Conversation
amd-mtrifuno
requested review from
NaveenElumalaiAMD,
TorreZuk,
amcamd,
babakpst,
bragadeesh,
daineAMD,
mahmoodw,
rkamd and
yoichiyoshida
as code owners
July 16, 2025 13:26
TorreZuk
previously approved these changes
Jul 16, 2025
TorreZuk
left a comment
TorreZuk
left a comment
Contributor
There was a problem hiding this comment.
Need Tensile review and so will tensile PR land first and tensile_tag commit be added?
amd-mtrifuno
added a commit
that referenced
this pull request
Jul 16, 2025
- Add missing changes for gfx1151 - rocBLAS PR with gfx1151 yaml files that requires Tensile changes: #699
assistant-librarian Bot
pushed a commit
to ROCm/Tensile
that referenced
this pull request
Jul 16, 2025
Complete gfx1151 Tensile support (#696) - Add missing changes for gfx1151 - rocBLAS PR with gfx1151 yaml files that requires Tensile changes: ROCm/rocm-libraries#699
amd-mtrifuno
added a commit
that referenced
this pull request
Jul 16, 2025
- Add missing changes for gfx1151 - rocBLAS PR with gfx1151 yaml files that requires Tensile changes: #699
Contributor
Author
Tensile PR merged, I updated tensile_tag. |
…ete_gfx1151_support
TorreZuk
approved these changes
Jul 17, 2025
Contributor
|
@yoichiyoshida or @babakpst can you also review. All tests appear to pass. |
babakpst
approved these changes
Jul 17, 2025
bstefanuk
approved these changes
Jul 17, 2025
amcamd
approved these changes
Jul 17, 2025
amcamd
deleted the
import/develop/amd-mtrifuno_rocBLAS/complete_gfx1151_support
branch
July 17, 2025 16:56
assistant-librarian Bot
pushed a commit
to ROCm/rocBLAS
that referenced
this pull request
Jul 17, 2025
Add rocBLAS yaml files for gfx1151 (#699) - Add Strix Halo yaml files that are copy of Navi33 yaml files (the same changes added for Strix Point) - Need to update tensile_tag.txt after PR with Tensile changes is merged in rocm-libraries and automatically in ROCm/Tensile (ROCm/rocm-libraries#696) --------- Co-authored-by: Torre Zuk <42548444+TorreZuk@users.noreply.github.com>
amd-mtrifuno
added a commit
that referenced
this pull request
Jul 18, 2025
- Add Strix Halo yaml files that are copy of Navi33 yaml files (the same changes added for Strix Point) - Need to update tensile_tag.txt after PR with Tensile changes is merged (#708) --------- Co-authored-by: Torre Zuk <42548444+TorreZuk@users.noreply.github.com>
|
Did you manage to backport this into 7.0.1? |
Contributor
Author
Yes, PR (#744) has been merged into release/rocm-rel-7.0 branch. |
1 task
DDEle
added a commit
that referenced
this pull request
Apr 21, 2026
## Motivation The AITER and FA test dockers (`Dockerfile.aiter`, `Dockerfile.fa`) inherit from the `rocm/pytorch` base image. Recent updates to that base image dropped the `render` group from `/etc/group`, so every parallel test stage now fails on the test agents with: ``` docker: Error response from daemon: Unable to find group render: no matching entries in group file. ``` Jenkins resolves `--group-add render` against the **container's** `/etc/group`, not the host's, so even though the test agents have render in their `/etc/group` (GID 109), the container lookup fails. This pattern affects every recent develop build ([#673](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/673), [#674](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/674), [#686](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/686), [#688](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/688), [#699](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/699), [#708](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/708) — 6 days in a row), where AITER tests fail in seconds and the cascading failure aborts all downstream Build/FMHA/TILE_ENGINE stages. ## Technical Details Add `groupadd -f render` to both `Dockerfile.aiter` and `Dockerfile.fa`, mirroring what the main `Dockerfile` already does (`Dockerfile:96`) and what `Dockerfile.pytorch` does (`Dockerfile.pytorch:4`). The `-f` flag makes it idempotent — silently succeeds if the group already exists. This guarantees the `render` group is always present in the container, regardless of whether the base image happens to ship it. ## Test Plan Triggering AITER CI job: ## Test Result ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
aledudek
pushed a commit
that referenced
this pull request
May 20, 2026
## Motivation The AITER and FA test dockers (`Dockerfile.aiter`, `Dockerfile.fa`) inherit from the `rocm/pytorch` base image. Recent updates to that base image dropped the `render` group from `/etc/group`, so every parallel test stage now fails on the test agents with: ``` docker: Error response from daemon: Unable to find group render: no matching entries in group file. ``` Jenkins resolves `--group-add render` against the **container's** `/etc/group`, not the host's, so even though the test agents have render in their `/etc/group` (GID 109), the container lookup fails. This pattern affects every recent develop build ([#673](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/673), [#674](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/674), [#686](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/686), [#688](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/688), [#699](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/699), [#708](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/708) — 6 days in a row), where AITER tests fail in seconds and the cascading failure aborts all downstream Build/FMHA/TILE_ENGINE stages. ## Technical Details Add `groupadd -f render` to both `Dockerfile.aiter` and `Dockerfile.fa`, mirroring what the main `Dockerfile` already does (`Dockerfile:96`) and what `Dockerfile.pytorch` does (`Dockerfile.pytorch:4`). The `-f` flag makes it idempotent — silently succeeds if the group already exists. This guarantees the `render` group is always present in the container, regardless of whether the base image happens to ship it. ## Test Plan Triggering AITER CI job: ## Test Result ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
ammallya
pushed a commit
that referenced
this pull request
Jun 8, 2026
#699) * Add initial implementation of tensor add tensor host * Update both sources with respective strides * Initial working version of tensor add tensor host without broadcasting * Add broadcast support file and make modifications to tensor_add_tensor.cpp file * Current test patch here * Cleanup tensor_add_tensor and changes to accomodate shift of source pointer at beginning * Make tensor_add_tensor_recursive templated * Add initial implementation for F16 implementation for tensor_add_tensor * Changes currently in test suite for testing tensor operations * Tensor operations for arithmetic * Add support for subtraction, division and multiply across HOST * Fix linker errors and other issues * Make changes for F16 test invocation for tensor arithmetic operations * Add declarations for tensor operations F16 bit depth in host_tensor_executors.hpp * Fix compilation issues * Add AVX implementation for broadcastNDim = 1 * Make changes to have code for else case inside broadcaseNDim = 1 * Updates for src2shape == 1 for broadcastNDim = 1 * Add code for src1shape == 1 * Add code for broadcastNDim = 2 * Add updates for numDims=3 * Fix issues with broadcastNDim = 3 * Add vectorized versions of divide, multiply and sub operations for various dims * Fix compilation with subtract, multiply and divide * Update the leftover part processed by raw C * Bug fixes for src1shape == 1 and broadastNDim = 2 * Add AVX Version for F16 version of tensor_multiply_tensor * Update the store function * Update Rpp32f* to Rpp16f* * Modify to cast the pointer * Updates to case when shape of input = 1 * Test setup * Add updates to reduce redundant GetShapes function calls * Fix compiler issues * Fix compiler issues new * Fix bug * Further improvements * Initial cleanup * Add vectorized F16 implementations for add, divide and subtract * Add initial implementation for enclosing all tensor ops in a single unit * Add list of enum operations * Fix issues with division * Update to use references * Updates to rppt calls * Compile fixes * Initial implementation for integer bit depths - raw C * Add vectorized implementations for each data type * Initial set of changes for HIP Arithmetic implementation * Add data types support * Add fixes for compilation * Modify test suite for testing purposes * Update test suite * Fix issues with I8 implementation * Add separate code path for I16 implementation * Fix accuracy issues with 16 bit integers * Create separate path for I32 integers * Fix compilation issues' * Make initial updates for separate division module * Fix compilation issues * Add initial dummy code for tensor divide tensor * Update the tensor binary operations and fix compilation issues * SIMD Updates for tensor_binary_operations.cpp * Fixes for SIMD Version * Further code * Add required fixes * Update float based division on HIP Side * Update AVX code for U8/I8 datatypes * Fix compilation issues * Fix all accuracy issues * Add QA support for U8 tensor add, subtract and multiply * Update the reference outputs * Changes to the miscellaneous test suite * Add input bin files * Add vector code for 2D cases * Updates for 1D version * Updates to vectorize 3D case of broadcasting * Fixes for 3D vector version * Remove unwanted code * Some initial cleanup * Further cleanup * Changed from function ptr to Structure member func * Changed from function ptr to Structure member func * API changes and bin files consolidation * Bin file consolidation and cleanup * Added Bitdepth and structures * Add Arithmetic divide datatype conversions * Fix Divide Segmentation issue, Modified Broadcast conditions and Added ROI in Testsuite. * Fix F32 test case in Testsuite file. * Fix i32 not working and return type of simd set value function * Add Output statement to check AVX and Scalar. * Fix 3D AVX roi changes, recursive function and Divide templates * Fix the rename of BitdepthTestmode and equivalent Dataype conversion format. * Fix corrections in Comments and Datatype I32 & U32 format range. * Fix remove cerr statements and I8 datatype change. * Add unified api for arithmetic operators * Modified test suite and updated the rpp version * resolve review comments * resolve review comments * Add copyright * resolve review comments * add missing bitdepths * resolve review comments * Fix dropout issues * PR684 update with multiple fixes for performance improvement and test suit fix * addressed review comments * Update rpp_test_suite_image.h * fix review comments * added braces for for loop for clarity --------- Co-authored-by: Srihari-mcw <srihari@multicorewareinc.com> Co-authored-by: ManasaDattaT <tammisetti.manasadatta@multicorewareinc.com> Co-authored-by: Mukesh <mukesh.jayakodi@multicorewareinc.com> Co-authored-by: HazarathKumarM <119284987+HazarathKumarM@users.noreply.github.com> Co-authored-by: HazarathKumarM <hazarathkumar@multicorewareinc.com> Co-authored-by: Abishek <52214183+r-abishek@users.noreply.github.com> Co-authored-by: Lakshmi Kumar <lakshmi.kumar@amd.com>
kliegeois
pushed a commit
to kliegeois/rocm-libraries
that referenced
this pull request
Jun 8, 2026
ROCm#699) * Add initial implementation of tensor add tensor host * Update both sources with respective strides * Initial working version of tensor add tensor host without broadcasting * Add broadcast support file and make modifications to tensor_add_tensor.cpp file * Current test patch here * Cleanup tensor_add_tensor and changes to accomodate shift of source pointer at beginning * Make tensor_add_tensor_recursive templated * Add initial implementation for F16 implementation for tensor_add_tensor * Changes currently in test suite for testing tensor operations * Tensor operations for arithmetic * Add support for subtraction, division and multiply across HOST * Fix linker errors and other issues * Make changes for F16 test invocation for tensor arithmetic operations * Add declarations for tensor operations F16 bit depth in host_tensor_executors.hpp * Fix compilation issues * Add AVX implementation for broadcastNDim = 1 * Make changes to have code for else case inside broadcaseNDim = 1 * Updates for src2shape == 1 for broadcastNDim = 1 * Add code for src1shape == 1 * Add code for broadcastNDim = 2 * Add updates for numDims=3 * Fix issues with broadcastNDim = 3 * Add vectorized versions of divide, multiply and sub operations for various dims * Fix compilation with subtract, multiply and divide * Update the leftover part processed by raw C * Bug fixes for src1shape == 1 and broadastNDim = 2 * Add AVX Version for F16 version of tensor_multiply_tensor * Update the store function * Update Rpp32f* to Rpp16f* * Modify to cast the pointer * Updates to case when shape of input = 1 * Test setup * Add updates to reduce redundant GetShapes function calls * Fix compiler issues * Fix compiler issues new * Fix bug * Further improvements * Initial cleanup * Add vectorized F16 implementations for add, divide and subtract * Add initial implementation for enclosing all tensor ops in a single unit * Add list of enum operations * Fix issues with division * Update to use references * Updates to rppt calls * Compile fixes * Initial implementation for integer bit depths - raw C * Add vectorized implementations for each data type * Initial set of changes for HIP Arithmetic implementation * Add data types support * Add fixes for compilation * Modify test suite for testing purposes * Update test suite * Fix issues with I8 implementation * Add separate code path for I16 implementation * Fix accuracy issues with 16 bit integers * Create separate path for I32 integers * Fix compilation issues' * Make initial updates for separate division module * Fix compilation issues * Add initial dummy code for tensor divide tensor * Update the tensor binary operations and fix compilation issues * SIMD Updates for tensor_binary_operations.cpp * Fixes for SIMD Version * Further code * Add required fixes * Update float based division on HIP Side * Update AVX code for U8/I8 datatypes * Fix compilation issues * Fix all accuracy issues * Add QA support for U8 tensor add, subtract and multiply * Update the reference outputs * Changes to the miscellaneous test suite * Add input bin files * Add vector code for 2D cases * Updates for 1D version * Updates to vectorize 3D case of broadcasting * Fixes for 3D vector version * Remove unwanted code * Some initial cleanup * Further cleanup * Changed from function ptr to Structure member func * Changed from function ptr to Structure member func * API changes and bin files consolidation * Bin file consolidation and cleanup * Added Bitdepth and structures * Add Arithmetic divide datatype conversions * Fix Divide Segmentation issue, Modified Broadcast conditions and Added ROI in Testsuite. * Fix F32 test case in Testsuite file. * Fix i32 not working and return type of simd set value function * Add Output statement to check AVX and Scalar. * Fix 3D AVX roi changes, recursive function and Divide templates * Fix the rename of BitdepthTestmode and equivalent Dataype conversion format. * Fix corrections in Comments and Datatype I32 & U32 format range. * Fix remove cerr statements and I8 datatype change. * Add unified api for arithmetic operators * Modified test suite and updated the rpp version * resolve review comments * resolve review comments * Add copyright * resolve review comments * add missing bitdepths * resolve review comments * Fix dropout issues * PR684 update with multiple fixes for performance improvement and test suit fix * addressed review comments * Update rpp_test_suite_image.h * fix review comments * added braces for for loop for clarity --------- Co-authored-by: Srihari-mcw <srihari@multicorewareinc.com> Co-authored-by: ManasaDattaT <tammisetti.manasadatta@multicorewareinc.com> Co-authored-by: Mukesh <mukesh.jayakodi@multicorewareinc.com> Co-authored-by: HazarathKumarM <119284987+HazarathKumarM@users.noreply.github.com> Co-authored-by: HazarathKumarM <hazarathkumar@multicorewareinc.com> Co-authored-by: Abishek <52214183+r-abishek@users.noreply.github.com> Co-authored-by: Lakshmi Kumar <lakshmi.kumar@amd.com> [ROCm/rpp commit: 8ab057f]
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add Strix Halo yaml files that are copy of Navi33 yaml files (the same changes added for Strix Point)
Need to update tensile_tag.txt after PR with Tensile changes is merged in rocm-libraries and automatically in ROCm/Tensile (Complete gfx1151 Tensile support #696)