Skip to content

Update hipDNN roadmap for end of Q1 2026 #5569

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
dpeabody-amd merged 1 commit into from
Mar 18, 2026
Merged

Update hipDNN roadmap for end of Q1 2026 #5569

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
41 changes: 25 additions & 16 deletions projects/hipdnn/docs/Roadmap.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,9 +9,9 @@ This document outlines the development roadmap for hipDNN, a comprehensive graph
>
> ⏳ = In progress

## P0 ~ Q1 2026 (Current milestone)
## P0 ~ Q1 2026

**Focus:** Stable foundation, core operations & initial PyTorch integration
**Focus:** Stable foundation & core operations

### Conv
- **Convolution MIOpen plugin support** ✅
Expand All @@ -21,36 +21,44 @@ This document outlines the development roadmap for hipDNN, a comprehensive graph
### Normalization
- **Batch normalization MIOpen plugin support** ✅
- Including basic fusions ✅
- **Batch normalization Fusilli plugin support** ✅
- **LayerNorm & RMSNorm frontend API** ✅

### GEMM
- **Initial frontend GEMM API support** ✅
- Fusilli plugin integration (see note) ✅
- hipBLASLt plugin initial enablement ⏳
- hipBLASLt plugin initial enablement ✅

### SDPA
- **SDPA frontend API & backend descriptors** ✅

### Core
- **Stable, robust library to build upon** ⏳
- Kernel engine settings (Engine knob configurations API + implementation) ⏳
- **Stable, robust library to build upon** ✅
- Kernel engine settings (Engine knob configurations API + implementation) ✅
- Ex. Flag for enabling benchmarking mode on MIOpen plugin
- Initial Python bindings POC ✅
- Initial benchmarking & performance tooling ✅

### PyTorch
- **PyTorch integration for opt-in hipDNN backend** ⏳
> **Notes:**
> - Fusilli plugin is opt-in, and not defaulted on yet.
> - PyTorch integration was moved to early Q2

> **Notes:** Fusilli plugin is opt-in, and not defaulted on yet.
## P1 ~ Q2 2026 (Current milestone)

## P1 ~ Q2 2026
**Focus:** PyTorch integration, more operations, basic engine selection heuristic & core improvements

**Focus:** More operations, basic engine selection heuristic & core improvements
### PyTorch
- **PyTorch integration for opt-in hipDNN backend** ⏳(Early Q2)

### GEMM
- **hipBLASLt plugin expanded operation & datatype support**
- **hipBLASLt plugin expanded operation & datatype support** ⏳

### SDPA
- Initial limited SDPA frontend API support
- Initial limited SDPA kernel provider (TBD)
- Limited SDPA provided by Fusilli, CK & ASM ⏳

### Normalization
- Adding new **HIP kernel provider plugin** to expand normalization support
- Expanded operation API & coverage to support Layernorm & RMS
- Adding new **HIP kernel provider plugin** to expand normalization support ⏳
- Expanded operation API & coverage to support Layernorm & RMS ⏳
- Expanded layout & datatype coverage for batchnorm

### Heuristics
Expand All @@ -61,6 +69,7 @@ This document outlines the development roadmap for hipDNN, a comprehensive graph
- Plugin SDK utilities to streamline plugin development for new providers ⏳
- Benchmarking & performance python tools ⏳
- Python API wrappers ⏳
- **Client auto-tuning API**

## P2 ~ Q3 2026

Expand All @@ -77,7 +86,7 @@ This document outlines the development roadmap for hipDNN, a comprehensive graph
- Kernel engine tagging & filtering
- Behavioral & numeric notes for filtering
- Client API to enable filtering
- **Client auto-tuning API**


## P3 ~ Q4 2026 & beyond

Expand Down
Loading