[DSpark] Remove overlap scheduling host syncs - #27

Closed

tanth47 wants to merge 745 commits into

mainfrom

codex/glm52-dspark-zero-overhead-upstream

This pull request is big! We're only showing the most recent 250 commits

Commits on Jul 18, 2026

Remove QServe and FBGEMM FP8 quantization (#31109)
b8zhong
authored
238b2b2
ci: add MLX to coverage report backend display order (#31442)
alisonshao
authored
e48eabb
[JIT] Refactor dtype traits into DTypeTrait and unify warp reductions (#30838)

authored
67e7f8d
[Bugfix][NPU] Fix/Refactor routed scaling factor application in MoE routing (#31449)
xdtbynd
authored
359009f
[MUSA] Fix sglang-kernel build (#31634)
yeahdongcn
authored
87dc211
[Diffusion] Use SGLang server for ERNIE-Image prompt enhancement (#31354)

authored
44e4999
Let CI server launches wait longer for ports held by a dying predecessor (#31281)
alisonshao
authored
a5c0b94
[plugin] oot torch profiler activity support (#31580)

DevashishLal-CB
and
Devashish Lal
authored
ab3d421
[Fix] Account zero-logprob sequences correctly in chunked logprob stitching (#31639)
hnyls2002
authored
19c53c4
[Refactor] Move output logprob processing into the logprob_processor layer (#31624)
hnyls2002
authored
639261f
[CP] Migrate MLA prefill CP (DeepSeek V3) to CP-v2 zigzag strategy (#31619)

kpham-sgl
and
claude
authored
7a89621
[AMD] register 8 JIT kernel benchmarks to jit-kernel-benchmark-test-amd (#31492)
michaelzhang-ai
authored
7fbe91c
[Spec] DFlash: remove per-step host syncs so the CPU runs a full step ahead (spec-v2 overlap) (#31468)

thanhhao98
and
Hao Phan
authored
72c4ed1
fix(rpc) Synchronize RPC requests only within the TP group. (#25213)

twb1235
and
xiezhq-hermann
authored
071e649
Fix SM120 NVFP4 KV cache test OOM (#31653)
Fridge003
authored
38b29dc
perf: avoid excessive prefill CUDA graph padding (#31487)
mickqian
authored
6c6175f
Implement SM120 DeepSeek V4 flashinfer_mxfp4 moe runner backend + TP2 (#30272)
lucifer1004
authored
faf6894
CI: synchronize prefill graph test fixtures (#31665)
mickqian
authored
573c075
vlm: batch cross-request vit encoding and reuse attention metadata (#24013)

authored
9306278
[Kernel] Sweep decoupled scattered kernels into sglang.kernels.ops (RFC #29630) (#31582)

BBuf
and
claude
authored
216b750
[NPU] fix: skip Triton embedding kernel on NPU to avoid kernel launch failure (#31636)
chx96642264
authored
d7b9425
Reset only the used mamba state on unified radix cache (#31648)
ispobock
authored
48ae829
[NPU] FIX CMB illusion of garbled characters acc problems, in prefix cache mtp scenarios. (#31659)
McZyWu
authored
ece02ff
Reset only the used mamba state on radix cache hit (#31643)
ispobock
authored
5609f8e
metrics: allow extra labels on HTTP request/response Prometheus metrics (#31530)
sshleifer
authored
d86ae51
model_runner: extract post-memory-pool wiring into _init_post_memory (#31601)
sshleifer
authored
b3a0185
[Fix] Respect cache_protected_len in ChunkCache and disabled-radix release paths (#31662)
hnyls2002
authored
10908a6
[flashinfer] Pass window_left at plan time for the SWA paged prefill wrapper (#31501)
luccafong
authored
99f5a6f

Commits on Jul 19, 2026

Commits on Jul 20, 2026

Commits on Jul 21, 2026

Commits on Jul 22, 2026

Commits on Jul 23, 2026

Commits on Jul 24, 2026