iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
22,547 Commits 3 Branches 0 Tags
eb68fb9105eab33888ce252001a269a8c52c19a5
Commit Graph
1085 Commits
Author SHA1 Message Date
MOVZXandHinata eeea036c12 [TopK] In-kernel canonical emission (MERLIN_TOPK_SORTED_EMIT) 
Emits the selected indices in canonical ascending order inside the kernel so
the Python QSA sort can be skipped. Output-equivalent to the Python sort;
reference model, property tests and GPU arms included. Default off in this
commit.

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>
 2026-10-09 20:04:37 +07:00
MOVZXandHinata 281491646d [Worker] E2-identity and rank-tagged e2-toggles startup lines 
Each worker logs its launch identity and, after warm-up and capture, one
e2-toggles line with the effective toggle state read back from the C++
extensions through two new tensor-free query ops. Both extensions must be
rebuilt.

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>
 2026-10-09 19:56:00 +07:00
MOVZXandHinata bef464a69d [TopK] Diagnostic-only build defines and qualification test cells 
Compile-time defines (default 0, device code unchanged by default) for the
delayed-publication stress, the pre-repair control and a forced-fault cell;
test cells for the OFF-arm controls, the QSA_SORT contract, FULL_K attribution
on the Marlin MoE GEMM and legacy-tail repeated-row membership.

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>
 2026-10-09 19:56:00 +07:00
MOVZXandHinata cff46af086 [TopK] Follow-up fixes to the persistent top-k kernels 
- Counter enum declared above the decode template (two-phase lookup).
- Emission-fault flag set atomically.
- Close the legacy large tie loop with a block barrier (inherited).
- Large-path fault injectors registered as hashed passthroughs.
- The tensor-free counter-slots op registered for CompositeExplicitAutograd.
- Tie measurements recorded only on rows neither fault verdict failed.
- Tests scoped to the properties the build promises.

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>
 2026-10-09 19:56:00 +07:00
MOVZXandHinata 2918e3fda1 [TopK] Determinism toggles for greedy repeatability 
- MERLIN_FULL_K: deterministic Marlin MoE split-K (after unmerged upstream #46639).
- MERLIN_QSA_SORT: canonical order for the QSA selected blocks.
- MERLIN_TIE_DET: deterministic tie-break at the persistent_topk cutoff.
- MERLIN_TIE_RECENT: cutoff ties keep the highest indices.
- Decode candidate truncation is detected; one tie direction across the
  medium path; the kernel switches are registered.

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>
 2026-10-09 19:56:00 +07:00
MOVZXandHinata 512ca2fc0b [TopK] Large-path hardening 
- Inline-PTX acquire/release helpers declare their memory effects.
- The fault verdict is aggregated across the CTA group before emission, so a
  faulted row is failed closed by every CTA.
- 64-bit device counters and an int64 reader (counter protocol v2).
- Tests: adversarial distributions, slot-residue coverage, an sm_90 gate for
  tests that need it.

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>
 2026-10-09 19:55:59 +07:00
MOVZXandHinata 63e2d8be20 [TopK] Deterministic placement and tie membership for the large-path radix_topk 
Above RADIX_THRESHOLD a row is split across a group of CTAs. The pivot was
already order-independent; cross-CTA placement, intra-CTA placement and tie
membership were arrival-order races. They are now deterministic behind
MERLIN_TOPK_DET_LARGE (default on).

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>
 2026-10-09 19:55:59 +07:00
MOVZXandHinata daa56356c8 [TopK] Barrier the histogram round before the leader publishes arrival 
In radix_topk's large-path round loop the leader published arrival with no
block barrier after the histogram merge, so a peer CTA could observe an
incomplete histogram. Inherited by earlier releases; fixed below the placement
work so both arms of any comparison carry it.

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>
 2026-10-09 19:55:59 +07:00
MOVZXandHinata 754fb3e29d [TopK] Exact restart of the medium path on candidate-buffer overflow 
When the persistent top-k medium-path candidate buffer overflows, the row is
restarted exactly instead of being truncated. Capture-safe counters with atomic
publication and a coherent snapshot record every restart.

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>
 2026-10-09 19:43:58 +07:00
AIwork4meAIwork4meJartX
 		4ac0d0eac2 [ROCm][RDNA3] Fix W4A16 split-K accuracy and determinism (#54706) 
Signed-off-by: AIwork4me <AIwork4me@users.noreply.github.com>
Co-authored-by: AIwork4me <AIwork4me@users.noreply.github.com>
Co-authored-by: JartX <sagformas@epdcenter.es>
 2026-10-03 23:44:17 +08:00
chowdmiaMiajan ChowdhuryRobert Shawmergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
 		6660926d92 [Bugfix][CPU] Fix macOS build on Apple Clang < 17: structured binding… (#58932) 
Signed-off-by: Miajan Chowdhury <chowdhurymiajan@gmail.com>
Co-authored-by: Miajan Chowdhury <chowdhurymiajan@gmail.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
 2026-10-02 10:01:07 +08:00
Yongye ZhuandClaude Opus 5.5 7566d83bd3 [Attention][MiniMax-M3] NVFP4 KV cache on the MSA sparse attention path (#59300) 
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-01 14:39:55 -07:00
almayneandLi, Jiang <jiang1.li@intel.com> 93dd80f636 [CPU] Conv1d optimised kernel for aarch64 (#54093) 
Signed-off-by: Anna Mayne <anna.mayne@arm.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>
 2026-09-30 20:23:42 +08:00
Rukhaiya2004andLi, Jiang <jiang1.li@intel.com> 72e7874fa6 [Hardware][Power] Enable W4A16 (AWQ & GPTQ) quantization on POWER10 using VSX (#59149) 
Signed-off-by: Rukhaiya <bibirukhaiya123@gmail.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>
 2026-09-30 19:44:48 +08:00
Rehan KhanandLi, Jiang <jiang1.li@intel.com> 583637f126 [CPU][Perf] Add vectorized Sampler Kernel (#53913) 
Signed-off-by: Rehan Khan <Rehan.Khan7@ibm.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>
 2026-09-30 15:16:28 +08:00
Matthew BonanniandClaude Opus 5.5 2798f66860 [Bugfix][HiSparse] Resolve MTP verification rows with a union residency kernel (#59235) 
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-30 03:01:45 +00:00
Yongye ZhuandClaude Opus 5.5 77871126f9 [Kernel][DSV4.1] Fuse MoE finalize into the TP all-reduce + mHC boundary (#58586) 
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-27 03:08:36 +08:00
MonishverClaude Fable 5.1Misha Goin
 		38cc054fc2 [Perf][Kernel] Vectorized flat abs-max for dynamic per-tensor FP8 quantization (#58194) 
Signed-off-by: Monishver Chandrasekaran <monishverchandrasekaran@gmail.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Co-authored-by: Misha Goin <mgoin64@gmail.com>
 2026-09-25 09:19:48 -06:00
chuan932andMisha Goin f550ad8bf7 [Kernel][Perf] Register-resident path for per-token-group 8-bit quant (#55330) 
Signed-off-by: chao.huan <chao.huan@nio.com>
Co-authored-by: Misha Goin <mgoin64@gmail.com>
 2026-09-25 08:11:42 -07:00
Yongye ZhuandSummer Yang 442bd00d9f [Perf][Distributed] Add low-SM multimem reduce-scatter for SM100/SM103 (#55072) 
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Signed-off-by: Summer Yang <girasoleyang@gmail.com>
Co-authored-by: Summer Yang <girasoleyang@gmail.com>
 2026-09-25 03:27:21 +00:00
Ganesh RCursorLi, Jiang <jiang1.li@intel.com>
 		f34a0e0797 [CPU] Gate the AVX10.2 paths on compiler support (#58133) 
Signed-off-by: R <Ganesh.R@amd.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>
 2026-09-24 18:07:53 +08:00
Ankit-Jaiswal-AMDandDouglas Lehr 88afb9dcac [CPU] Adds support for fp32 attention sinks (#56252) 
Signed-off-by: Ankit Jaiswal <ankit.jaiswal@amd.com>
Co-authored-by: Douglas Lehr <91553416+dllehr-amd@users.noreply.github.com>
 2026-09-23 08:55:09 -05:00
Gabriel WuandKimi Code c843f0cab7 [Bugfix][SM120][MLA] Support NoPE sparse MLA (GLM-5.3-Flash) on the FlashInfer SM120 backend (#55277) 
Signed-off-by: Zihua Wu <13583761+lucifer1004@users.noreply.github.com>
Co-authored-by: Kimi Code <noreply@moonshot.cn>
 2026-09-23 02:03:20 -07:00
Misha Goin 25965ee583 [Kernel] Remove AllSpark INT8 W8A16 GEMM backend (#58001) 
Signed-off-by: mgoin <mgoin64@gmail.com>
 2026-09-22 14:07:34 -07:00
Gilles TurpinandWentao Ye 74370a0e30 [Bugfix] hadacore_transform: respect inplace parameter to fix garbage outputs with QuIP transforms (#43462) 
Signed-off-by: Gilles Turpin <turpingilles15@gmail.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
 2026-09-22 12:48:20 -04:00
Woosuk KwonandCodex 9a70c233cd [Perf][DSV4.1] Fuse TP all-reduce with mHC input preparation (#57643) 
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
Co-authored-by: Codex <noreply@openai.com>
 2026-09-20 20:28:06 -07:00
Wentao Ye 4868312128 [Refactor] Remove dead kernel code (#57621) 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
 2026-09-20 10:39:23 -07:00
Jinzhen LinCodexmgoin
 		7d99c2c4fd [Feature][Humming] Humming feature integration (#56685) 
Signed-off-by: jinzhen.ljz <jinzhen.ljz@antgroup.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
 2026-09-20 08:11:01 -07:00
Fadi ArafehCodexLi, Jiang <jiang1.li@intel.com>
 		92b40f5a14 [CPU][Perf] refactor paged attention for Arm CPUs (#56045) 
Signed-off-by: Fadi Arafeh <fadi.arafeh@arm.com>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>
 2026-09-20 12:25:13 +08:00
Ariel BereslavskyAriel BereslavskyCodexMisha GoinYongye Zhumergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
 		a7fda4c88b [Bugfix] Fix SM100 fp8_ds_mla cache scales (#49435) 
Signed-off-by: Ariel Bereslavsky <arielb@getimpala.ai>
Signed-off-by: Ariel Bereslavsky <ariel@getimpala.ai>
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Ariel Bereslavsky <arielb@getimpala.ai>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Misha Goin <mgoin64@gmail.com>
Co-authored-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
 2026-09-19 17:15:07 -07:00
Yuwen ZhouandLi, Jiang <jiang1.li@intel.com> b346479065 [CPU] Add CPU FP8 W8A8 linear/MoE support (#49942) 
Signed-off-by: yuwenzho <yuwen.zhou@intel.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>
 2026-09-18 17:59:42 +08:00
Thien TranandCodex 8da75d6190 C3x SM100 FP8 blockwise - Pad activation scales to multiple of 4 (#57377) 
Signed-off-by: Thien Tran <gau.nernst@yahoo.com.sg>
Co-authored-by: Codex <noreply@openai.com>
 2026-09-18 08:39:27 +00:00
Rukhaiya2004Akash kaothalkarAntigravityLi, Jiang <jiang1.li@intel.com>
 		117cf43fb6 [HARDWARE][POWER] Enable W8A8 INT8 MoE on POWER (#55316) 
Signed-off-by: Rukhaiya <bibirukhaiya123@gmail.com>
Signed-off-by: Akash kaothalkar <akash.kaothalkar@ibm.com>
Co-authored-by: Akash kaothalkar <akash.kaothalkar@ibm.com>
Co-authored-by: Antigravity <antigravity@google.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>
 2026-09-17 21:45:13 +08:00
Wentao Ye 91a4c40b45 [Compile] Fix compile warning #177-D (#56609) 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
 2026-09-17 05:04:56 -07:00
Harry MellorandClaude Opus 5 8c1557a79c Add pydocstyle to the ruff rules (#52136) 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Signed-off-by: Harry Mellor <hej.mellor@gmail.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
 2026-09-16 16:44:29 +00:00
Yongye ZhuYongye ZhuClaude Opus 5Yunxiao NingAI Assistant
 		d6a1677d55 [Model][DSv4.1] FlashMLA mega attention and the NVFP4 compressed KV cache (#56935) 
Signed-off-by: Yongye Zhu <yongye@inferact.ai>
Signed-off-by: 云挚 <ningyunxiao.nyx@antgroup.com>
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Yongye Zhu <yongye@inferact.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Yunxiao Ning <73625538+foraxe@users.noreply.github.com>
Co-authored-by: AI Assistant <noreply@openai.com>
 2026-09-16 03:07:50 -07:00
0z5a0z5aCodexYongye ZhuClaude Opus 5
 		6ecd97f1b7 [Perf][DSpark] Add KV-only context insertion across V4.1 cache formats (#56441) 
Signed-off-by: 0z5a <0z5a@users.noreply.github.com>
Signed-off-by: Yongye Zhu <yongye@inferact.ai>
Co-authored-by: 0z5a <0z5a@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
 2026-09-15 21:34:55 -07:00
Misha GoinandOpenAI Codex c6fa1f05d1 [Perf][Kernel] Add sampled filtering for persistent top-k (#56346) 
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: OpenAI Codex <noreply@openai.com>
 2026-09-15 14:15:59 -07:00
Fangzhou Ai dffbb714e4 [ROCm][Perf] Optimize DSV4.1 K=512 decode top-k on gfx950 (#56743) 
Signed-off-by: fai <fangzhouai@gmail.com>
 2026-09-15 18:34:24 +00:00
Yongye ZhuYongye ZhuClaude Opus 5
 		e6eb0d120c [Model][DSv4.1] Store the whole KV in MXFP8 (FlashMLA V4.1 record) (#56893) 
Signed-off-by: Yongye Zhu <yongye@inferact.ai>
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Yongye Zhu <yongye@inferact.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
 2026-09-14 20:12:55 -07:00
sychen52 7702ee87db [Bugfix][DSA] Write nvfp4_ds_mla from the fused norm+rope kernel (#55538) 
Signed-off-by: Shiyang Chen <shiychen@nvidia.com>
 2026-09-15 00:25:33 +04:00
Matthias Gehre dc2e8f1157 [ROCm][Perf] W4A16: keep skinny GEMM zero-points packed 4-bit (#54965) 
Signed-off-by: Matthias Gehre <matthias.gehre@amd.com>
 2026-09-14 17:40:25 +08:00
Ayushman SinghandAyushman Singh c2ea9f1a5f [Kernel][Perf][Quantization] Fix odd-row performance cliff in per-token-group quantization (#56478) 
Signed-off-by: Ayushman Singh <ayush1399@users.noreply.github.com>
Co-authored-by: Ayushman Singh <ayush1399@users.noreply.github.com>
 2026-09-12 20:27:08 -07:00
Nicolò LucchesiLucas WilkinsonOpenAI Codex
 		29332cf936 [4/N] Expose HiSparse cache metrics via KV connector stats (#56061) 
Signed-off-by: NickLucche <nicolo.lucchesi@mistral.ai>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
 2026-09-12 10:47:55 -07:00
Matthew BonanniOpenAI CodexLucas WilkinsonLucas WilkinsonNicolò LucchesiOpenAI CodexClaude Opus 5Lucas Wilkinson
 		d43bb2f37f [3/N] HiSparse: host-resident sparse-MLA decode hot-buffering (#53781) 
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
Signed-off-by: Matthew Bonanni <matthew@neuralmagic.com>
Signed-off-by: Nicolò Lucchesi <nicolo.lucchesi@mistral.ai>
Co-authored-by: OpenAI Codex <codex@openai.com>
Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
Co-authored-by: Nicolò Lucchesi <nicolo.lucchesi@mistral.ai>
Co-authored-by: OpenAI Codex <noreply@openai.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
 2026-09-12 00:36:47 -04:00
Misha Goin 8a7f98c8c3 [Kernel][MoE] Optimize batched_moe_align_block_size with cooperative writes (#53280) 
Signed-off-by: mgoin <mgoin64@gmail.com>
 2026-09-11 12:05:22 -07:00
Wentao Ye 1e1060f998 [Kimi Perf] Group fp8 mla cahche insertion, 4~6x kernel level performance improvement for small batch (#55356) 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
 2026-09-11 08:16:57 -07:00
Fadi ArafehandLi, Jiang <jiang1.li@intel.com> e3f755b732 [CPU] Speedup LM Head on Arm CPUs (#55352) 
Signed-off-by: Fadi Arafeh <fadi.arafeh@arm.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>
 2026-09-11 16:21:14 +08:00
MattandClaude Opus 5 9163190dda [ROCm][Bugfix][Perf] Tune multi-stream shared experts use; wvSplitKrc fixes (#56098) 
Signed-off-by: Matthew Wong <Matthew.Wong2@amd.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
 2026-09-10 15:14:22 -07:00
kliuaeClaudeTJian
 		9521c60bdc [ROCm][Perf] Kimi-K3 Fused kernels for KDA prefill reland (#54038) 
Signed-off-by: kliuae <kuanfu.liu@embeddedllm.com>
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
Signed-off-by: kliuae <17350011+kliuae@users.noreply.github.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: TJian <tunjian.tan@embeddedllm.com>
 2026-09-10 15:37:54 +08:00