Releases: Genesis-Embodied-AI/quadrants
Release list
v1.1.4b3
Pre-release v1.1.4b3
This pre-release is to test migrating metal float atomic add to use native atomics, rather than CAS.
What's Changed
AdStack
- [AdStack] Gate the adstack size-expr subtraction fusion on shared source-loop provenance. by @duburcqa in #795
- [AdStack] Propagate the adstack size-expr onto scalarized tensor components. by @duburcqa in #796
- [AdStack] Fix the reverse-mode gradient of a loop-carried vector rebuilt across a dynamic loop. by @duburcqa in #799
- [AdStack] Fix SPIR-V dominance of the adstack heap stride loads. by @duburcqa in #806
- [AdStack] Run the ndarray autodiff tests on Metal and restrict f64 to where it is required. by @duburcqa in #800
Data Oriented
- [Data oriented] Add template_primitives parameter default True by @hughperkins in #789
Doc
- [Doc] Fix doc quality check flagged issues on atomics.md by @hughperkins in #792
Graph
- [Graph] Migrate to qd.graph namespace by @hughperkins in #798
Build
- [Build] Raise macOS minimum deployment target to 13.0 by @hughperkins in #794
Atomics
- [Atomics] Migrate metal to use native float atomics by @hughperkins in #788
Full Changelog: v1.1.3...v1.1.4b3
v1.1.4b2
Pre-release v1.1.4b2
This pre-release is to test changing mac minimum version from 11 to 13.
- [Build] Raise macOS minimum deployment target to 13.0 by @hughperkins in #794 (comment)
Full Changelog: v1.1.3...v1.1.4b2
v1.1.4b1
Pre-release v1.1.4b1
This pre-release is to test running CSE per offload rather than over whole kernel.
- [Optimization] Make CSE per offload rather than whole kernel by @hughperkins in #790
Full Changelog: v1.1.3...v1.1.4b1
v1.1.3
Release v1.1.3
What's Changed
AMDGPU
- [AMDGPU] Don't emit llvm.amdgcn.permlane64 on CDNA by @hughperkins in #746
SPIRV
Tiles
- [Tiles] Use unpacked vector field for Tile16x16/Tile32x32 register storage by @hughperkins in #722
Graph
- [Graph] Support member ndarrays for qd.checkpoint and qd.graph_do_while by @hughperkins in #760
- [Graph] Add graph parallel by @hughperkins in #756
Lang
- [Lang] Make qd.static not bypass the pure-kernel purity check by @hughperkins in #733
New Contributors
Full Changelog: v1.1.2...v1.1.3
v1.1.3b2
Pre-release v1.1.3b2
This pre-release is to test making CSE per-offload, rather than per whole kernel.
What's Changed
Optimization
- [Optimization] Make CSE per offload rather than whole kernel by @hughperkins in #790
AMDGPU
- [AMDGPU] Don't emit llvm.amdgcn.permlane64 on CDNA by @hughperkins in #746
SPIRV
Tiles
- [Tiles] Use unpacked vector field for Tile16x16/Tile32x32 register storage by @hughperkins in #722
Graph
- [Graph] Support member ndarrays for qd.checkpoint and qd.graph_do_while by @hughperkins in #760
- [Graph] Add graph parallel by @hughperkins in #756
New Contributors
Full Changelog: v1.1.2...v1.1.3b2
v1.1.3b1
Pre-release v1.1.3b1
This pre-release is to test migrating f32 atomic add/sub on metal to use native op, rather than CAS.
What's Changed
Atomics
- [Atomics] Migrate metal to use native float atomics by @hughperkins in #788
AMDGPU
- [AMDGPU] Don't emit llvm.amdgcn.permlane64 on CDNA by @hughperkins in #746
SPIRV
Tiles
- [Tiles] Use unpacked vector field for Tile16x16/Tile32x32 register storage by @hughperkins in #722
Graph
- [Graph] Support member ndarrays for qd.checkpoint and qd.graph_do_while by @hughperkins in #760
- [Graph] Add graph parallel by @hughperkins in #756
New Contributors
Full Changelog: v1.1.2...v1.1.3b1
v1.1.2
Release v1.1.2
This release aims to fix a bug where Quadrants would no longer run on cuda toolkit 12.8 drivers.
What's Changed
CI
- [CI] Add CI non ascii check by @hughperkins in #778
- [CI] Fix non-ASCII scanner treating "+++" added lines as file headers by @hughperkins in #784
- [CI] Render PR change report in job summary instead of a standalone check by @hughperkins in #783
Doc
- [Doc] Fix fastcache doc by @hughperkins in #782
- [Doc] Fix subgroup md doc check by @hughperkins in #781
- [Doc] Fix graph doc issues by @hughperkins in #762
Graph
- [Graph] Re-add SM110 support, and use Toolkit 12.8 for SM90,100,120 by @hughperkins in #780
Full Changelog: v1.1.1...v1.1.2
v1.1.1
Release v1.1.1
This release silences nanobind leak warnings, that were introduced in v1.1.0.
What's Changed
CI
- [CI] Exempt advanced-section terms from doc-quality undefined-term rule by @hughperkins in #763
- [CI] Add doc sequence check to doc quality check by @hughperkins in #764
Doc
- [Doc] Use American spelling throughout user guide by @hughperkins in #765
- [Doc] Fix compound types doc by @hughperkins in #767
Build
- [Build] Use CMAKE_ARGS directly, drop legacy QUADRANTS_CMAKE_ARGS by @hughperkins in #753
AMDGPU
[DLPack]
Nanobind
- [Nanobind] Silence interpreter-shutdown leak warnings by @hughperkins in #779
New Contributors
Full Changelog: v1.1.0...v1.1.1
v1.1.1b2
Pre-release v1.1.1b2
This pre-release is in order to test the nanobind warnings suppression for Genesis.
What's Changed
- [CI] Exempt advanced-section terms from doc-quality undefined-term rule by @hughperkins in #763
- [Docs] Use American spelling throughout user guide by @hughperkins in #765
- [CI] Add doc sequence check to doc quality check by @hughperkins in #764
- [Build] Use CMAKE_ARGS directly, drop legacy QUADRANTS_CMAKE_ARGS by @hughperkins in #753
- [Nanobind] Silence interpreter-shutdown leak warnings by @hughperkins in #779
Full Changelog: v1.1.0...v1.1.1b2
v1.1.1b1
Pre-release v1.1.1b1
This pre-release is in order to test the nanobind warnings suppression for Genesis.
What's Changed
- [CI] Exempt advanced-section terms from doc-quality undefined-term rule by @hughperkins in #763
- [Docs] Use American spelling throughout user guide by @hughperkins in #765
- [CI] Add doc sequence check to doc quality check by @hughperkins in #764
- [Build] Use CMAKE_ARGS directly, drop legacy QUADRANTS_CMAKE_ARGS by @hughperkins in #753
- [Nanobind] Silence interpreter-shutdown leak warnings by @hughperkins in #779
Full Changelog: v1.1.0...v1.1.1b1