Skip to content

v1.4.0

Latest

Choose a tag to compare

@hughperkins hughperkins released this 04 Nov 13:39
· 19 commits to main since this release
b2f658e

Release v1.4.0b3

New:

  • support for python 3.14
  • fill= parameter for ndarray, field, tensor, during allocation
  • added qd.block_idx()
  • add cpu_per_worker_min_block_dim for cpu range loops

Backwards incompatible changes:

  • removed SourceBuilder
  • removed Python backend
  • removed support for Python 3.10 and 3.11
  • dropped support for SPIRV < 1.5

What's Changed

CI

Misc

CPU

  • [CPU] Clarify runtime task and loop-body names by @hughperkins in #951
  • [CPU] Configurable scheduling: revamp naming and add cpu_per_worker_min_block_dim by @hughperkins in #933

AMDGPU

  • [AMDGPU] Fix direct HIP tensor access and require HIP 6+ by @hughperkins in #955
    AI/quadrants/pull/1034
  • [AMDGPU] Cache AMD runtime HSACO bytes across resets by @hughperkins in #1023
  • [AMDGPU] Remove OpenCL and empty DAZ bitcode inputs; select ABI 600 by @hughperkins in #1047
  • [AMDGPU] Centralize device library selection and ABI configuration in CMake by @hughperkins in #1049

SPIRV

  • [SPIRV] Clarify SPIR-V module assembly and separate header storage by @hughperkins in #960
  • [SPIRV] Separate operations from IR builder tooling by @hughperkins in #961
  • [SPIRV] Extend block_idx to Vulkan and Metal through GLSL helpers by @hughperkins in #952
  • [SPIRV] Require SPIR-V 1.3 and Vulkan 1.1 by @hughperkins in #964
  • [SPIRV] Use SPIR-V 1.5 for Metal and require a 1.5 minimum by @hughperkins in #965
  • [SPIRV] Order ndarray host writes through the graphics command queue by @hughperkins in #1046

Vulkan

Docs

Tensor

  • [Tensor] Fix layout permutation for list tensor indices by @hughperkins in #989
  • [Tensor] Move test_tensor_pr560_fixes.py tests into other test files by @hughperkins in #996
  • [Tensor] Support shape-axis layouts for vector and matrix tensors by @hughperkins in #991

Ndarray

Build

Lang

Tests

Field

Graph

  • [Graph] CUDA: upload per-launch graph args with async HtoD on the launch stream by @arcusbuilds in #950

Template

  • [Template] Fix qd.Template function annotations and migrate template tests by @hughperkins in #1055

GPU

  • [GPU] Align allocation streams and replace exhaustion test with mocks by @hughperkins in #1042

New Contributors

Full Changelog: v1.4.0b3...v1.4.0