Quadrants 1.0.0 - #41

Merged

npoulad1 merged 181 commits into

amd-integrationfrom

npoulad/quadrants-1.0.0-rebase

Jun 8, 2026

Commits on Mar 26, 2026

[Misc] Warn user to disable caching when print_ir/QD_DUMP_IR enabled (#425)

v01dXYZ
and
v01dxyz
authored
0fc25c9

Commits on Mar 29, 2026

[Build] Pin torch version to CUDA 12.8 for CUDA tests (#428)
hughperkins
authored
27c34a2
[Misc] Fixing up taichi-dev urls (#429)
hughperkins
authored
412108b
[Perf] Rename cuda_graph to gpu_graph across the codebase (#430)
hughperkins
authored
55cbdd3

Commits on Mar 30, 2026

Misc: fix typo integeral -> integral (#434)

v01dXYZ
and
v01dxyz
authored
1dae239

Commits on Mar 31, 2026

Commits on Apr 1, 2026

Commits on Apr 2, 2026

[Misc] Allow disabling GPU graph via QD_GPU_GRAPH=0 env var (#439)
hughperkins
authored
f2fc6db
[Misc] Add named top-level loops (#440)
hughperkins
authored
8b8f12c
[Misc] Rename gpu_graph to graph (#446)
hughperkins
authored
a451902

Commits on Apr 8, 2026

Commits on Apr 9, 2026

Commits on Apr 10, 2026

[Misc] CUDA graph 5 Add fatbin (#464)
hughperkins
authored
83e4b3b
[Bug] Reuse VkInstance across init/reset cycles (#465)
hughperkins
authored
f81ecab

Commits on Apr 12, 2026

Commits on Apr 13, 2026

Commits on Apr 14, 2026

[Build] Fix flaky test_clock_accuracy (#436)
hughperkins
authored
6ac8d2e
Fix AARCH64 emitting invalid asm in CUDA kernels. (#473)

duburcqa
and
hughperkins
authored
5e8efa6

Commits on Apr 15, 2026

Commits on Apr 17, 2026

Commits on Apr 20, 2026

Commits on Apr 21, 2026

Commits on Apr 22, 2026

Commits on Apr 23, 2026

Commits on Apr 24, 2026

Commits on Apr 25, 2026

Commits on Apr 26, 2026

[CI] CI wrap check feeds only diffs to agent (#567)
hughperkins
authored
f6c231a

Commits on Apr 27, 2026

Commits on Apr 28, 2026

Commits on Apr 29, 2026

Commits on Apr 30, 2026

[SPIR-V] Shrink reverse-grad kernel MSL by ~50% (#591)
duburcqa
authored
da5e8e0
[CI] Add CI check that PR changes have test coverage (#596)
hughperkins
authored
2520e13
[Perf] Enable zero-copy in to_torch() and to_numpy() (#450)
hughperkins
authored
7f4632b
Add BufferView: safe sub-range ndarray access for kernels (#585)

alanray-tech
and
hughperkins
authored
4d98ceb
[Doc] Add user-facing fastcache documentation (#597)

hughperkins
and
hugh
authored
84b0636
[Misc] Upgrade to enable v1 dlpack so to_numpy(copy=False) writable (#598)

hughperkins
and
root
authored
8d20beb

Commits on May 1, 2026

[AutoDiff] Cut reverse-mode adstack memory usage 10x on all backends (#599)
duburcqa
authored
cda2944

Commits on May 2, 2026

Commits on May 4, 2026

[Misc] Forward copy= through qd.Tensor, add copy=None option (#616)

hughperkins
and
cursoragent
authored
b5fb5db
[Doc] Update README (#617)

hughperkins
and
cursoragent
authored
95520fe
[CI] Fix coverage report showing def lines as uncovered (#623)

hughperkins
and
cursoragent
authored
5c6872c
[Perf] Generic launcher: persistent context, JIT-pointer reuse, Metal compute encoder, LLVM-GPU async memory ops (Part 1/2) (#619)
duburcqa
authored
4e35bd5
[CI] Encode Python-first testing policy in coverage-check prompt (#622)

hughperkins
and
cursoragent
authored
ab20656
[CI] Add PR Line change report (#624)

hughperkins
and
cursoragent
authored
d16692c

Commits on May 5, 2026

Commits on May 6, 2026

[AutoDiff] Adstack chain-clone safety: mutated-SNode leaf reject + load_top consumer-aware guard (#634)
duburcqa
authored
5884fad

Commits on May 9, 2026

Commits on May 10, 2026

Commits on May 12, 2026

[AMDGPU] Use syncscope("agent") for atomix xor to avoid CAS livelock (#672)
hughperkins
authored
60391b8
[GPU] New bit ops for QIPC (#679)
hughperkins
authored
8dbad65
[GPU] Subgroup ops cross-gpu (#665)
hughperkins
authored
d08b0f1

Commits on May 13, 2026

Commits on May 14, 2026

Pass dataclass sub-structs into qd.func (#698)
hughperkins
authored
1b32142

Commits on May 15, 2026

[AMDGPU] HIP graph runtime support for @qd.kernel(graph=True) (#692)
hughperkins
authored
bbc5bd4
[CI] Add per-file timing report to Mac Metal test job (#695)

hughperkins
and
cursoragent
authored
2cfe475
[CI] Enable kernel disk cache during tests (#696)
hughperkins
authored
a5b9b68
[Math] New QIPC ops for single-threaded linalg (#683)
hughperkins
authored
188cd49
[BREAKING][GPU] New QIPC ops for subgroups (#676)
hughperkins
authored
26f0c72
[GPU] New QIPC ops for block (#684)
hughperkins
authored
a22cc2d

Commits on May 18, 2026

Commits on May 19, 2026

[Vulkan] Declare GroupNonUniform SPIR-V caps and enable shaderSubgroupExtendedTypes (#707)
hughperkins
authored
5d68fc4

Commits on Jun 5, 2026

Merge upstream Quadrants v1.0.0 (5d68fc4) into amd-integration

npoulad1
and
cursoragent
committed
ba8a481
Fix duplicate HIP graph driver-function declarations after v1.0.0 merge

npoulad1
and
cursoragent
committed
c7a1da0
Fix AMDGPU launcher coherence and num_instructions visibility after v1.0.0 merge

npoulad1
and
cursoragent
committed
8cc6c18
Restore async result D2H and hoist kernarg vectors in AMDGPU launcher

npoulad1
and
cursoragent
committed
2281dbc

Commits on Jun 6, 2026

amdgpu: default to LDS permlane64 emulation; drop host-x86 barrier asm on retarget

npoulad1
and
cursoragent
committed
cec7bc8

Commits on Jun 8, 2026

style: apply clang-format (v19.1.7) to AMDGPU fn_attrs and launcher sources

npoulad1
and
cursoragent
committed
c4d06c6
fix(amdgpu): use CreateNeg for branchless i32 sgn instead of CreateSub(0, input)

npoulad1
and
cursoragent
committed
7874fc8