Skip to content

Navigation Menu

Appearance settings

- AI CODE CREATION
  GitHub CopilotWrite better code with AI
  GitHub Copilot appDirect agents from issue to merge
  MCP Registry^NewIntegrate external tools
- DEVELOPER WORKFLOWS
  ActionsAutomate any workflow
  CodespacesInstant dev environments
  IssuesPlan and track work
  Code ReviewManage code changes
- APPLICATION SECURITY
  GitHub Advanced SecurityFind and fix vulnerabilities
  Code securitySecure your code as you build
  Secret protectionStop leaks before they start
- EXPLORE
  Why GitHub
  Documentation
  Blog
  Changelog
  Marketplace
View all features
- BY COMPANY SIZE
  Enterprises
  Small and medium teams
  Startups
  Nonprofits
- BY USE CASE
  App Modernization
  DevSecOps
  DevOps
  CI/CD
  View all use cases
- BY INDUSTRY
  Healthcare
  Financial services
  Manufacturing
  Government
  View all industries
View all solutions
- EXPLORE BY TOPIC
  AI
  Software Development
  DevOps
  Security
  View all topics
- EXPLORE BY TYPE
  Customer stories
  Events & webinars
  Ebooks & reports
  Business insights
  GitHub Skills
- SUPPORT & SERVICES
  Documentation
  Customer support
  Community forum
  Trust center
  Partners
View all resources
- COMMUNITY
  GitHub SponsorsFund open source developers
- PROGRAMS
  Security Lab
  Maintainer Community
  Accelerator
  GitHub Stars
  Archive Program
- REPOSITORIES
  Topics
  Trending
  Collections
- ENTERPRISE SOLUTIONS
  Enterprise platformAI-powered developer platform
- AVAILABLE ADD-ONS
  GitHub Advanced SecurityEnterprise-grade security features
  Copilot for BusinessEnterprise-grade AI features
  Premium SupportEnterprise-grade 24/7 support
Pricing

Search code, repositories, users, issues, pull requests...

Search

Clear

Search syntax tips

Provide feedback

We read every piece of feedback, and take your input very seriously.

Include my email address so I can be contacted

Saved searches

Use saved searches to filter your results more quickly

Name

Query

To see all available qualifiers, see our documentation.

Appearance settings

You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session.

Dismiss alert

ypapadop-amd / ggml Public

forked from ggml-org/ggml

Notifications You must be signed in to change notification settings
Fork 4
Star 33

Code
Issues 16
Pull requests 3
Actions
Projects
Security and quality
Insights

Additional navigation options

Code
Issues
Pull requests
Actions
Projects
Security and quality
Insights

Layers for MNIST example#177

Draft

ypapadop-amd wants to merge 141 commits into

hsa-backendypapadop-amd/ggml:hsa-backendfrom

ypapadop-amd/mnist-layersypapadop-amd/ggml:ypapadop-amd/mnist-layersCopy head branch name to clipboard

Conversation Commits141 (141)Checks Files changed

Draft

Layers for MNIST example#177
ypapadop-amd wants to merge 141 commits into
hsa-backendfrom
ypapadop-amd/mnist-layers

Commits

Commits on Feb 27, 2026

HSA backend initialization (#3)
ypapadop-amd
committed
Adding agent information to internal datastructures (#4)
ypapadop-amd
committed
Memory pool information and buffer allocation support (#5)
ypapadop-amd
committed
HSA tensor data functions (#6)
ypapadop-amd
committed
Support in HSA backend for small examples support (#7)
ypapadop-amd
committed
Fix for correctly choosing memories and reporting size (#8)
ypapadop-amd
committed
HSA backend function implementations (#9)
ypapadop-amd
committed
More HSA backend function implementations (#10)
ypapadop-amd
committed
HSA operations fallback to CPU (#11)
ypapadop-amd
committed
HSA queue support (#12)
ypapadop-amd
committed
Use GGML_ABORT instead of abort() (#13)
ypapadop-amd
committed
Support for CPY kernel (#14)
ypapadop-amd
committed
CPU backend fallback (#15)
ypapadop-amd
committed
Replacing mutex and bool with call_once (#17)
ypapadop-amd
committed
Support for loading pre-compiled NPU kernels (#18)
ypapadop-amd
committed
Exception and error handling (#25)
ypapadop-amd
committed
Clang format support (#26)
ypapadop-amd
committed
Free packet memory after synchronization (#28)
ypapadop-amd
committed
ggml_backend_buffer_init_tensor support (#29)
ypapadop-amd
committed
Using alias for std::filesystem (#30)
ypapadop-amd
committed
Matrix multiplication kernel (#31)
ypapadop-amd
committed
Binary instructions format loading (#32)
ypapadop-amd
committed
Release memory allocated for packets (#33)
ypapadop-amd
committed
Directly create PDI instead through xclbin (#34)
ypapadop-amd
committed
HSA tensor extra metadata (#37)
ypapadop-amd
committed
Various fixes (#38)
ypapadop-amd
committed
CMake Refactor (#39)
ypapadop-amd
committed
Stand-alone kernel discovery header (#40)
ypapadop-amd
committed
Renaming kernel to the IRON equivalent (#41)
ypapadop-amd
committed
Fix CPU emulation ops (#42)
ypapadop-amd
committed
Miscelaneous cleanups (#43)
ypapadop-amd
committed
Support user kernel directory (#44)
ypapadop-amd
committed
Using GGML_STATUS_ALLOC_FAILED when allocation has failed (#45)
ypapadop-amd
committed
Using correct size for inputs and outputs (#46)
ypapadop-amd
committed
Support non uniform sizes (#47)
ypapadop-amd
committed
JIT compilation (alpha) (#49)
ypapadop-amd
committed
Invoking python script via pybind11 (#50)
ypapadop-amd
committed
Adding copyright (#51)
ypapadop-amd
committed
Fixing MLIR-AIE discovery (#52)
ypapadop-amd
committed
Using find_package(AIE) (#53)
ypapadop-amd
committed
Cleanups (#54)
ypapadop-amd
committed
Using existing ggml_n_dims (#55)
ypapadop-amd
committed
JIT for core functions via Peano (#56)
ypapadop-amd
committed
JIT cleanups (#57)
ypapadop-amd
committed
Fix for CMake arg (#58)
ypapadop-amd
committed
Use compiler wrapper in CMake (#59)
ypapadop-amd
committed
Flatten IRON kernel support directories (#60)
ypapadop-amd
committed
Vector-vector ops (#61)
ypapadop-amd
committed
Rename files to follow naming scheme (#62)
ypapadop-amd
committed
Support for clearing kernel cache (#63)
ypapadop-amd
committed
Unary op support (#64)
ypapadop-amd
committed
Jit cleanup (#65)
ypapadop-amd
committed
Associating core functions with kernels (#66)
ypapadop-amd
committed
Remove redundant tensor (#67)
ypapadop-amd
committed
Core function refactor (#68)
ypapadop-amd
committed
Unary ops implementations (#69)
ypapadop-amd
committed
Flatten tensors (#71)
ypapadop-amd
committed
JIT verbose output (#72)
ypapadop-amd
committed
Mulmat fixes (#73)
ypapadop-amd
committed
Host implementations for CPY, DUP, CONT (#74)
ypapadop-amd
committed
Remove prebuilt kernels (#75)
ypapadop-amd
committed
Use correct environment variable name (#77)
ypapadop-amd
committed
Fixes for MUL_MAT (#78)
ypapadop-amd
committed
Moving kernels to per-architecture directory (#79)
ypapadop-amd
committed
JIT refactor (#80)
ypapadop-amd
committed
Various cleanups (#81)
ypapadop-amd
committed
Replacing NPU1Col4 with NPU1 (#82)
ypapadop-amd
committed
Remove AIEDevice.npu1_4col (#83)
ypapadop-amd
committed
Remove NPU1Col4 (#84)
ypapadop-amd
committed
Flatten view support (#85)
ypapadop-amd
committed
Using HSA_AMD_MEMORY_POOL_INFO_RUNTIME_ALLOC_REC_GRANULE to detect AIE dev heap (#86)
ypapadop-amd
committed
update iron mm.cc file and includes, modify tile sizes for core function compilation of different data types, small bug fix in ggml-hsa.cpp (#88)

efurst
authored and
ypapadop-amd
committed
Mul mat fixes (#89)
ypapadop-amd
committed
Per-device alignment requirements (#90)
ypapadop-amd
committed
Fix check if a tensor can be flattened (#91)
ypapadop-amd
committed
Tensor view copy (#92)
ypapadop-amd
committed
Better errors (#93)
ypapadop-amd
committed
Updating for latest IRON (#94)
ypapadop-amd
committed
IRON environment (#95)
ypapadop-amd
committed
Fixing import to IRON dtype_to_str
ypapadop-amd
committed
Replace CoreFunction with ExternalFunction (#97)
ypapadop-amd
committed
Kernel compilation at tensor initialization (#98)
ypapadop-amd
committed
Update mul_mat from IRON (#99)
ypapadop-amd
committed
Abstract kernels (#100)
ypapadop-amd
committed
Using matmul kernel column major support to avoid temporary tensors (#101)
ypapadop-amd
committed
Kernel refactor (#102)
ypapadop-amd
committed
Relocate core files (#103)
ypapadop-amd
committed
Use datatype and number of elements in tile calculation (#104)

ypapadop-amd
and
Copilot
committed
Remove emulated tensors (#105)
ypapadop-amd
committed
Updated environment instructions (#106)
ypapadop-amd
committed
Convert FP16 to BF16 (#107)
ypapadop-amd
committed
Updating IRON environment set-up
ypapadop-amd
committed
Update IRON environment set-up (#108)
ypapadop-amd
committed
Adding pytest to Python requirements (#109)
ypapadop-amd
committed
Operations refactor (#110)
ypapadop-amd
committed
Aligning tensor sizes for bf16 / int8 / int16 (#111)
ypapadop-amd
committed
Return false is_host for HSA memory and refactor error messages (#112)
ypapadop-amd
committed
Adding new ggml_backend_i member (#115)
ypapadop-amd
committed
Run time logging (#116)
ypapadop-amd
committed
Move tests to test/ggml-hsa (#122)
ypapadop-amd
committed
Generic iron kernels (#123)

ypapadop-amd
and
Copilot
committed
Add GitHub Copilot instructions for ggml-hsa (#133)

Copilot
and
ypapadop-amd
committed
Update latest iron (#136)
ypapadop-amd
committed
Remove core function (#137)
ypapadop-amd
committed
Doc and python check updates (#139)

ypapadop-amd
and
Copilot
committed
Adding new unary ops (#141)
ypapadop-amd
committed
Removing define
ypapadop-amd
committed
Update README on supported NPUs and prerequisites (#144)

ypapadop-amd
and
Copilot
committed
Remove python file registration from C++ (#145)

ypapadop-amd
and
Copilot
committed
Generic backend kernel support (#146)
ypapadop-amd
committed
TensorDesc as dataclass (#147)

ypapadop-amd
and
Copilot
committed
Update README on how to compile (#148)

ypapadop-amd
and
Copilot
committed
Pass op_params argument to kernels in compilation

artulab
authored and
ypapadop-amd
committed
aie2p MAT_MUL (#150)
ypapadop-amd
committed
Implement SCALE kernel

artulab
authored and
ypapadop-amd
committed
Formatting
ypapadop-amd
committed
Bump MLIR-AIE and ROCm versions (#156)
ypapadop-amd
committed
Bump mlir-aie to 1.2.0 (#158)

ypapadop-amd
and
Copilot
committed
Use correct import for NPU1 / NPU2 (#159)
ypapadop-amd
committed
Implement softmax kernel

artulab
authored and
ypapadop-amd
committed
Vector ops refactoring (#160)

ypapadop-amd
and
Copilot
committed
feat(ggml-hsa): bump mlir-aie dependency from v1.2.0 to v1.2.1 (#163)
ypapadop-amd
committed
GGML_OP_CLAMP support (#164)

ypapadop-amd
and
Copilot
committed
Simplify core function compilation (#161)

committed
Support kernel names without ggml_op entry (#165)
ypapadop-amd
committed
Fix circular buffer overrun (#166)

ypapadop-amd
and
Copilot
authored

Commits on Mar 6, 2026

Deferred compilation (#167)

ypapadop-amd
and
Copilot
authored
ci: add auto-format workflow for ggml-hsa directory (#168)
ypapadop-amd
authored

Commits on Mar 7, 2026

Fallback to data memory if kernarg is not available (#169)

authored
Add scalar ARGMAX kernel for HSA backend (#171)

authored

Commits on Mar 9, 2026

feat(ggml-hsa): add COUNT_EQUAL operation kernel (#173)

authored
Implement CROSS_ENTROPY_LOSS, and improve vec_exp in SOFT_MAX
artulab
committed
Auto-format code in src/ggml-hsa

github-actions[bot]
authored and
artulab
committed
Vector op broadcast (#174)

authored
Code documentation (#176)
ypapadop-amd
authored

Commits on Mar 10, 2026

feat(hsa): add null pointer assertions in AIE kernel dispatch (#178)
ypapadop-amd
authored
refactor(hsa): move scalar math functions to shared header
ypapadop-amd
committed
feat(hsa): move pow2, floor_log2 and alibi_slope math functions
ypapadop-amd
committed
Auto-format code in src/ggml-hsa

github-actions[bot]
authored and
ypapadop-amd
committed
feat(ggml-hsa): add scalar f32 matmul support for AIE2/AIE2p
ypapadop-amd
committed
f32 support in MUL_MAT
ypapadop-amd
committed

Footer

© 2026 GitHub, Inc.

Footer navigation

Terms
Privacy
Security
Status
Community
Docs
Contact

You can’t perform that action at this time.