[PR #185] [CLOSED] [HIP] Add support for gfx90c and gfx1010 GPU architectures with tiled GEMM fallback #203

New issue

Closed

opened 2026-10-26 22:21:32 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 22:21:32 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/Comfy-Org/comfy-kitchen/pull/185
Author: @Zakhrov
Created: 9/17/2026
Status: ❌ Closed

Base: main ← Head: main

📝 Commits (10+)

9507848 Add support for gfx90c and gfx1010 GPU architectures and optimize flash decode kernel
e0075df Refactor HIP backend for Sage attention kernels and enhance testing
83f53c3 Update error message for wave32 requirement
5e0ffef Refactor tests and improve code readability
da75655 Merge branch 'Comfy-Org:main' into main
54f400b Fix failing tests on gfx90c
a4cd6c0 Update README to reflect gfx90c and gfx1010 support
3750a88 Enable software GEMM tiling and attention for Vega, RDNA1 and RDNA2
7c226c3 Refactor HIP backend for Sage attention kernels and enhance testing
a9030ef Merge branch 'Comfy-Org:main' into main

📊 Changes

28 files changed (+6322 additions, -3696 deletions)

View changed files

📝 README.md (+39 -26)
📝 comfy_kitchen/backends/eager/quantization.py (+220 -70)
📝 comfy_kitchen/backends/hip/CMakeLists.txt (+34 -2)
📝 comfy_kitchen/backends/hip/__init__.py (+775 -256)
📝 comfy_kitchen/backends/hip/architecture_config.h.in (+9 -0)
📝 comfy_kitchen/backends/hip/architectures.json (+3 -1)
📝 comfy_kitchen/backends/hip/dlpack_bindings.cpp (+2369 -2023)
📝 comfy_kitchen/backends/hip/mma.h (+522 -304)
📝 comfy_kitchen/backends/hip/ops/flash_decode.hip (+299 -264)
📝 comfy_kitchen/backends/hip/rope_math.h (+60 -58)
📝 comfy_kitchen/backends/hip/sage_attention/sage_common.h (+229 -193)
📝 comfy_kitchen/backends/triton/__init__.py (+54 -8)
📝 comfy_kitchen/backends/triton/adaln.py (+53 -16)
📝 comfy_kitchen/backends/triton/na.py (+80 -21)
📝 comfy_kitchen/backends/triton/quantization.py (+214 -58)
📝 comfy_kitchen/backends/triton/rms_rope.py (+46 -15)
📝 comfy_kitchen/backends/triton/rope.py (+94 -26)
📝 comfy_kitchen/tensor/base.py (+18 -4)
📝 comfy_kitchen/tensor/int8.py (+51 -14)
📝 tests/test_constraints.py (+29 -3)

...and 8 more files

📄 Description

Updated architectures.json to include support for gfx90c and gfx1010.
Refactored flash_decode.hip to improve code readability and maintainability.
Enhanced the flash_decode_kernel to handle different GPU architectures with conditional compilation.
Optimized memory access patterns and calculations within the kernel for better performance.
Ensured compatibility with existing functionality while improving overall efficiency.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/Comfy-Org/comfy-kitchen/pull/185 **Author:** [@Zakhrov](https://github.com/Zakhrov) **Created:** 9/17/2026 **Status:** ❌ Closed **Base:** `main` ← **Head:** `main` --- ### 📝 Commits (10+) - [`9507848`](https://github.com/Comfy-Org/comfy-kitchen/commit/9507848e58de940a821d90383f33b27f6f1c9492) Add support for gfx90c and gfx1010 GPU architectures and optimize flash decode kernel - [`e0075df`](https://github.com/Comfy-Org/comfy-kitchen/commit/e0075df9201df401fd89bad966860385e26d1081) Refactor HIP backend for Sage attention kernels and enhance testing - [`83f53c3`](https://github.com/Comfy-Org/comfy-kitchen/commit/83f53c3622b04597edf70c7af4473af85a365947) Update error message for wave32 requirement - [`5e0ffef`](https://github.com/Comfy-Org/comfy-kitchen/commit/5e0ffef2b098944ee1dbbcff79b4c733e050c524) Refactor tests and improve code readability - [`da75655`](https://github.com/Comfy-Org/comfy-kitchen/commit/da756553cb53a4fbc60ccb80a97cb637f27115d9) Merge branch 'Comfy-Org:main' into main - [`54f400b`](https://github.com/Comfy-Org/comfy-kitchen/commit/54f400b9f568cdceb6cb0f281cb996ecd587ec6b) Fix failing tests on gfx90c - [`a4cd6c0`](https://github.com/Comfy-Org/comfy-kitchen/commit/a4cd6c0e7c1d76176886d9d6b4b562a1e78c916e) Update README to reflect gfx90c and gfx1010 support - [`3750a88`](https://github.com/Comfy-Org/comfy-kitchen/commit/3750a8806675f3c4479153bcff4813350f156bdf) Enable software GEMM tiling and attention for Vega, RDNA1 and RDNA2 - [`7c226c3`](https://github.com/Comfy-Org/comfy-kitchen/commit/7c226c36858be335d4098614137953e038c44a19) Refactor HIP backend for Sage attention kernels and enhance testing - [`a9030ef`](https://github.com/Comfy-Org/comfy-kitchen/commit/a9030efb1f20b1b01c0b2cdcad6a9d0220dbc151) Merge branch 'Comfy-Org:main' into main ### 📊 Changes **28 files changed** (+6322 additions, -3696 deletions) <details> <summary>View changed files</summary> 📝 `README.md` (+39 -26) 📝 `comfy_kitchen/backends/eager/quantization.py` (+220 -70) 📝 `comfy_kitchen/backends/hip/CMakeLists.txt` (+34 -2) 📝 `comfy_kitchen/backends/hip/__init__.py` (+775 -256) 📝 `comfy_kitchen/backends/hip/architecture_config.h.in` (+9 -0) 📝 `comfy_kitchen/backends/hip/architectures.json` (+3 -1) 📝 `comfy_kitchen/backends/hip/dlpack_bindings.cpp` (+2369 -2023) 📝 `comfy_kitchen/backends/hip/mma.h` (+522 -304) 📝 `comfy_kitchen/backends/hip/ops/flash_decode.hip` (+299 -264) 📝 `comfy_kitchen/backends/hip/rope_math.h` (+60 -58) 📝 `comfy_kitchen/backends/hip/sage_attention/sage_common.h` (+229 -193) 📝 `comfy_kitchen/backends/triton/__init__.py` (+54 -8) 📝 `comfy_kitchen/backends/triton/adaln.py` (+53 -16) 📝 `comfy_kitchen/backends/triton/na.py` (+80 -21) 📝 `comfy_kitchen/backends/triton/quantization.py` (+214 -58) 📝 `comfy_kitchen/backends/triton/rms_rope.py` (+46 -15) 📝 `comfy_kitchen/backends/triton/rope.py` (+94 -26) 📝 `comfy_kitchen/tensor/base.py` (+18 -4) 📝 `comfy_kitchen/tensor/int8.py` (+51 -14) 📝 `tests/test_constraints.py` (+29 -3) _...and 8 more files_ </details> ### 📄 Description - Updated `architectures.json` to include support for gfx90c and gfx1010. - Refactored `flash_decode.hip` to improve code readability and maintainability. - Enhanced the `flash_decode_kernel` to handle different GPU architectures with conditional compilation. - Optimized memory access patterns and calculations within the kernel for better performance. - Ensured compatibility with existing functionality while improving overall efficiency. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-26 22:21:32 +00:00