__: Significantly positive - reliability and correctness are much more valuable than the minimal memory overhead In practice, this fix follows the RAII (Resource Acquisition Is Initialization) principle and ensures proper ownership semantics, which is standard best practice in modern C++. The performance impact should be unnoticeable in real-world usage.

…being used after being freed. (ROCm#5220) ## Motivation A `heap-use-after-free` error was triggered by AddressSanitizer on test `CPU_Dump_NAN_FP32.testDump`. ## Technical Details Root Cause Analysis: The AddressSanitizer error occurred because the HIPOCProgramImpl constructor was not storing the binary data passed to it. When LoadProgram called LoadBinary and created a HIPOCProgram with the returned vector, the temporary vector would go out of scope, but COMGR still needed to access the binary data later, causing a use-after-free. - The fix ensures that the HIPOCProgramImpl object owns the binary data for its entire lifetime - Both constructors now consistently store the binary data in the `binary` member variable (std::vector) - The uint8_t constructor converts the data to char format using iterator range construction - This prevents the use-after-free that occurred when COMGR tried to access freed memory ## Test Plan Test output before change: ``` HSA_XNACK=1 ASAN_OPTIONS=symbolize=1 ./build/ml-libs/MIOpen/build/bin/miopen_gtest --gtest_filter="*CPU_Dump_NAN_FP32*" PRNG seed: 12345678 Note: Google Test filter = *CPU_Dump_NAN_FP32* [==========] Running 1 test from 1 test suite. [----------] Global test environment set-up. [----------] 1 test from CPU_Dump_NAN_FP32 [ RUN ] CPU_Dump_NAN_FP32.testDump ================================================================= ==3639==ERROR: AddressSanitizer: heap-use-after-free on address 0x7e0f08c50200 at pc 0x7f5f8d7a6554 bp 0x7ffcb7c4a730 sp 0x7ffcb7c49ee8 READ of size 26088 at 0x7e0f08c50200 thread T0 #0 0x7f5f8d7a6553 in memcpy /data/nhanna/repos/TheRock/compiler/amd-llvm/compiler-rt/lib/asan/../sanitizer_common/sanitizer_common_interceptors_memintrinsics.inc:117:5 ROCm#1 0x7f5f23d61d78 in COMGR::setCStr(char*&, llvm::StringRef, unsigned long*) /data/nhanna/repos/TheRock/compiler/amd-llvm/amd/comgr/src/comgr.cpp:216:9 ROCm#2 0x7f5f23d61d78 in COMGR::DataObject::setData(llvm::StringRef) /data/nhanna/repos/TheRock/compiler/amd-llvm/amd/comgr/src/comgr.cpp:334:17 ROCm#3 0x7f5f23d61d78 in amd_comgr_set_data /data/nhanna/repos/TheRock/compiler/amd-llvm/amd/comgr/src/comgr.cpp:606:24 ROCm#4 0x7f5f221dc1d3 in amd::Comgr::set_data(amd_comgr_data_s, unsigned long, char const*) /data/nhanna/repos/TheRock/rocm-systems/projects/clr/rocclr/device/comgrctx.hpp:252:12 ROCm#5 0x7f5f221dc1d3 in amd::device::Program::getSymbolsFromCodeObj(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>*, amd_comgr_symbol_type_s) const /data/nhanna/repos/TheRock/rocm-systems/projects/clr/rocclr/device/devprogram.cpp:2061:14 ROCm#6 0x7f5f219e6f7c in hip::DynCO::populateDynGlobalVars() /data/nhanna/repos/TheRock/rocm-systems/projects/clr/hipamd/src/hip_code_object.cpp:216:22 ROCm#7 0x7f5f219e8e6a in hip::DynCO::getDynFunc(ihipModuleSymbol_t**, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>) /data/nhanna/repos/TheRock/rocm-systems/projects/clr/hipamd/src/hip_code_object.cpp:125:22 ROCm#8 0x7f5f21f842ba in hip::PlatformState::GetDynFunc(ihipModuleSymbol_t**, ihipModule_t*, char const*) /data/nhanna/repos/TheRock/rocm-systems/projects/clr/hipamd/src/hip_platform.cpp:884:22 ROCm#9 0x7f5f21ec2d71 in hip::hipModuleGetFunction(ihipModuleSymbol_t**, ihipModule_t*, char const*) /data/nhanna/repos/TheRock/rocm-systems/projects/clr/hipamd/src/hip_module.cpp:89:47 ROCm#10 0x7f5f2212c588 in hipModuleGetFunction /data/nhanna/repos/TheRock/rocm-systems/projects/clr/hipamd/src/hip_table_interface.cpp:1926:10 ROCm#11 0x7f5f7478d806 in miopen::HIPOCKernel::HIPOCKernel(miopen::HIPOCProgram, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::vector<unsigned long, std::allocator<unsigned long>>, std::vector<unsigned long, std::allocator<unsigned long>>) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/include/miopen/hipoc_kernel.hpp:225:25 ROCm#12 0x7f5f766febb7 in miopen::KernelCache::AddKernel(miopen::Handle const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, miopen::HIPOCProgram*) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/kernel_cache.cpp:161:18 ROCm#13 0x7f5f76b6f0e4 in miopen::Handle::AddKernel(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/hip/handlehip.cpp:450:34 ROCm#14 0x7f5f7411b52f in miopen::checkNumericsImpl(miopen::Handle const&, int, miopen::TensorDescriptor const&, void const*, bool) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/check_numerics.cpp:107:12 ROCm#15 0x55e87c72ebee in void testDumpWithNan<float>(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:130:8 ROCm#16 0x55e87c72d4e8 in CPU_Dump_NAN_FP32_testDump_Test::TestBody() /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:157:37 ROCm#17 0x55e87ef19d5e in void testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2653:27 ROCm#18 0x55e87ef19d5e in void testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2689:52 ROCm#19 0x55e87ef04cdd in testing::Test::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2728:50 ROCm#20 0x55e87ef04cdd in testing::Test::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2718:6 ROCm#21 0x55e87ef04e64 in testing::TestInfo::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2874:14 ROCm#22 0x55e87ef0500e in testing::TestSuite::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:3052:33 ROCm#23 0x55e87ef0500e in testing::TestSuite::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:3006:6 ROCm#24 0x55e87ef0d20b in testing::internal::UnitTestImpl::RunAllTests() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:6004:47 ROCm#25 0x55e87ef1a1de in bool testing::internal::HandleSehExceptionsInMethodIfSupported<testing::internal::UnitTestImpl, bool>(testing::internal::UnitTestImpl*, bool (testing::internal::UnitTestImpl::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2653:27 ROCm#26 0x55e87ef1a1de in bool testing::internal::HandleExceptionsInMethodIfSupported<testing::internal::UnitTestImpl, bool>(testing::internal::UnitTestImpl*, bool (testing::internal::UnitTestImpl::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2689:52 ROCm#27 0x55e87ef051b5 in testing::UnitTest::Run() /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:5583:55 ROCm#28 0x55e87eee3f9b in RUN_ALL_TESTS() /data/nhanna/repos/TheRock/build/third-party/googletest/dist/include/gtest/gtest.h:2334:73 ROCm#29 0x55e87eee3f9b in main /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/main_hip.cpp:34:12 ROCm#30 0x7f5f20a587e4 in __libc_start_main (/lib64/libc.so.6+0x3a7e4) (BuildId: 889235a2805b8308b2d0274921bbe1890e9a1986) ROCm#31 0x55e87b0bcf2d in _start (/data/nhanna/repos/TheRock/build/ml-libs/MIOpen/build/bin/miopen_gtest+0x126bf2d) 0x7e0f08c50200 is located 0 bytes inside of 26088-byte region [0x7e0f08c50200,0x7e0f08c567e8) freed by thread T0 here: #0 0x7f5f8d7b8ba2 in operator delete(void*, unsigned long) /data/nhanna/repos/TheRock/compiler/amd-llvm/compiler-rt/lib/asan/asan_new_delete.cpp:190:3 ROCm#1 0x7f5f76b7317d in std::__new_allocator<char>::deallocate(char*, unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/new_allocator.h:172:2 ROCm#2 0x7f5f76b7317d in std::allocator<char>::deallocate(char*, unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/allocator.h:210:25 ROCm#3 0x7f5f76b7317d in std::allocator_traits<std::allocator<char>>::deallocate(std::allocator<char>&, char*, unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/alloc_traits.h:517:13 ROCm#4 0x7f5f76b7317d in std::_Vector_base<char, std::allocator<char>>::_M_deallocate(char*, unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:390:4 ROCm#5 0x7f5f76b7317d in std::_Vector_base<char, std::allocator<char>>::~_Vector_base() /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:369:2 ROCm#6 0x7f5f76b7317d in std::vector<char, std::allocator<char>>::~vector() /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:738:7 ROCm#7 0x7f5f76b7317d in miopen::Handle::LoadProgram(std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, bool) const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/hip/handlehip.cpp:633:5 ROCm#8 0x7f5f766fda82 in miopen::KernelCache::AddKernel(miopen::Handle const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, miopen::HIPOCProgram*)::'lambda'()::operator()() const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/kernel_cache.cpp:143:30 ROCm#9 0x7f5f766fda82 in miopen::KernelCache::AddKernel(miopen::Handle const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, miopen::HIPOCProgram*) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/kernel_cache.cpp:124:26 ROCm#10 0x7f5f76b6f0e4 in miopen::Handle::AddKernel(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/hip/handlehip.cpp:450:34 ROCm#11 0x7f5f7411b52f in miopen::checkNumericsImpl(miopen::Handle const&, int, miopen::TensorDescriptor const&, void const*, bool) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/check_numerics.cpp:107:12 ROCm#12 0x55e87c72ebee in void testDumpWithNan<float>(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:130:8 ROCm#13 0x55e87c72d4e8 in CPU_Dump_NAN_FP32_testDump_Test::TestBody() /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:157:37 ROCm#14 0x55e87ef19d5e in void testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2653:27 ROCm#15 0x55e87ef19d5e in void testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2689:52 previously allocated by thread T0 here: #0 0x7f5f8d7b7f9d in operator new(unsigned long) /data/nhanna/repos/TheRock/compiler/amd-llvm/compiler-rt/lib/asan/asan_new_delete.cpp:109:35 ROCm#1 0x7f5f76b720a5 in std::__new_allocator<char>::allocate(unsigned long, void const*) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/new_allocator.h:151:27 ROCm#2 0x7f5f76b720a5 in std::allocator<char>::allocate(unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/allocator.h:198:32 ROCm#3 0x7f5f76b720a5 in std::allocator_traits<std::allocator<char>>::allocate(std::allocator<char>&, unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/alloc_traits.h:482:20 ROCm#4 0x7f5f76b720a5 in std::_Vector_base<char, std::allocator<char>>::_M_allocate(unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:381:20 ROCm#5 0x7f5f76b720a5 in std::_Vector_base<char, std::allocator<char>>::_M_create_storage(unsigned long) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:398:33 ROCm#6 0x7f5f76b720a5 in std::_Vector_base<char, std::allocator<char>>::_Vector_base(unsigned long, std::allocator<char> const&) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:335:9 ROCm#7 0x7f5f76b720a5 in std::vector<char, std::allocator<char>>::vector(std::vector<char, std::allocator<char>> const&) /opt/rh/gcc-toolset-13/root/usr/lib/gcc/x86_64-redhat-linux/13/../../../../include/c++/13/bits/stl_vector.h:602:9 ROCm#8 0x7f5f76b720a5 in miopen::Handle::LoadProgram(std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, bool) const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/hip/handlehip.cpp:623:27 ROCm#9 0x7f5f766fda82 in miopen::KernelCache::AddKernel(miopen::Handle const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, miopen::HIPOCProgram*)::'lambda'()::operator()() const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/kernel_cache.cpp:143:30 ROCm#10 0x7f5f766fda82 in miopen::KernelCache::AddKernel(miopen::Handle const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, miopen::HIPOCProgram*) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/kernel_cache.cpp:124:26 ROCm#11 0x7f5f76b6f0e4 in miopen::Handle::AddKernel(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::filesystem::__cxx11::path const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::vector<unsigned long, std::allocator<unsigned long>> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&, unsigned long, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) const /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/hip/handlehip.cpp:450:34 ROCm#12 0x7f5f7411b52f in miopen::checkNumericsImpl(miopen::Handle const&, int, miopen::TensorDescriptor const&, void const*, bool) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/src/check_numerics.cpp:107:12 ROCm#13 0x55e87c72ebee in void testDumpWithNan<float>(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>> const&) /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:130:8 ROCm#14 0x55e87c72d4e8 in CPU_Dump_NAN_FP32_testDump_Test::TestBody() /data/nhanna/repos/TheRock/rocm-libraries/projects/miopen/test/gtest/dumpTensorTest.cpp:157:37 ROCm#15 0x55e87ef19d5e in void testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2653:27 ROCm#16 0x55e87ef19d5e in void testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /data/nhanna/repos/TheRock/build/third-party/googletest/source/googletest/src/gtest.cc:2689:52 SUMMARY: AddressSanitizer: heap-use-after-free /data/nhanna/repos/TheRock/compiler/amd-llvm/amd/comgr/src/comgr.cpp:216:9 in COMGR::setCStr(char*&, llvm::StringRef, unsigned long*) Shadow bytes around the buggy address: 0x7e0f08c4ff80: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa 0x7e0f08c50000: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa 0x7e0f08c50080: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa 0x7e0f08c50100: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa 0x7e0f08c50180: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa =>0x7e0f08c50200:[fd]fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd 0x7e0f08c50280: fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd 0x7e0f08c50300: fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd 0x7e0f08c50380: fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd 0x7e0f08c50400: fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd 0x7e0f08c50480: fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd Shadow byte legend (one shadow byte represents 8 application bytes): Addressable: 00 Partially addressable: 01 02 03 04 05 06 07 Heap left redzone: fa Freed heap region: fd Stack left redzone: f1 Stack mid redzone: f2 Stack right redzone: f3 Stack after return: f5 Stack use after scope: f8 Global redzone: f9 Global init order: f6 Poisoned by user: f7 Container overflow: fc Array cookie: ac Intra object redzone: bb ASan internal: fe Left alloca redzone: ca Right alloca redzone: cb ==3639==ABORTING ``` ## Test Result Test output after change: ``` HSA_XNACK=1 ASAN_OPTIONS=symbolize=1 ./build/ml-libs/MIOpen/build/bin/miopen_gtest --gtest_filter="*CPU_Dump_NAN_FP32*" PRNG seed: 12345678 Note: Google Test filter = *CPU_Dump_NAN_FP32* [==========] Running 1 test from 1 test suite. [----------] Global test environment set-up. [----------] 1 test from CPU_Dump_NAN_FP32 [ RUN ] CPU_Dump_NAN_FP32.testDump [ OK ] CPU_Dump_NAN_FP32.testDump (51 ms) [----------] 1 test from CPU_Dump_NAN_FP32 (51 ms total) [----------] Global test environment tear-down [==========] 1 test from 1 test suite ran. (52 ms total) [ PASSED ] 1 test. ``` ## Cline Analysis ### Test Coverage Analysis: __1. LoadProgram Code Path (std::vector constructor):__ - __Primary Test__: `rocm-libraries/projects/miopen/test/gtest/db_sync.cpp` - __Function__: `BuildKernel()` calls `handle.LoadProgram(program_file, program_args, "")` - __Coverage__: This test extensively exercises the LoadProgram → LoadBinary → HIPOCProgramImpl constructor path - __Scope__: Tests multiple GPU architectures (gfx908, gfx90a, gfx942, gfx1030) with different CU counts - __Frequency__: Runs on thousands of kernel configurations in the database sync tests __2. Solution Binary Serialization (std::vector usage):__ - __Primary Test__: `rocm-libraries/projects/miopen/test/gtest/find_2_conv.cpp` - __Function__: `miopenSaveSolution()` and `miopenLoadSolution()` with `std::vector<char> solution_binary` - __Coverage__: Tests the save/load cycle of solution binaries - __Scope__: Tests all convolution directions (Forward, BackwardData, BackwardWeights) __3. Additional Coverage:__ - __Cache Tests__: `rocm-libraries/projects/miopen/test/gtest/cache.cpp` tests compression/decompression with `std::vector<char>` - __Dropout Tests__: Uses `std::vector<unsigned char>` for reserve space (related pattern) __Test Quality Assessment:__ ✅ __Both constructors are well-tested__: - The `std::vector<char>` constructor is heavily exercised through database sync tests - The `std::vector<uint8_t>` constructor would be tested through any code paths that use uint8_t binary data ✅ __Real-world scenarios covered__: - Database synchronization (production kernel loading) - Solution serialization (runtime binary handling) - Multi-threaded execution (db_sync uses up to 32 threads) ✅ __Comprehensive architecture coverage__: - Tests run on multiple GPU architectures - Different compute unit configurations tested __Confidence Level__: Very High ### Performance Analysis: Regarding the performance impact of this fix, it's actually quite minimal and represents good engineering practice: __Memory Impact:__ - __Additional Memory Usage__: Each HIPOCProgramImpl object now stores a copy of the binary data in its `binary` member variable - __Typical Size__: GPU code objects are usually relatively small (typically a few KB to a few MB depending on kernel complexity) - __Lifetime__: The memory is only held for the lifetime of the HIPOCProgram object, which is typically short-lived during kernel loading __Performance Characteristics:__ - __One-time Copy Cost__: There's a single memory copy operation during construction (std::vector copy or iterator range construction) - __No Runtime Overhead__: Once constructed, there's no additional performance cost during kernel execution - __Memory Safety Benefit__: Eliminates potential crashes and undefined behavior, which far outweighs the small memory cost __Context in MIOpen:__ - This occurs during the kernel loading phase, not during actual ML inference/training - Kernel loading is already an expensive operation involving compilation, module creation, etc. - The additional memory copy is negligible compared to the overall kernel loading time __Trade-off Analysis:__ - __Cost__: Small increase in memory usage during kernel loading - __Benefit__: Eliminates memory safety bugs that could cause crashes or data corruption - __Net Result__: Significantly positive - reliability and correctness are much more valuable than the minimal memory overhead In practice, this fix follows the RAII (Resource Acquisition Is Initialization) principle and ensures proper ownership semantics, which is standard best practice in modern C++. The performance impact should be unnoticeable in real-world usage.

* Add fp4 mfma support * Allow using Zeros, Ones and Identity for MX types * Display scales for MX types * Fix non subtileImpl path bug * Fix display issue on MX types (PrintTensor option)

Add --type tf32 CLI option: uses float storage with XFloat32 math-op truncation (10-bit mantissa). The slow reference path already handles f32XdlMathOp == XFloat32 via ReferenceSolution<TypedGemm_S_S_S, float, XFloat32>. Changes: - Add isTF32 parameter to runGemm, sets f32XdlMathOp on contraction - Golden reference uses columnMajorGemm<float, XFloat32> for TF32 - TF32 validation tolerance set to 1.0f (13 mantissa bits lost) - Console output shows MathOp=XFloat32 for TF32 runs - 10 new slow-path tf32 tests (transpose combos, beta, bias, features, scaleAB) Fast path still rejects XFloat32 — fast-path tf32 tests come next. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

#8, stacked on #1, #3] Template solveCPUFastInF32 on MathOpAccumT so the inner reduction applies XFloat32 truncation when MathOpAccumT=XFloat32. Remove the XFloat32 rejection guard from isFastPathEligible. Update SolveGemmCPU dispatch to branch on f32XdlMathOp. When MathOpAccumT=float (default), casts are no-ops with identical codegen to the previous implementation. Changes: - solveCPUFastInF32 gains template<MathOpAccumT=float> - innerReduction uses float(MathOpAccumT(val)) cast chain - isFastPathEligible no longer rejects XFloat32 - SolveGemmCPU dispatches solveCPUFastInF32<XFloat32> for TF32 - 10 new fast-path tf32 tests (20 total tf32 tests now) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Add --type tf32 CLI option: uses float storage with XFloat32 math-op truncation (10-bit mantissa). The slow reference path already handles f32XdlMathOp == XFloat32 via ReferenceSolution<TypedGemm_S_S_S, float, XFloat32>. Changes: - Add isTF32 parameter to runGemm, sets f32XdlMathOp on contraction - Golden reference uses columnMajorGemm<float, XFloat32> for TF32 - TF32 validation tolerance set to 1.0f (13 mantissa bits lost) - Console output shows MathOp=XFloat32 for TF32 runs - 10 new slow-path tf32 tests (transpose combos, beta, bias, features, scaleAB) Fast path still rejects XFloat32 — fast-path tf32 tests come next. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

#8, stacked on #1, #3] Template solveCPUFastInF32 on MathOpAccumT so the inner reduction applies XFloat32 truncation when MathOpAccumT=XFloat32. Remove the XFloat32 rejection guard from isFastPathEligible. Update SolveGemmCPU dispatch to branch on f32XdlMathOp. When MathOpAccumT=float (default), casts are no-ops with identical codegen to the previous implementation. Changes: - solveCPUFastInF32 gains template<MathOpAccumT=float> - innerReduction uses float(MathOpAccumT(val)) cast chain - isFastPathEligible no longer rejects XFloat32 - SolveGemmCPU dispatches solveCPUFastInF32<XFloat32> for TF32 - 10 new fast-path tf32 tests (20 total tf32 tests now) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…ternalproject Link to a locally built OpenBLAS static library

This entire device template (DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1) emits MFMA intrinsics, which exist only on CDNA (gfx9). The shared is_xdl_wmma_supported() helper returns true for FP16/BF16 with 16x16 on gfx11/gfx12 because it is also used by WMMA device templates. As a result the existing 16x16 XDL instances passed IsSupportedArgument on RDNA, were launched, and produced garbage (no MFMA hardware). This was exposed by RUN_ALL_TESTS=ON in build #8, which caused all FP16/BF16 TestGroupedConvndBwdData2d cases to fail on gfx11/gfx1201. FP32 silently "passed" only because is_xdl_wmma_supported<float,...>() returns false. Fix: in IsSupportedArgument, unconditionally reject the XDL template on gfx11/gfx12. The corresponding WMMA path lives in the sibling file device_grouped_conv_bwd_data_multiple_d_wmma_cshuffle.hpp and is unaffected. This replaces the narrower CDEBlockTransferScalarPerVector_NPerBlock==1 guard added previously, which only covered the new NoShuffle instances and did not protect the pre-existing 16x16 XDL instances. Made-with: Cursor

# Add gfx950 MXFP4 Subtile-based kernel implementation ## Summary This PR is a follow-up to #6499 ([hipblaslt] Add support for gfx950 mxfp4) and adds the **Subtile-based kernel implementation (`UseSubtileImpl=1`)** for hipBLASLt on **gfx950**. It introduces a new tile-decomposed code generation path optimized for **MXFP4** and **BF16** GEMMs, plus the solution-selection plumbing, validation, Origami logic yamls, and unit tests needed to make it production-usable. ## Motivation PR #6499 brought MX data type support online for gfx950, but the existing TensileLite codegen path leaves significant performance on the table for MXFP4-heavy workloads. The Subtile path restructures global-read / local-read / MFMA / store scheduling at a finer granularity, which **greatly improves MXFP4 GEMM performance when using `HIPBLASLT_MATMUL_MATRIX_SCALE_BLK32_UE8M0_32_8_EXT`** (added to the hipBLASLt CHANGELOG). ## What's included ### 1. New Subtile-based kernel components (Tensile) New modules under `projects/hipblaslt/tensilelite/Tensile/Components/`: * `SubtileBasedKernel.py` (~1850 LOC) — entry point and orchestration of the subtile codegen path; replaces large portions of the standard prefetch / unroll / store flow when `UseSubtileImpl=1`. * `SubtileBasedLogicalScheduler.py` (~2415 LOC) — logical scheduler that builds the subtile-grained instruction graph (GR loads, LR offsets, MFMA tiles, scale loads, stores) from kernel parameters. * `SubtileBasedInstructionScheduler.py` (~433 LOC) — converts the logical schedule to an emit order respecting wave / register / hazard constraints. * `SubtileBasedInstructionEmitter.py` (~216 LOC) — instruction emission helpers shared by the subtile components. ### 2. Kernel writer / common changes * **`KernelWriter.py`**, **`KernelWriterAssembly.py`**: integration points for the subtile path — prefetch, GR offset calculation, LR offset calculation, post-loop, MFMA macro accounting, optimized `storeD`, LDS buffer swap, MX FP4 scale emit, `SrdMXSA/B+2` handling, sgpr allocation / overflow guards, computeLoadSrd fix. * **`SolutionStructs/Solution.py`**, **`SolutionStructs/Problem.py`**: introduces the `UseSubtileImpl` parameter, MX-related reject conditions for non-Subtile paths on gfx950, and additional valid GEMM type combinations for MX inputs. * **`Common/ValidParameters.py`**, **`Common/RequiredParameters.py`**, **`Common/GlobalParameters.py`**: `UseSubtileImpl` registration and defaults. * **`Components/StreamK.py`**: subtile-aware StreamK fixup (incl. import union with the `BufferLoadB32` cache-coherence change from #6837). * **`Components/GlobalWriteBatch.py`**: optimized global write batching for the subtile path (~670 LOC of changes). * **`Components/ComputeStoreVgprs.py`**, **`Components/LSU.py`**, **`Components/WorkGroupMappingAlgos.py`**, **`AsmStoreState.py`**, **`KernelWriterModules.py`**: minor adjustments needed by the subtile pipeline. ### 3. rocisa / host / client * **`rocisa/rocisa/include/container.hpp`**: helpers needed by the new emitter. * **`tensile_host.cpp`**, **`include/Tensile/TensorDescriptor.hpp`**: small fixups for the subtile path and gfx950 build. * **`client/include/DataInitialization.hpp`**, **`client/src/DataInitialization.cpp`**, **`client/src/Reference.cpp`**, **`client/src/ReferenceValidator.cpp`**, **`client/include/TypedId.hpp`**: MX scale init and reference paths used by the new tests. * **`clients/common/include/testing_matmul.hpp`**, **`clients/common/include/norm.hpp`**, **`clients/common/include/hipblaslt_datatype2string.hpp`**, **`clients/common/src/mxDataGen.cpp`**: wiring for batched (>1) testing and MX init. ### 4. Origami / solution selection (gfx950 MXFP4) New auto-tuned logic yamls under `projects/hipblaslt/library/.../Tensile/Logic/asm_full/gfx950/gfx950/Origami/` covering the FP4 SS / HS / BS variants in three layouts: * `Origami/` (default) * `Origami/Origami_nta4/` (no-transpose-A FP4) * `Origami/Origami_ntb4/` (no-transpose-B FP4) (9 new `gfx950_Cijk_Alik_Bljk_F4{SS,HS,BS}_MXA32_MXB32_*_UserArgs.yaml` files in total.) ### 5. New tests **End-to-end gfx950 GEMM yamls** in `Tensile/Tests/common/gemm/gfx950/`: * `subtile_bf16.yaml`, `subtile_mxfp4.yaml` * `mx32f4_tn.yaml`, `mx32f8_tn.yaml` * `mxfp4_mxfp4_{fp32,bf16}_tn_act{,_groupgemm}.yaml` * `mxfp4_fp8_{fp32,bf16}_tn_act{,_groupgemm}.yaml` * `fp8_mxfp4_{fp32,bf16}_tn_act{,_groupgemm}.yaml` **StreamK + MX:** `Tensile/Tests/common/streamk/sk_mx32f4_quick.yaml`, `sk_mx32f8_quick.yaml`. **New unit tests** (`Tensile/Tests/unit/`): * `test_SubtileBasedLogicalScheduler.py` (~1735 LOC) * `test_SubtileBasedSchedulerRef.py` (~596 LOC) * `test_gr_lr_roundtrip.py` (~571 LOC) * `test_storeD_roundtrip.py` (~2420 LOC) * `test_graTileAssignment.py` (~354 LOC) * `test_lraTileAssignment.py` (~360 LOC) * `conftest.py`, `gpu_test_helpers.py` shared fixtures (~601 LOC) **New gtest:** `tensilelite/tests/MXScalePadding_test.cpp`. ### 6. Misc / hardening * Reject conditions: gfx950 MX + non-Subtile, DepthU constraints, GroupGEMM not yet supported with StreamK + MX, AssertSummationElementMultiple=256 for subtile MXFP4, missing-mxblock check for non-MX types. * Skip rocRoller for FP4-A/FP4-B with pre-swizzled scale layout (#42). * `forceDenorm=False` in `generateMXInput` (#11). * Several rebase fixes, copyright/year header updates, and review-comment fixes to `KernelWriter` / `KernelWriterAssembly`. ### 7. CHANGELOG Greatly improved MXFP4 GEMM performance when using HIPBLASLT_MATMUL_MATRIX_SCALE_BLK32_UE8M0_32_8_EXT ## How to use Set `UseSubtileImpl: 1` on a gfx950 MX-FP4 solution (see the new `subtile_mxfp4.yaml` / `mx32f4_tn.yaml` for canonical configs). The path is opt-in — non-MX and non-gfx950 kernels are unaffected. ## Backwards compatibility / risk * All new behavior is gated on `UseSubtileImpl=1` and gfx950. Existing solutions on other architectures or non-MX paths are unchanged. * `GroupGEMM + StreamK + MX` is intentionally rejected for now (TODO). * New Origami yamls only add solutions; nothing existing is modified. ## Test plan * New gtests + unit tests run automatically in CI (Tensilelite Python unit suite, `MXDataGen_test`, `MXScalePadding_test`). * New end-to-end gfx950 GEMM and StreamK yamls are added to the common test buckets. * Manual: run the gfx950 MXFP4 subtile suites (`pytest -k gfx950` after building Tensile, plus `tensilelite-client --yaml subtile_mxfp4.yaml` for sanity). ## Notes for reviewers * This branch was rebased onto current `develop` (post-#6499) by skipping the `users/nakajee/gfx950_mx_rebase_merge` history (which #6499 squash-merged) and replaying only the subtile-specific work as a single squashed commit. The actual code changes in this PR are limited to the files listed above (24 added, 56 modified; ~+170k / −2.6k including generated logic yamls). * The largest reviewable diffs are: * `Tensile/Components/SubtileBased{Kernel,LogicalScheduler,InstructionScheduler,InstructionEmitter}.py` (new files) * `Tensile/KernelWriter.py`, `Tensile/KernelWriterAssembly.py` * `Tensile/SolutionStructs/{Problem,Solution}.py` * `Tensile/Components/{GlobalWriteBatch,StreamK}.py` * `clients/common/include/testing_matmul.hpp` * `client/src/DataInitialization.cpp` * Description of all commits that were squashed for this feature branch: Subtile implementation for gfx950 MX FP4 --- 272f88d: Add sample subtile impl --- Author: brianshi <brianshi@amd.com> --- 60ecede: GR Offset calculation (#1) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> --- be69c1d: Enable post-loop code generation, and add some subroutines --- Author: b-shi <brianshi@amd.com> --- 646d102: LR offset calculation (#2) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> --- 71f4bca: Add GR load emit logic, and misc fixes (#3) --- Author: b-shi <brianshi@amd.com> --- 1fd0db9: Emit LR + init ACCVGPR (#4) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> --- 9d406b9: Add loop and ptr update code --- Author: b-shi <brianshi@amd.com> --- b6127bc: Update GR/LR offset calculation to fully support 2x2, 1x4, 4x1 waveConfigs (#7) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> --- 89ec87c: Account for valuC macro value in SK WS store code --- Author: b-shi <brianshi@amd.com> --- 6edf53d: Rebase fix --- Author: b-shi <brianshi@amd.com> --- 34e79fc: Enable fp4 (#8) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> --- d5a5c57: [Tensilelite] Add MX FP4 scale offset computation for subtile-based kernel (#6) --- Author: Archana Ramalingam <98564406+archana-ramalingam@users.noreply.github.com> --- 7a8a85a: Add lds buffer swap logic --- Author: b-shi <brianshi@amd.com> --- d24a8fe: Add optimized storeD code (#9) --- Author: b-shi <brianshi@amd.com> --- a45c20c: Fix MX scale tensor initialization: set forceDenorm=false in generateMXInput (#11) --- Author: T.J. Alumbaugh <T.J.Alumbaugh@amd.com> --- f945268: [Tensilelite] Enable the MX FP4 scale emit code in the subtile-based kernel (#10) --- Author: Archana Ramalingam <98564406+archana-ramalingam@users.noreply.github.com> --- cf37df4: Use fixed value for SrdMXSA/B+2 (#14) --- Author: Koji Nakajima <75698246+nakajee@users.noreply.github.com> --- f0c8dbc: Merge subtile_mx_f4_schedule to subtile_mx branch (#16) --- Author: b-shi <brianshi@amd.com> --- 543796f: Enable DU > 256, and reduce sgpr allocation (#18) --- Author: b-shi <brianshi@amd.com> --- c65bdb0: Add missing mxblock check for non-mx data types --- Author: b-shi <brianshi@amd.com> --- d64d226: Introduce UseSubtileImpl parameter (#20) --- Author: b-shi <brianshi@amd.com> Squash commits 20-35 from subtile_mx branch --- e4780da: Enable FixSrd2 for A/B (#23) --- Author: b-shi <brianshi@amd.com> * Enable FixSrd2 for A/B * Address comments from PR --------- --- e4c64a7: Add nt libs --- Author: b-shi <brianshi@amd.com> --- cd13ec1: [Tensilelite] Pad MX scale tensor dimensions for unaligned problem sizes (#21) --- Author: Archana Ramalingam <98564406+archana-ramalingam@users.noreply.github.com> * Add scale padding * Add tests * Remove redundant pre-swizzle path * Remove code from conflict * Fix reverted mxdatagen path for tensile tests * Add diverse test cases for scale padding in MXScalePadding_test and subtile.yaml - Expanded test cases to include non-multiple-of-32, even non-multiple-of-16, and odd dimensions. --- d87938f: Split subtile.yaml into subtile_bf16.yaml and subtile_mxfp4.yaml (#22) --- Author: James Newling <james.newling@gmail.com> Replace the 'monolithic' subtile.yaml with two focused test files. All original test coverage is preserved. Two new FP4 groups added. BF16 coverage (subtile_bf16.yaml, tests are essentially unchanged): # | Description | Dest | MIs | PGR | DU | SK | Sizes --+--------------------+------+-----+-----+---------+-----+------ 0 | BF16 TN main | b | 19 | 0 | 64 | 0,3 | 11 1 | BF16 TN large DU | b | 4 | 0 | 128,192 | 0,3 | 7 2 | BSS (f32 output) | s | 6 | 0 | 64 | 0,3 | 9 3 | BF16 bias | b | 2 | 0 | 64 | 0 | 1 FP4 coverage (subtile_mxfp4.yaml): # | Description | Dest | MIs | PGR | DU | SK | Sizes | Status --+--------------------+------+-----+-----+-----+-----+-------+-------- 0 | FP4 TN main | b | 15 | 0 | 256 | 0,3 | 23 | from original 1 | FP4 TN large DU | b | 4 | 0 | 512 | 0,3 | 13 | from original 2 | F4SS (f32 output) | s | 5 | 0 | 256 | 0,3 | 13 | from original 3 | FP4 bias | b | 2 | 0 | 256 | 0 | 1 | from original 4 | FP4 PGR=2 | b | 13 | 2 | 256 | 0 | 5 | new 5 | FP4 expanded MIWT | b | 24 | 0 | 256 | 0 | 5 | new 6 | PGR=2 WG 4x1/1x4 | | 6 | 2 | 256 | 0 | 1 | known failures (commented) Run times on gfx950 (8x MI350X): File | NEV=-1 | NEV=0 -------------------+--------+------ subtile_bf16.yaml | 23s | 23s subtile_mxfp4.yaml | 37s | 40s Where NEV is number of elements to validate. I (James) have checked these numbers, and weirdly it is true that NEV=0 is a bit faster than NEV=-1 for mxfp4. --- af04f0d: Dependency based instruction scheduling (#19) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> * Revert to single partition * Start using dependencies * as is * start using separate EmittedModules * remove reduntant wait * Add _extractPathsFromBeforeDeps * Continue simplification * Simplifying * Add more rules * cleanup * Add fp4 test * fix test * Add tests * Remove after field on emittedmodule * Refactoring instructionSchedule * Add comments * cleanup modules vs ops * Refactoring print functions * Test cleanup * Add more tests * Replace subgroup by partition * Remove unused unroll param * Add high level notes * Simplify NLL and NGLL GR removal * Add some comments * Force instruction insertion if no slots available * Fix test after rebase * Move scale before A/B and track inflight count * Fine-grain vmcnt calculation * Separate counts for scaleA and B * Avoid using m0 update and buffer_lod on same MFMA slot to avoid scalar instruction serialization * Fix test * Add vmcnt test * Fix duplicated loads for 1x4 and 4x1 * Fix placement in reverse order * Fix regression on PGR0 * add fallback to numMFMA=1 --- 3ec902b: Add some 1x4 and 4x1 origami solutions --- Author: b-shi <brianshi@amd.com> --- c5000d3: Fix typo --- Author: b-shi <brianshi@amd.com> --- 226ed84: [hipblaslt] Refactor Srd2 calculation for useFixedSrd2 (#30) --- Author: Koji Nakajima <75698246+nakajee@users.noreply.github.com> --- abf19d4: [Tensilelite] UseSubtileImpl: subtile-aligned edge check for store path (#29) --- Author: b-shi <brianshi@amd.com> * [Tensilelite] UseSubtileImpl: subtile-aligned edge check, OOB guard, and refactoring - Replace Size%MT edge check with subtile-aligned check: NonEdge paired store when trailing rows/cols are a multiple of the subtile block size (waveGroupM rows for M, 16 cols for N). Non-last workgroups always take NonEdge. - Add per-wave OOB guard (subtileM32ValidBlocksSgpr / subtileN16ValidBlocksSgpr) to skip stores outside valid M/N tile bounds in the NonEdge path. - Refactor duplicated OOB guard into _emitSubtileOobGuard helper; refactor M/N guard SGPR computation into _emitSubtileMGuard / _emitSubtileNGuard. - Fix orphan scalar store blockIdxM (was tt0, now (tt0*MatrixInstM)//mBlockSize). - Add quick-exit and edge/non-edge header comments to generated ASM. * Add some bias tests, combine M/N guard to single routine * Add OOB check for C loads, update storeD unit tests to check OOB, simplify quick exit checks * Address more PR comments: add M group skip, and skip to store end. simplified loadC OOB mask --------- --- 637881a: Fix unit tests & remove legacy code for subtile interleaving (#33) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> * Fix gr_lr_roundtrip test * Use non-interleaved version as ref code * Fix scheduler test * Removed legacy interleaved mode for LR/GR offset calculation --- e9cb889: Fix MX FP4 scale buffer allocation and initialization for batched GEMM (#25) --- Author: Archana Ramalingam <98564406+archana-ramalingam@users.noreply.github.com> * Fix bacth count issue * Add batch count tests * Fix bacth count issue * Address PR review: clarify FP4-specific byte stride and add non-aligned batched tests - Updated comments on dataBatchBytes computation to clarify FP4 packing assumption (2 elements/byte) and flag that non-FP4 block-scaling types would require updating this conversion. - Added batched test cases with non-multiple-of-32 M/N dimensions: FP4 DU=256: [48,48,2] and [33,65,2] FP4 DU=512: [63,63,2] BF16: [50,100,2] --------- --- a43247b: Update some test yamls (#31) --- Author: b-shi <brianshi@amd.com> --- e2f69c8: Add f4bs origami library with activation function support. Refactor sgpr allocation to reduce sgpr usage in post loop. Store code-path reorganization (#32) --- Author: b-shi <brianshi@amd.com> * Free swap/localwritebase sgprs before post-loop * Defer sgpr allocation to remove holds in sgpr pool. Add Origami library logic files for Cijk_Alik_Bljk_F4BS_MXA32_MXB32 (base, nta4, ntb4 variants). * Remove uneeded alignment and comment * Add more epilogue tests * Remove older origami library for f4bs * Reorder post-loop code blocks to after persistant loop Misc fixes * Fix build issues, relax longjump sgpr requirements * Fix GSU0 branch logic --------- --- 3f034bf: Add F4HS and F4SS Origami library logic for FP4→F16 and FP4→F32 GEMM (#35) --- Author: Majedul Sujon <85503863+msujon-AMD@users.noreply.github.com> * Add F4HS and F4SS Origami library logic for FP4→F16 and FP4→F32 GEMM - Add 6 new yaml files (F4HS, F4SS) across Origami, Origami_nta4, Origami_ntb4 - Update F4BS yaml files: AssertSummationElementMultiple 32→256 for K%256 enforcement - Add ("F4", "F4", "H", "S") to _validGEMMTypes and _HPATypes in Problem.py * Add F4HS test cases to subtile_mxfp4.yaml Add two new benchmark problem blocks for FP4→F16 (F4HS): - No-bias block: same wavetile and problem size coverage as F4SS - Bias epilogue block: BiasDataTypeList [s, h], relu/none activations * Add F4HS (FP4->Half) type support to Tensile client Add TypedGemm_F4_H_S typedef and corresponding reference CPU solver case so F4HS (FP4 input, Float16 output, Float compute) problems can be validated by the benchmark client. --------- --- d0bc8fd: Rewrite subtile-based scheduler. Fix DU>64 & enable very large MT (#36) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> * Initial support for DU>256 * Renaming * add option to do DU=512 in the tests * blocked K-major for scale * Change scaleSet swap logic * Update print functions * Put scales after values for avoid race conditions * Fix tests * more test * tweak printschedule display * Add PGR2 in the yaml tests * Add new scaleGROp * comment out failing tests * Revert "comment out failing tests" This reverts commit 1f5802c. * Draft new logical scheduler * Refactoring * Add more test on step1 * Add more tests on step1 * add bf16 320x320 test * reduce step1 code * Simplify step1 logic * validate some step1 test * Fix partition 2x2 test * more step1 test * 320x320 BF16 test * Add test DU512 + partition2x2 * Simplify step1 code * Add step2 tests * Fix multi-partition step2 * Add step2 du512, 2x2 partition test * Use common algo for all numPartitions * Draft for step3 tests * remove useless tests * New GR algo (draft) * [Step3] Add more test * Iteration on GR * Display ordered GR list with granularities * More test * Add some comments * Disable by default debug logs * Getting rid of step naming * Start remove AnnotatedOp (still there in group pass) * Split dependency Ops * Add todo on place_GRs pass * Valid test_annotate_deps_1x1_partition_DU256 * Test output looking better (still WIP) * single dep for LR tooo * Add remove_cross_deps pass * Fix bugs in dependency pass * insert_gr_lr_inc pass * Add group_lr_gr pass * Add emit pass * Quick port of instruction Emit code * Move emit function to separate file * Refactoring instructionEmitter * Port vgprTile tracking * Reworking second pass (WIP) * Display unrolling requirement * Unrolling check on 2nd pass * Generic validation for assign_vgpr pass * Fix unroll * Add inst schedule in standalone mode * Use lrGran for vgprTile size calculation * Fix bug in emit pass (missing depencency) * PreMFMA path + non-duplication scale load * missing globalReadLDSBufferSwap for GR_INC scales * add wairlr_sync on all LR->GR dep * add waitgr_sync op * remove_unnecessary_gr_deps * Change LR dispatch algo a bit to avoid too many waitgr_sync * Avoid duplicated loads in emitter * Fix bug on gr_emit code * GrInc pass. fix duplicated insertion for B * Fix missing LR_inc for SA/SB * preloop, NLL, NGLL * Simplify preloop * minor changes * Move unroll logic to scheduler * minor changes * Fix unroll id bug on NLL / NGLL * Disable post GRINC for now * Remove commented code * Handle 1x4, 4x1 gr read gran * Fix vmcnt computation * Use correct grCount mapping * Revert in emit logic on buffer_load for PGR0 needs * Add bf16 version in standalone test * Fix LR_Inc insertion on DU>64 * Add subIterK/Partition comment to codegen * Fix issue in GrInc placement * Remove last_mt * Fix LR MT index bug with muli-partition * Disable early LDS size check when subtileImpl is on * Add pass to remove redundant LR deps + fixed issue on dependency annotation pass * Remove more LR redundant deps * Only insert wait_lr_sync on deps * Simple algo to select partition config * Remove HC value for partitions... * Take into account all inflight GR (all tensors) * Fix tests and regressions on gr counts * Fix grCount merge calculation * Better display of dependencies * Add remove_wait_lr_sync after grouping * Add temporary non reg file * Change merge logic on GR grouping pass * Fix non necessary wait_lr_sync * Downgrade some waitlr_sync to sync + added 384x256 no reg test * non reg test 320x320 * Add larger MT * non reg test for fp4 256x256 * Moving out instructionScheduler * Remove old scheduler * Renaming scheduler * Re-work test * Add larger MT test cases * Rename non-ref test * Re-add standalone mode * Refactor DepOp * Remove dead code * Remove MFMATileSize class * Remove from_til_info * Avoid redundant tensor list creation * Remove hardcode granularities in vgrpTile allocation pass. Simplify code. * Re-enable # PGR=2 WG 4x1/1x4, K > DU tests * Remove unused GRScaleOp * DepRef renaming * Get rid of MT string representation * Remove TODO * EmmitedModule simplication * Use explicit pass dependencies * Renaming LogicalScheduler * Remove old test_InterleavingScheduler.py file * Commenting failing test for now * Remove debug logs * Disable lds padding when using UseSubtileImpl --- e8e8c09: Fix LR-GR dependency issue when DU>64 (#40) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> * Fix and simplify logic for remove_unnecessary_lr_deps * Add new ref tests for 128x128x(128,64) --- 4aa441a: Rebase fix --- Author: b-shi <brianshi@amd.com> --- 5ba911e: Skip rocRoller for FP4-A/FP4-B + pre-swizzled scale layout (#42) --- Author: Archana Ramalingam <98564406+archana-ramalingam@users.noreply.github.com> --- 9c74998: Rebase fix --- Author: b-shi <brianshi@amd.com> --- 842b149: Addressed review comments for KernelWriter and KernelWriterAssembly --- Author: Koji Nakajima <knakajim@amd.com> --- dce43b1: Fix computeLoadSrd issue --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- c075bbf: Fix preSolution CPU re-sync regressing subtile_mxfp4.yaml --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- ced840f: Fix computeLoadSrd issue (#43) --- Author: bnemanich <brad.nemanich@amd.com> --- bc2f6dd: Small update for gfx950 mx tests + more - enable UseSubtileImpl for all gfx950 non subtile mx tests - skip all gfx950 mxfp8 - use MXScaleFormat=1 as default - set AssertSummationElementMultiple=256 for subtile mxfp4 - fix isSwizzledSubtile in computeLoadSrd --- Author: Koji Nakajima <knakajim@amd.com> --- 5c794b7: Fix gsuasb.yaml failures --- Author: b-shi <brianshi@amd.com> --- 727f8db: tensilelite: add solution reject conditions for UseSubtileImpl=1 (#38) --- Author: Majedul Sujon <85503863+msujon-AMD@users.noreply.github.com> --- 8928fbb: Add more reject conditions for Subtile --- Author: Koji Nakajima <knakajim@amd.com> --- 6e63ab6: Fix kringshift test failures --- Author: b-shi <brianshi@amd.com> --- b3e9724: Update reject condtion for DepthU in subtile case. Plus, update DepthU setting for gfx950 mx test cases --- Author: Koji Nakajima <knakajim@amd.com> --- 5ab6009: Fix build errors --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- 4a4edca: Update more mxfp4 tensilelite test cases --- Author: Koji Nakajima <knakajim@amd.com> --- bbbc553: Update change log --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- 6476c04: Add more reject conditions for gfx950 subtile --- Author: Koji Nakajima <knakajim@amd.com> --- c5828c4: Updated gfx950 mxfp4 test cases - add StreamK setting - skip groupgemm tests for now (groupgemm does not support streamK) --- Author: Koji Nakajima <knakajim@amd.com> --- f1fc2f1: Fix hipblaslt build error of gfx950 --- Author: Koji Nakajima <knakajim@amd.com> --- 70cea1b: Updated subtile_mxfp4.yaml (add StreamK) --- Author: Koji Nakajima <knakajim@amd.com> --- c1c9b2a: Add uninit lsc,lsp, etc.. fields for subtile --- Author: b-shi <brianshi@amd.com> --- c0c1f72: Fixed merge error in testing_matmul.hpp --- Author: Koji Nakajima <knakajim@amd.com> --- 191e0cb: Add missed batch_count >1 changes --- Author: archana-ramalingam <Archana.Ramalingam@amd.com> --- 01c52f8: Addressed PR comments --- Author: Koji Nakajima <knakajim@amd.com> --- 4e89c91: Reduce mxfp4 test time --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- 3dac20f: Prevent overflow for wgmxcc sgpr allocation --- Author: b-shi <brianshi@amd.com> --- 18dec79: Fix error with problem type --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- 9e69ffd: Add a reject conditoin for gfx950 mx + non Subtile --- Author: Koji Nakajima <knakajim@amd.com> --- 0eed3ba: Add more valid GEMM types --- Author: Brad Nemanich <brad.nemanich@amd.com> --- 8b5514e: Fix missing b build error --- Author: archana-ramalingam <Archana.Ramalingam@amd.com> --- f981ff5: Fix 1250 tests --- Author: Brad Nemanich <brad.nemanich@amd.com> --- d1e69d9: Add more FP4 tests --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- e3a688f: Add MXScaleFormat: 1 to all gfx950 mx test yaml --- Author: Koji Nakajima <knakajim@amd.com> --- aaef3f5: Add DataTypeMXSA,B setting in gfx950 mxfp4 logic yaml --- Author: Koji Nakajima <knakajim@amd.com> --- 861ef8e: Add DataTypeMXSA,B setting in gfx950 mxfp4 logic yaml (nta4,ntb4) --- Author: Koji Nakajima <knakajim@amd.com> Co-authored-by: Archana Ramalingam <Archana.Ramalingam@amd.com> Co-authored-by: Brad Nemanich <Brad.Nemanich@amd.com> Co-authored-by: Brian Shi <Brian.Shi@amd.com> Co-authored-by: James Newling <James.Newling@amd.com> Co-authored-by: Koji Nakajima <Koji.Nakajima@amd.com> Co-authored-by: Majedul Sujon <Majed.Sujon@amd.com> Co-authored-by: Sebastien Vince <Sebastien.Vince@amd.com> Co-authored-by: T.J. Alumbaugh <T.J.Alumbaugh@amd.com> ## Submission Checklist - [ ] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests. --------- Co-authored-by: Archana Ramalingam <Archana.Ramalingam@amd.com> Co-authored-by: Brad Nemanich <Brad.Nemanich@amd.com> Co-authored-by: Brian Shi <Brian.Shi@amd.com> Co-authored-by: James Newling <James.Newling@amd.com> Co-authored-by: Koji Nakajima <Koji.Nakajima@amd.com> Co-authored-by: Majedul Sujon <Majed.Sujon@amd.com> Co-authored-by: Sebastien Vince <Sebastien.Vince@amd.com> Co-authored-by: T.J. Alumbaugh <T.J.Alumbaugh@amd.com>

Adds _node_with_pos(node, capture) — combines _node_label (per-category- stream [N] index, plain MFMA omits) with bare '@ idx=M' position render. Single helper for the canonical Failure node reference shape, replacing three different prior styles: - #1, #2: manual `_node_label + " " + "@ idx=N"` concatenation - #4, #5: `category[name] format_position(...)` (rendered the FULL name inside brackets, e.g. `LRA0[LRA0[0]]`, plus a cross-category list suffix that duplicates [N]'s purpose) - #6: `category format_position(...)` (no brackets) All five formatters now route through _node_with_pos. Plain MFMA stays bracket-less per _node_label's MFMA discriminator; PackMFMAs (categories PackA*/PackB*) keep [N] because CMS reschedules them. Waits in #5 stay as bare '@ idx=N' — the surrounding 'SWaitCnt' word already names the kind; rendering the SYNC category as `SYNC[N] @ idx=M` would just duplicate that. Skipped: - #10 SCCConflict: brackets carry rocisa class name (e.g. [SCSelectB32]) not [N] index; semantic conflict, separate audit. - #7 WrongInterleaving / #8 TimingTooClose: use `name` field for Pack identity (MiddlePack_a/_b/_c, CVT0_a/_b); replacing with [N] would lose the a/b/c discriminator. - #13 ConstraintViolation: slated for deletion in bead `pcz`. 3 new pinning tests verifying [N] actually appears when capture is given (one per Failure: #4 LRA0[1], #5 LRA0[1] + plain MFMA bracket-less, #6 GRA[1] in trailing reference). Test count: 564 passed (+3).

…index Routes #7 WrongInterleavingFailure (pack/expected_next/actual_next) and #8 TimingTooCloseFailure (producer/consumer) through _node_with_pos for the canonical 'category[N] @ idx=M' rendering, dropping the per-formatter _idx() closures that read node.name + issued_at directly. Removes two redundant artifacts: 1. The duplicated vmfma_index in the prior message — GraphNode.name is `f"{category}@{vmfma}.{sub}"` so today's output renders as `PackA0@5.2 @ idx=5 has wrong interleaving...` (vmfma printed twice). Now: `PackA0[N] @ idx=5 has wrong interleaving...` (single source). 2. The dual-shape (`issued_at` vs `position`) detection. The structural- side ValidatorInstruction emitter was deleted in `ola.4` phase 2 (Pack rule); both Failures are now graph-side only. _node_with_pos's own getattr(node, 'position', None) or node.issued_at fallback handles any residual ValidatorInstruction caller without per-formatter closures. Tests rewritten: - test_wrong_interleaving_failure_format: drops the MiddlePack_a/b/c name-string assertions (vestigial — name was the OLD GraphNode label, no longer in the message). Pins the exact new rendering with empty capture. - test_timing_too_close_failure_format: same shape; pins exact new output. - 2 new tests added: test_wrong_interleaving_failure_format_with_capture_brackets and test_timing_too_close_failure_format_with_capture_brackets — verify [N] indexing kicks in when capture is real. Suite: 565 passed (was 563, +2 new bracket tests) / 2 skipped / 1 xfailed.

…rows in 8t9 audit DZL_SCOPE_REASSESSMENT.md §1.1 SCC: adds source-level confirmation that CommonInstruction (parent of every SCC-touching SOPx class) has only dst/dst1/srcs fields, with getDstParams()/getSrcParams() bodies shown verbatim from instruction.hpp:382. Rules out the q9j-style "binding gap" hypothesis at the source level, not just by Python dir() probing. ROCISA_WORKAROUNDS_AUDIT.md: adds a "Status update (2026-05-07)" preamble that flags which rows have been resolved or reframed by subsequent investigations: - Row #3 (_OPERAND_RULES, ~600 LoC estimate) → reassessed by q9j reassessment to ~12 lines C++ binding + ~260 LoC validator migration (the C++ API already exists). - Rows #7, #8, #19 (VCC machinery) → REMOVED by uraq; no rocisa-side replacement planned. VCC dataflow tracking is permanently dropped from the validator's scope. - MFMA acc special-casing → RECLASSIFIED to q9j Category A (acc and acc2 are already in the C++ getDstParams/getSrcParams partition). - Category C narrowed from SCC/VCC/m0/acc to SCC/m0 only. The audit's table and prose are preserved as the original artifact; the preamble lets a reader find the current canonical reframing for each affected row without rewriting the historical record.

# Add gfx950 MXFP4 Subtile-based kernel implementation ## Summary This PR is a follow-up to #6499 ([hipblaslt] Add support for gfx950 mxfp4) and adds the **Subtile-based kernel implementation (`UseSubtileImpl=1`)** for hipBLASLt on **gfx950**. It introduces a new tile-decomposed code generation path optimized for **MXFP4** and **BF16** GEMMs, plus the solution-selection plumbing, validation, Origami logic yamls, and unit tests needed to make it production-usable. ## Motivation PR #6499 brought MX data type support online for gfx950, but the existing TensileLite codegen path leaves significant performance on the table for MXFP4-heavy workloads. The Subtile path restructures global-read / local-read / MFMA / store scheduling at a finer granularity, which **greatly improves MXFP4 GEMM performance when using `HIPBLASLT_MATMUL_MATRIX_SCALE_BLK32_UE8M0_32_8_EXT`** (added to the hipBLASLt CHANGELOG). ## What's included ### 1. New Subtile-based kernel components (Tensile) New modules under `projects/hipblaslt/tensilelite/Tensile/Components/`: * `SubtileBasedKernel.py` (~1850 LOC) — entry point and orchestration of the subtile codegen path; replaces large portions of the standard prefetch / unroll / store flow when `UseSubtileImpl=1`. * `SubtileBasedLogicalScheduler.py` (~2415 LOC) — logical scheduler that builds the subtile-grained instruction graph (GR loads, LR offsets, MFMA tiles, scale loads, stores) from kernel parameters. * `SubtileBasedInstructionScheduler.py` (~433 LOC) — converts the logical schedule to an emit order respecting wave / register / hazard constraints. * `SubtileBasedInstructionEmitter.py` (~216 LOC) — instruction emission helpers shared by the subtile components. ### 2. Kernel writer / common changes * **`KernelWriter.py`**, **`KernelWriterAssembly.py`**: integration points for the subtile path — prefetch, GR offset calculation, LR offset calculation, post-loop, MFMA macro accounting, optimized `storeD`, LDS buffer swap, MX FP4 scale emit, `SrdMXSA/B+2` handling, sgpr allocation / overflow guards, computeLoadSrd fix. * **`SolutionStructs/Solution.py`**, **`SolutionStructs/Problem.py`**: introduces the `UseSubtileImpl` parameter, MX-related reject conditions for non-Subtile paths on gfx950, and additional valid GEMM type combinations for MX inputs. * **`Common/ValidParameters.py`**, **`Common/RequiredParameters.py`**, **`Common/GlobalParameters.py`**: `UseSubtileImpl` registration and defaults. * **`Components/StreamK.py`**: subtile-aware StreamK fixup (incl. import union with the `BufferLoadB32` cache-coherence change from #6837). * **`Components/GlobalWriteBatch.py`**: optimized global write batching for the subtile path (~670 LOC of changes). * **`Components/ComputeStoreVgprs.py`**, **`Components/LSU.py`**, **`Components/WorkGroupMappingAlgos.py`**, **`AsmStoreState.py`**, **`KernelWriterModules.py`**: minor adjustments needed by the subtile pipeline. ### 3. rocisa / host / client * **`rocisa/rocisa/include/container.hpp`**: helpers needed by the new emitter. * **`tensile_host.cpp`**, **`include/Tensile/TensorDescriptor.hpp`**: small fixups for the subtile path and gfx950 build. * **`client/include/DataInitialization.hpp`**, **`client/src/DataInitialization.cpp`**, **`client/src/Reference.cpp`**, **`client/src/ReferenceValidator.cpp`**, **`client/include/TypedId.hpp`**: MX scale init and reference paths used by the new tests. * **`clients/common/include/testing_matmul.hpp`**, **`clients/common/include/norm.hpp`**, **`clients/common/include/hipblaslt_datatype2string.hpp`**, **`clients/common/src/mxDataGen.cpp`**: wiring for batched (>1) testing and MX init. ### 4. Origami / solution selection (gfx950 MXFP4) New auto-tuned logic yamls under `projects/hipblaslt/library/.../Tensile/Logic/asm_full/gfx950/gfx950/Origami/` covering the FP4 SS / HS / BS variants in three layouts: * `Origami/` (default) * `Origami/Origami_nta4/` (no-transpose-A FP4) * `Origami/Origami_ntb4/` (no-transpose-B FP4) (9 new `gfx950_Cijk_Alik_Bljk_F4{SS,HS,BS}_MXA32_MXB32_*_UserArgs.yaml` files in total.) ### 5. New tests **End-to-end gfx950 GEMM yamls** in `Tensile/Tests/common/gemm/gfx950/`: * `subtile_bf16.yaml`, `subtile_mxfp4.yaml` * `mx32f4_tn.yaml`, `mx32f8_tn.yaml` * `mxfp4_mxfp4_{fp32,bf16}_tn_act{,_groupgemm}.yaml` * `mxfp4_fp8_{fp32,bf16}_tn_act{,_groupgemm}.yaml` * `fp8_mxfp4_{fp32,bf16}_tn_act{,_groupgemm}.yaml` **StreamK + MX:** `Tensile/Tests/common/streamk/sk_mx32f4_quick.yaml`, `sk_mx32f8_quick.yaml`. **New unit tests** (`Tensile/Tests/unit/`): * `test_SubtileBasedLogicalScheduler.py` (~1735 LOC) * `test_SubtileBasedSchedulerRef.py` (~596 LOC) * `test_gr_lr_roundtrip.py` (~571 LOC) * `test_storeD_roundtrip.py` (~2420 LOC) * `test_graTileAssignment.py` (~354 LOC) * `test_lraTileAssignment.py` (~360 LOC) * `conftest.py`, `gpu_test_helpers.py` shared fixtures (~601 LOC) **New gtest:** `tensilelite/tests/MXScalePadding_test.cpp`. ### 6. Misc / hardening * Reject conditions: gfx950 MX + non-Subtile, DepthU constraints, GroupGEMM not yet supported with StreamK + MX, AssertSummationElementMultiple=256 for subtile MXFP4, missing-mxblock check for non-MX types. * Skip rocRoller for FP4-A/FP4-B with pre-swizzled scale layout (#42). * `forceDenorm=False` in `generateMXInput` (#11). * Several rebase fixes, copyright/year header updates, and review-comment fixes to `KernelWriter` / `KernelWriterAssembly`. ### 7. CHANGELOG Greatly improved MXFP4 GEMM performance when using HIPBLASLT_MATMUL_MATRIX_SCALE_BLK32_UE8M0_32_8_EXT ## How to use Set `UseSubtileImpl: 1` on a gfx950 MX-FP4 solution (see the new `subtile_mxfp4.yaml` / `mx32f4_tn.yaml` for canonical configs). The path is opt-in — non-MX and non-gfx950 kernels are unaffected. ## Backwards compatibility / risk * All new behavior is gated on `UseSubtileImpl=1` and gfx950. Existing solutions on other architectures or non-MX paths are unchanged. * `GroupGEMM + StreamK + MX` is intentionally rejected for now (TODO). * New Origami yamls only add solutions; nothing existing is modified. ## Test plan * New gtests + unit tests run automatically in CI (Tensilelite Python unit suite, `MXDataGen_test`, `MXScalePadding_test`). * New end-to-end gfx950 GEMM and StreamK yamls are added to the common test buckets. * Manual: run the gfx950 MXFP4 subtile suites (`pytest -k gfx950` after building Tensile, plus `tensilelite-client --yaml subtile_mxfp4.yaml` for sanity). ## Notes for reviewers * This branch was rebased onto current `develop` (post-#6499) by skipping the `users/nakajee/gfx950_mx_rebase_merge` history (which #6499 squash-merged) and replaying only the subtile-specific work as a single squashed commit. The actual code changes in this PR are limited to the files listed above (24 added, 56 modified; ~+170k / −2.6k including generated logic yamls). * The largest reviewable diffs are: * `Tensile/Components/SubtileBased{Kernel,LogicalScheduler,InstructionScheduler,InstructionEmitter}.py` (new files) * `Tensile/KernelWriter.py`, `Tensile/KernelWriterAssembly.py` * `Tensile/SolutionStructs/{Problem,Solution}.py` * `Tensile/Components/{GlobalWriteBatch,StreamK}.py` * `clients/common/include/testing_matmul.hpp` * `client/src/DataInitialization.cpp` * Description of all commits that were squashed for this feature branch: Subtile implementation for gfx950 MX FP4 --- 272f88d: Add sample subtile impl --- Author: brianshi <brianshi@amd.com> --- 60ecede: GR Offset calculation (#1) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> --- be69c1d: Enable post-loop code generation, and add some subroutines --- Author: b-shi <brianshi@amd.com> --- 646d102: LR offset calculation (#2) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> --- 71f4bca: Add GR load emit logic, and misc fixes (#3) --- Author: b-shi <brianshi@amd.com> --- 1fd0db9: Emit LR + init ACCVGPR (#4) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> --- 9d406b9: Add loop and ptr update code --- Author: b-shi <brianshi@amd.com> --- b6127bc: Update GR/LR offset calculation to fully support 2x2, 1x4, 4x1 waveConfigs (#7) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> --- 89ec87c: Account for valuC macro value in SK WS store code --- Author: b-shi <brianshi@amd.com> --- 6edf53d: Rebase fix --- Author: b-shi <brianshi@amd.com> --- 34e79fc: Enable fp4 (#8) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> --- d5a5c57: [Tensilelite] Add MX FP4 scale offset computation for subtile-based kernel (#6) --- Author: Archana Ramalingam <98564406+archana-ramalingam@users.noreply.github.com> --- 7a8a85a: Add lds buffer swap logic --- Author: b-shi <brianshi@amd.com> --- d24a8fe: Add optimized storeD code (#9) --- Author: b-shi <brianshi@amd.com> --- a45c20c: Fix MX scale tensor initialization: set forceDenorm=false in generateMXInput (#11) --- Author: T.J. Alumbaugh <T.J.Alumbaugh@amd.com> --- f945268: [Tensilelite] Enable the MX FP4 scale emit code in the subtile-based kernel (#10) --- Author: Archana Ramalingam <98564406+archana-ramalingam@users.noreply.github.com> --- cf37df4: Use fixed value for SrdMXSA/B+2 (#14) --- Author: Koji Nakajima <75698246+nakajee@users.noreply.github.com> --- f0c8dbc: Merge subtile_mx_f4_schedule to subtile_mx branch (#16) --- Author: b-shi <brianshi@amd.com> --- 543796f: Enable DU > 256, and reduce sgpr allocation (#18) --- Author: b-shi <brianshi@amd.com> --- c65bdb0: Add missing mxblock check for non-mx data types --- Author: b-shi <brianshi@amd.com> --- d64d226: Introduce UseSubtileImpl parameter (#20) --- Author: b-shi <brianshi@amd.com> Squash commits 20-35 from subtile_mx branch --- e4780da: Enable FixSrd2 for A/B (#23) --- Author: b-shi <brianshi@amd.com> * Enable FixSrd2 for A/B * Address comments from PR --------- --- e4c64a7: Add nt libs --- Author: b-shi <brianshi@amd.com> --- cd13ec1: [Tensilelite] Pad MX scale tensor dimensions for unaligned problem sizes (#21) --- Author: Archana Ramalingam <98564406+archana-ramalingam@users.noreply.github.com> * Add scale padding * Add tests * Remove redundant pre-swizzle path * Remove code from conflict * Fix reverted mxdatagen path for tensile tests * Add diverse test cases for scale padding in MXScalePadding_test and subtile.yaml - Expanded test cases to include non-multiple-of-32, even non-multiple-of-16, and odd dimensions. --- d87938f: Split subtile.yaml into subtile_bf16.yaml and subtile_mxfp4.yaml (#22) --- Author: James Newling <james.newling@gmail.com> Replace the 'monolithic' subtile.yaml with two focused test files. All original test coverage is preserved. Two new FP4 groups added. BF16 coverage (subtile_bf16.yaml, tests are essentially unchanged): # | Description | Dest | MIs | PGR | DU | SK | Sizes --+--------------------+------+-----+-----+---------+-----+------ 0 | BF16 TN main | b | 19 | 0 | 64 | 0,3 | 11 1 | BF16 TN large DU | b | 4 | 0 | 128,192 | 0,3 | 7 2 | BSS (f32 output) | s | 6 | 0 | 64 | 0,3 | 9 3 | BF16 bias | b | 2 | 0 | 64 | 0 | 1 FP4 coverage (subtile_mxfp4.yaml): # | Description | Dest | MIs | PGR | DU | SK | Sizes | Status --+--------------------+------+-----+-----+-----+-----+-------+-------- 0 | FP4 TN main | b | 15 | 0 | 256 | 0,3 | 23 | from original 1 | FP4 TN large DU | b | 4 | 0 | 512 | 0,3 | 13 | from original 2 | F4SS (f32 output) | s | 5 | 0 | 256 | 0,3 | 13 | from original 3 | FP4 bias | b | 2 | 0 | 256 | 0 | 1 | from original 4 | FP4 PGR=2 | b | 13 | 2 | 256 | 0 | 5 | new 5 | FP4 expanded MIWT | b | 24 | 0 | 256 | 0 | 5 | new 6 | PGR=2 WG 4x1/1x4 | | 6 | 2 | 256 | 0 | 1 | known failures (commented) Run times on gfx950 (8x MI350X): File | NEV=-1 | NEV=0 -------------------+--------+------ subtile_bf16.yaml | 23s | 23s subtile_mxfp4.yaml | 37s | 40s Where NEV is number of elements to validate. I (James) have checked these numbers, and weirdly it is true that NEV=0 is a bit faster than NEV=-1 for mxfp4. --- af04f0d: Dependency based instruction scheduling (#19) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> * Revert to single partition * Start using dependencies * as is * start using separate EmittedModules * remove reduntant wait * Add _extractPathsFromBeforeDeps * Continue simplification * Simplifying * Add more rules * cleanup * Add fp4 test * fix test * Add tests * Remove after field on emittedmodule * Refactoring instructionSchedule * Add comments * cleanup modules vs ops * Refactoring print functions * Test cleanup * Add more tests * Replace subgroup by partition * Remove unused unroll param * Add high level notes * Simplify NLL and NGLL GR removal * Add some comments * Force instruction insertion if no slots available * Fix test after rebase * Move scale before A/B and track inflight count * Fine-grain vmcnt calculation * Separate counts for scaleA and B * Avoid using m0 update and buffer_lod on same MFMA slot to avoid scalar instruction serialization * Fix test * Add vmcnt test * Fix duplicated loads for 1x4 and 4x1 * Fix placement in reverse order * Fix regression on PGR0 * add fallback to numMFMA=1 --- 3ec902b: Add some 1x4 and 4x1 origami solutions --- Author: b-shi <brianshi@amd.com> --- c5000d3: Fix typo --- Author: b-shi <brianshi@amd.com> --- 226ed84: [hipblaslt] Refactor Srd2 calculation for useFixedSrd2 (#30) --- Author: Koji Nakajima <75698246+nakajee@users.noreply.github.com> --- abf19d4: [Tensilelite] UseSubtileImpl: subtile-aligned edge check for store path (#29) --- Author: b-shi <brianshi@amd.com> * [Tensilelite] UseSubtileImpl: subtile-aligned edge check, OOB guard, and refactoring - Replace Size%MT edge check with subtile-aligned check: NonEdge paired store when trailing rows/cols are a multiple of the subtile block size (waveGroupM rows for M, 16 cols for N). Non-last workgroups always take NonEdge. - Add per-wave OOB guard (subtileM32ValidBlocksSgpr / subtileN16ValidBlocksSgpr) to skip stores outside valid M/N tile bounds in the NonEdge path. - Refactor duplicated OOB guard into _emitSubtileOobGuard helper; refactor M/N guard SGPR computation into _emitSubtileMGuard / _emitSubtileNGuard. - Fix orphan scalar store blockIdxM (was tt0, now (tt0*MatrixInstM)//mBlockSize). - Add quick-exit and edge/non-edge header comments to generated ASM. * Add some bias tests, combine M/N guard to single routine * Add OOB check for C loads, update storeD unit tests to check OOB, simplify quick exit checks * Address more PR comments: add M group skip, and skip to store end. simplified loadC OOB mask --------- --- 637881a: Fix unit tests & remove legacy code for subtile interleaving (#33) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> * Fix gr_lr_roundtrip test * Use non-interleaved version as ref code * Fix scheduler test * Removed legacy interleaved mode for LR/GR offset calculation --- e9cb889: Fix MX FP4 scale buffer allocation and initialization for batched GEMM (#25) --- Author: Archana Ramalingam <98564406+archana-ramalingam@users.noreply.github.com> * Fix bacth count issue * Add batch count tests * Fix bacth count issue * Address PR review: clarify FP4-specific byte stride and add non-aligned batched tests - Updated comments on dataBatchBytes computation to clarify FP4 packing assumption (2 elements/byte) and flag that non-FP4 block-scaling types would require updating this conversion. - Added batched test cases with non-multiple-of-32 M/N dimensions: FP4 DU=256: [48,48,2] and [33,65,2] FP4 DU=512: [63,63,2] BF16: [50,100,2] --------- --- a43247b: Update some test yamls (#31) --- Author: b-shi <brianshi@amd.com> --- e2f69c8: Add f4bs origami library with activation function support. Refactor sgpr allocation to reduce sgpr usage in post loop. Store code-path reorganization (#32) --- Author: b-shi <brianshi@amd.com> * Free swap/localwritebase sgprs before post-loop * Defer sgpr allocation to remove holds in sgpr pool. Add Origami library logic files for Cijk_Alik_Bljk_F4BS_MXA32_MXB32 (base, nta4, ntb4 variants). * Remove uneeded alignment and comment * Add more epilogue tests * Remove older origami library for f4bs * Reorder post-loop code blocks to after persistant loop Misc fixes * Fix build issues, relax longjump sgpr requirements * Fix GSU0 branch logic --------- --- 3f034bf: Add F4HS and F4SS Origami library logic for FP4→F16 and FP4→F32 GEMM (#35) --- Author: Majedul Sujon <85503863+msujon-AMD@users.noreply.github.com> * Add F4HS and F4SS Origami library logic for FP4→F16 and FP4→F32 GEMM - Add 6 new yaml files (F4HS, F4SS) across Origami, Origami_nta4, Origami_ntb4 - Update F4BS yaml files: AssertSummationElementMultiple 32→256 for K%256 enforcement - Add ("F4", "F4", "H", "S") to _validGEMMTypes and _HPATypes in Problem.py * Add F4HS test cases to subtile_mxfp4.yaml Add two new benchmark problem blocks for FP4→F16 (F4HS): - No-bias block: same wavetile and problem size coverage as F4SS - Bias epilogue block: BiasDataTypeList [s, h], relu/none activations * Add F4HS (FP4->Half) type support to Tensile client Add TypedGemm_F4_H_S typedef and corresponding reference CPU solver case so F4HS (FP4 input, Float16 output, Float compute) problems can be validated by the benchmark client. --------- --- d0bc8fd: Rewrite subtile-based scheduler. Fix DU>64 & enable very large MT (#36) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> * Initial support for DU>256 * Renaming * add option to do DU=512 in the tests * blocked K-major for scale * Change scaleSet swap logic * Update print functions * Put scales after values for avoid race conditions * Fix tests * more test * tweak printschedule display * Add PGR2 in the yaml tests * Add new scaleGROp * comment out failing tests * Revert "comment out failing tests" This reverts commit 1f5802c. * Draft new logical scheduler * Refactoring * Add more test on step1 * Add more tests on step1 * add bf16 320x320 test * reduce step1 code * Simplify step1 logic * validate some step1 test * Fix partition 2x2 test * more step1 test * 320x320 BF16 test * Add test DU512 + partition2x2 * Simplify step1 code * Add step2 tests * Fix multi-partition step2 * Add step2 du512, 2x2 partition test * Use common algo for all numPartitions * Draft for step3 tests * remove useless tests * New GR algo (draft) * [Step3] Add more test * Iteration on GR * Display ordered GR list with granularities * More test * Add some comments * Disable by default debug logs * Getting rid of step naming * Start remove AnnotatedOp (still there in group pass) * Split dependency Ops * Add todo on place_GRs pass * Valid test_annotate_deps_1x1_partition_DU256 * Test output looking better (still WIP) * single dep for LR tooo * Add remove_cross_deps pass * Fix bugs in dependency pass * insert_gr_lr_inc pass * Add group_lr_gr pass * Add emit pass * Quick port of instruction Emit code * Move emit function to separate file * Refactoring instructionEmitter * Port vgprTile tracking * Reworking second pass (WIP) * Display unrolling requirement * Unrolling check on 2nd pass * Generic validation for assign_vgpr pass * Fix unroll * Add inst schedule in standalone mode * Use lrGran for vgprTile size calculation * Fix bug in emit pass (missing depencency) * PreMFMA path + non-duplication scale load * missing globalReadLDSBufferSwap for GR_INC scales * add wairlr_sync on all LR->GR dep * add waitgr_sync op * remove_unnecessary_gr_deps * Change LR dispatch algo a bit to avoid too many waitgr_sync * Avoid duplicated loads in emitter * Fix bug on gr_emit code * GrInc pass. fix duplicated insertion for B * Fix missing LR_inc for SA/SB * preloop, NLL, NGLL * Simplify preloop * minor changes * Move unroll logic to scheduler * minor changes * Fix unroll id bug on NLL / NGLL * Disable post GRINC for now * Remove commented code * Handle 1x4, 4x1 gr read gran * Fix vmcnt computation * Use correct grCount mapping * Revert in emit logic on buffer_load for PGR0 needs * Add bf16 version in standalone test * Fix LR_Inc insertion on DU>64 * Add subIterK/Partition comment to codegen * Fix issue in GrInc placement * Remove last_mt * Fix LR MT index bug with muli-partition * Disable early LDS size check when subtileImpl is on * Add pass to remove redundant LR deps + fixed issue on dependency annotation pass * Remove more LR redundant deps * Only insert wait_lr_sync on deps * Simple algo to select partition config * Remove HC value for partitions... * Take into account all inflight GR (all tensors) * Fix tests and regressions on gr counts * Fix grCount merge calculation * Better display of dependencies * Add remove_wait_lr_sync after grouping * Add temporary non reg file * Change merge logic on GR grouping pass * Fix non necessary wait_lr_sync * Downgrade some waitlr_sync to sync + added 384x256 no reg test * non reg test 320x320 * Add larger MT * non reg test for fp4 256x256 * Moving out instructionScheduler * Remove old scheduler * Renaming scheduler * Re-work test * Add larger MT test cases * Rename non-ref test * Re-add standalone mode * Refactor DepOp * Remove dead code * Remove MFMATileSize class * Remove from_til_info * Avoid redundant tensor list creation * Remove hardcode granularities in vgrpTile allocation pass. Simplify code. * Re-enable # PGR=2 WG 4x1/1x4, K > DU tests * Remove unused GRScaleOp * DepRef renaming * Get rid of MT string representation * Remove TODO * EmmitedModule simplication * Use explicit pass dependencies * Renaming LogicalScheduler * Remove old test_InterleavingScheduler.py file * Commenting failing test for now * Remove debug logs * Disable lds padding when using UseSubtileImpl --- e8e8c09: Fix LR-GR dependency issue when DU>64 (#40) --- Author: sebvince <115461989+sebvince@users.noreply.github.com> * Fix and simplify logic for remove_unnecessary_lr_deps * Add new ref tests for 128x128x(128,64) --- 4aa441a: Rebase fix --- Author: b-shi <brianshi@amd.com> --- 5ba911e: Skip rocRoller for FP4-A/FP4-B + pre-swizzled scale layout (#42) --- Author: Archana Ramalingam <98564406+archana-ramalingam@users.noreply.github.com> --- 9c74998: Rebase fix --- Author: b-shi <brianshi@amd.com> --- 842b149: Addressed review comments for KernelWriter and KernelWriterAssembly --- Author: Koji Nakajima <knakajim@amd.com> --- dce43b1: Fix computeLoadSrd issue --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- c075bbf: Fix preSolution CPU re-sync regressing subtile_mxfp4.yaml --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- ced840f: Fix computeLoadSrd issue (#43) --- Author: bnemanich <brad.nemanich@amd.com> --- bc2f6dd: Small update for gfx950 mx tests + more - enable UseSubtileImpl for all gfx950 non subtile mx tests - skip all gfx950 mxfp8 - use MXScaleFormat=1 as default - set AssertSummationElementMultiple=256 for subtile mxfp4 - fix isSwizzledSubtile in computeLoadSrd --- Author: Koji Nakajima <knakajim@amd.com> --- 5c794b7: Fix gsuasb.yaml failures --- Author: b-shi <brianshi@amd.com> --- 727f8db: tensilelite: add solution reject conditions for UseSubtileImpl=1 (#38) --- Author: Majedul Sujon <85503863+msujon-AMD@users.noreply.github.com> --- 8928fbb: Add more reject conditions for Subtile --- Author: Koji Nakajima <knakajim@amd.com> --- 6e63ab6: Fix kringshift test failures --- Author: b-shi <brianshi@amd.com> --- b3e9724: Update reject condtion for DepthU in subtile case. Plus, update DepthU setting for gfx950 mx test cases --- Author: Koji Nakajima <knakajim@amd.com> --- 5ab6009: Fix build errors --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- 4a4edca: Update more mxfp4 tensilelite test cases --- Author: Koji Nakajima <knakajim@amd.com> --- bbbc553: Update change log --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- 6476c04: Add more reject conditions for gfx950 subtile --- Author: Koji Nakajima <knakajim@amd.com> --- c5828c4: Updated gfx950 mxfp4 test cases - add StreamK setting - skip groupgemm tests for now (groupgemm does not support streamK) --- Author: Koji Nakajima <knakajim@amd.com> --- f1fc2f1: Fix hipblaslt build error of gfx950 --- Author: Koji Nakajima <knakajim@amd.com> --- 70cea1b: Updated subtile_mxfp4.yaml (add StreamK) --- Author: Koji Nakajima <knakajim@amd.com> --- c1c9b2a: Add uninit lsc,lsp, etc.. fields for subtile --- Author: b-shi <brianshi@amd.com> --- c0c1f72: Fixed merge error in testing_matmul.hpp --- Author: Koji Nakajima <knakajim@amd.com> --- 191e0cb: Add missed batch_count >1 changes --- Author: archana-ramalingam <Archana.Ramalingam@amd.com> --- 01c52f8: Addressed PR comments --- Author: Koji Nakajima <knakajim@amd.com> --- 4e89c91: Reduce mxfp4 test time --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- 3dac20f: Prevent overflow for wgmxcc sgpr allocation --- Author: b-shi <brianshi@amd.com> --- 18dec79: Fix error with problem type --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- 9e69ffd: Add a reject conditoin for gfx950 mx + non Subtile --- Author: Koji Nakajima <knakajim@amd.com> --- 0eed3ba: Add more valid GEMM types --- Author: Brad Nemanich <brad.nemanich@amd.com> --- 8b5514e: Fix missing b build error --- Author: archana-ramalingam <Archana.Ramalingam@amd.com> --- f981ff5: Fix 1250 tests --- Author: Brad Nemanich <brad.nemanich@amd.com> --- d1e69d9: Add more FP4 tests --- Author: Brad Nemanich <Brad.Nemanich@amd.com> --- e3a688f: Add MXScaleFormat: 1 to all gfx950 mx test yaml --- Author: Koji Nakajima <knakajim@amd.com> --- aaef3f5: Add DataTypeMXSA,B setting in gfx950 mxfp4 logic yaml --- Author: Koji Nakajima <knakajim@amd.com> --- 861ef8e: Add DataTypeMXSA,B setting in gfx950 mxfp4 logic yaml (nta4,ntb4) --- Author: Koji Nakajima <knakajim@amd.com> Co-authored-by: Archana Ramalingam <Archana.Ramalingam@amd.com> Co-authored-by: Brad Nemanich <Brad.Nemanich@amd.com> Co-authored-by: Brian Shi <Brian.Shi@amd.com> Co-authored-by: James Newling <James.Newling@amd.com> Co-authored-by: Koji Nakajima <Koji.Nakajima@amd.com> Co-authored-by: Majedul Sujon <Majed.Sujon@amd.com> Co-authored-by: Sebastien Vince <Sebastien.Vince@amd.com> Co-authored-by: T.J. Alumbaugh <T.J.Alumbaugh@amd.com> ## Submission Checklist - [ ] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests. --------- Co-authored-by: Archana Ramalingam <Archana.Ramalingam@amd.com> Co-authored-by: Brad Nemanich <Brad.Nemanich@amd.com> Co-authored-by: Brian Shi <Brian.Shi@amd.com> Co-authored-by: James Newling <James.Newling@amd.com> Co-authored-by: Koji Nakajima <Koji.Nakajima@amd.com> Co-authored-by: Majedul Sujon <Majed.Sujon@amd.com> Co-authored-by: Sebastien Vince <Sebastien.Vince@amd.com> Co-authored-by: T.J. Alumbaugh <T.J.Alumbaugh@amd.com>

jayhawk-commits requested review from amd-aakash, ammallya, eidenyoshida, idass1990, naveenrajm7 and rahulc-gh May 1, 2025 19:22

jayhawk-commits added 3 commits May 1, 2025 15:24

Re-adding sparse checkout block in auto-label workflow

93be618

Add step to install requests python module on runners

4e616fd

Make the json loading the same with both scripts

f72c834

jayhawk-commits requested review from ScottTodd and geomin12 May 1, 2025 20:28

Label application/removal is done by the yaml, so remove this else case

b1390f3

geomin12 reviewed May 1, 2025

View reviewed changes

ScottTodd reviewed May 2, 2025

View reviewed changes

Comment thread .github/workflows/pr-auto-label.yml Outdated

Comment thread .github/scripts/pr-auto-label.py Outdated

Comment thread .github/scripts/pr-fanout.py Outdated

Comment thread .github/scripts/pr-auto-label.py Outdated

jayhawk-commits added 6 commits May 4, 2025 19:58

Merge branch 'develop' into joseph-workflowImprovements

76fca4f

Renaming python scripts to use underscores instead of hyphens

4091c83

Update pr-fanout.yml

33ed542

Additional code cleanup

5b26c63

Fix python script path

a139748

jayhawk-commits added 8 commits May 5, 2025 12:34

Interface change when moving from rest api to gh cli

216de30

Update pr_auto_label.py

a553a33

Remove token parameter

9c8ceef

More refactoring to use the client class

3f584d2

Create a model defining the repos-config.json

118e9e8

- Use pydantic to define the entries in the json file and for validation when the json file is consumed. - reponame field no longer used by any existing scripts.

Further refactoring and implementation

7d9b611

- gh cli doesn't support writing GitHub checks, so REST API created to support this. - Implement the functions that synchronize the custom labels and GitHub Checks between the monorepo PR and the fanned out PRs.

Fix bugs from first checks

787eaea

Missing backslashes in multi-line strings

e429faf

assistant-librarian Bot mentioned this pull request Jul 14, 2025

rocsparse-test fails to run on Fedora #627

Closed

assistant-librarian Bot mentioned this pull request Jul 23, 2025

hipblas-test double free error on Fedora. #807

Closed

evbo mentioned this pull request Aug 14, 2025

[Feature] Tensile gfx1032 support #1202

Open

assistant-librarian Bot mentioned this pull request Sep 18, 2025

Kernrl oops on illegal instruction for gfx1151 on Fedora Rawhide #1675

Closed

evetsso pushed a commit to evetsso/rocm-libraries that referenced this pull request Dec 31, 2025

Merge pull request ROCm#8 from AMD-ROCm-Internal/amd/dev/dsicarov/SWD…

fd3fa8f

…EV-536565 SWDEV-536565 Add missing includes to fix build

jmachado-amd pushed a commit to jmachado-amd/rocm-libraries that referenced this pull request Apr 28, 2026

Merge pull request ROCm#8 from jmachado-amd/cpu_geev_static_lapack_ex…

271f209

…ternalproject Link to a locally built OpenBLAS static library

bnemanich mentioned this pull request May 3, 2026

Add gfx950 MXFP4 and BF16 Subtile-based kernel implementation #7017

Merged

1 task

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[GitHub Actions] python-based workflows based on files changed in a pull request#8

[GitHub Actions] python-based workflows based on files changed in a pull request#8
jayhawk-commits merged 31 commits into
developfrom
joseph-workflowImprovements

jayhawk-commits commented May 1, 2025 •

edited

Loading

Uh oh!

geomin12 left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jayhawk-commits commented May 5, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

jayhawk-commits commented May 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Key Workflows

Key Considerations

Test Sequences

Uh oh!

geomin12 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jayhawk-commits commented May 5, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

jayhawk-commits commented May 1, 2025 •

edited

Loading