Lei WangandClaude Fable 5 7525499330 [CUDA][Cache] Publish immutable binary cache directories (#3177) ...

* [CUDA][Cache] Publish immutable binary cache directories

* [Example] Drop stale torch JIT build locks in minference example

The vertical_slash_index extension builds into a cache directory shared
across CI runs. torch's FileBaton lock file is not tied to the owning
process, so a run cancelled mid-compile leaves it behind and every later
load() polls on it forever (hung CI at test_vs_sparse_attn). Guard the
build with an OS-level flock, which dies with its holder, and remove any
leftover FileBaton lock before invoking torch.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

2026-09-08 02:46:47 +08:00

..

amd

[ROCm] Support wave64 Hadamard transforms (#3154)

2026-09-06 14:14:25 +08:00

analysis

[Analysis] Reject T.Parallel indexing of local buffers (#3041)

2026-08-17 14:31:41 +08:00

arith

[Compiler][Z3] Bump 3rdparty/tvm to materialize Z3 solvers lazily (#3105)

2026-08-29 00:51:12 +08:00

autotune

[Cache] Store cached kernel params as JSON instead of cloudpickle (#3143)

2026-09-06 16:06:20 +08:00

backend

[Backend] Add unified backend resolution policy (#2318)

2026-08-06 02:41:57 +08:00

cache

[CUDA][Cache] Publish immutable binary cache directories (#3177)

2026-09-08 02:46:47 +08:00

carver

[BugFix][Carver] Compare sm_version numerically in plan_rasterization (#3144)

2026-09-03 15:51:32 +08:00

components

[Language] Remove deprecated T.symbolic alias (#3122)

2026-09-04 23:56:52 +08:00

contrib

[BugFix] Isolate cross-compiler options per invocation (#2728)

2026-07-22 19:00:59 +08:00

cpu

[Analysis] Reject T.Parallel indexing of local buffers (#3041)

2026-08-17 14:31:41 +08:00

cuda

[BugFix][Hopper][Blackwell] Read the WGMMA/UMMA K-panel stride from the layout instead of the operand extent (#2965)

2026-09-02 18:13:38 +08:00

debug

[Debug][TIR][JIT] Unify pass instrumentation per compilation (#2923)

2026-08-10 15:37:25 +08:00

fastmath

[BugFix] Fix failed ties-away-from-zero round compile on bfloat16/float8 (#2873)

2026-08-10 16:44:34 +08:00

issue

[Cache] Store cached kernel params as JSON instead of cloudpickle (#3143)

2026-09-06 16:06:20 +08:00

jit

[Runtime][Cache] Export libraries after disk-cache hits (#3116)

2026-08-31 19:22:37 +08:00

kernel

[BugFix][Hopper][Blackwell] Read the WGMMA/UMMA K-panel stride from the layout instead of the operand extent (#2965)

2026-09-02 18:13:38 +08:00

language

[BugFix][Transform] Keep shared FP32 atomics scalar on SM90 (#3129)

2026-09-07 11:37:13 +08:00

layout

[CUDA] Unify TMA copy lowering on CuTe algebra (#3106)

2026-08-29 15:43:29 +08:00

llvm

[Analysis] Reject T.Parallel indexing of local buffers (#3041)

2026-08-17 14:31:41 +08:00

math

[BugFix][CUDA] Provide htan overloads for fp16/bf16 tangent (#2894)

2026-08-11 15:22:06 +08:00

metal

[BugFix][Metal] Preserve pointer address spaces for byte offsets (#2925)

2026-08-14 16:54:58 +08:00

profiler

[tilelang] Fix CUPTI cache flush filtering (#2244)

2026-05-22 16:15:53 +08:00

quantize

[CUDA] Add SM120 NVF4 block-scale MMA support (#2364)

2026-07-30 18:10:10 +08:00

runtime

[Language] Remove deprecated T.symbolic alias (#3122)

2026-09-04 23:56:52 +08:00

target

[Backend] Add unified backend resolution policy (#2318)

2026-08-06 02:41:57 +08:00

tilelibrary

[Refactor] Move example-only helpers out of tilelang package (#2761)

2026-07-24 18:03:34 +08:00

tools

[Tool] Add official compile-only CLI (#3045)

2026-08-25 14:11:20 +08:00

transform

[Layout] Consider scalar reducer plans in register-count search (#3171)

2026-09-07 13:32:20 +08:00

utils

[Refactor] Move example-only helpers out of tilelang package (#2761)

2026-07-24 18:03:34 +08:00

webgpu

[Enhancement] Update examples and tests for improved type handling functionality (#1448)

2025-12-17 14:41:25 +08:00

test_version_provider.py

[Build] Fix sdist version metadata (#2374)

2026-06-11 20:40:16 +08:00