Issue Station
Sign In
Tools hereToolbox

vllm - ✅(Solved) Fix [Model Runner V2][Bug]: The _gumbel_sample_kernel exhibits poor performance on H800. [1 pull requests, 1 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
vllm-project/vllm#40755•Fetched 2026-04-24 10:36:28
View on GitHub
Comments
0
Participants
1
Timeline
3
Reactions
0
Author
lio1226lio1226
Participants
lio1226lio1226
Timeline (top)
labeled ×1mentioned ×1subscribed ×1

Root Cause

I have tested the performance of Model Runner V2 with Eagle3 enabled on H800. The result is inferior to that of Model Runner V1. Current analysis identifies that the excessive latency of the _gumbel_sample_kernel is the root cause, as shown in the figure below. <img width="2484" height="1034" alt="Image" src="https://github.com/user-attachments/assets/0e2227a4-40e6-4e40-a501-beffc2b7a083" />

Fix Action

Fix / Workaround

============================== CPU Info

Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Address sizes: 52 bits physical, 57 bits virtual Byte Order: Little Endian CPU(s): 192 On-line CPU(s) list: 0-191 Vendor ID: GenuineIntel Model name: Intel(R) Xeon(R) Platinum 8468 CPU family: 6 Model: 143 Thread(s) per core: 2 Core(s) per socket: 48 Socket(s): 2 Stepping: 8 CPU max MHz: 3800.0000 CPU min MHz: 800.0000 BogoMIPS: 4200.00 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cat_l2 cdp_l3 invpcid_single intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities Virtualization: VT-x L1d cache: 4.5 MiB (96 instances) L1i cache: 3 MiB (96 instances) L2 cache: 192 MiB (96 instances) L3 cache: 210 MiB (2 instances) NUMA node(s): 2 NUMA node0 CPU(s): 0-47,96-143 NUMA node1 CPU(s): 48-95,144-191 Vulnerability Gather data sampling: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling: Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization Vulnerability Spectre v2: Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S Vulnerability Srbds: Not affected Vulnerability Tsx async abort: Not affected

PR fix notes

PR #34854: [Model Runner V2] Use FP32 for Gumbel Noise

Description (problem / solution / changelog)

  • Use FP32 for gumbel noise
  • Merge tl.max and tl.argmax into one tl.max(..., return_indices=True)

Up to 50% performance improvement for large batch sizes:


if __name__ == "__main__":
    import time
    BATCH_SIZES = [1, 1024]
    VOCAB_SIZES = [200 * 1000]
    for batch_size in BATCH_SIZES:
        for vocab_size in VOCAB_SIZES:
            logits = torch.randn(batch_size, vocab_size, dtype=torch.float32, device="cuda")
            idx_mapping = torch.randint(0, batch_size, (batch_size,), dtype=torch.int32, device="cuda")
            temperature = torch.rand(batch_size, dtype=torch.float32, device="cuda")
            temperature.zero_()
            seed = torch.randint(0, 1000000, (batch_size,), dtype=torch.int64, device="cuda")
            pos = torch.randint(0, 100, (batch_size,), dtype=torch.int64, device="cuda")
            sampled = gumbel_sample(logits, idx_mapping, temperature, seed, pos, apply_temperature=False)

            torch.cuda.synchronize()

            start_time = time.perf_counter()
            for _ in range(1000):
                sampled = gumbel_sample(logits, idx_mapping, temperature, seed, pos, apply_temperature=False)
            torch.cuda.synchronize()
            end_time = time.perf_counter()
            print(f"Batch size: {batch_size}, Vocab size: {vocab_size}, Time: {end_time - start_time} ms")

Changed files

  • vllm/v1/worker/gpu/sample/gumbel.py (modified, +8/-9)

Code Example

Collecting environment information...
==============================
        System Info
==============================
OS                           : Ubuntu 22.04.5 LTS (x86_64)
GCC version                  : (Ubuntu 11.4.0-1ubuntu1~22.04.3) 11.4.0
Clang version                : Could not collect
CMake version                : Could not collect
Libc version                 : glibc-2.35

==============================
       PyTorch Info
==============================
PyTorch version              : 2.11.0+cu130
Is debug build               : False
CUDA used to build PyTorch   : 13.0
ROCM used to build PyTorch   : N/A
XPU used to build PyTorch    : N/A

==============================
      Python Environment
==============================
Python version               : 3.12.13 (main, Mar  4 2026, 09:23:07) [GCC 11.4.0] (64-bit runtime)
Python platform              : Linux-5.15.0-119-generic-x86_64-with-glibc2.35
    
==============================
       CUDA / GPU Info
==============================
Is CUDA available            : True
CUDA runtime version         : 13.0.88
CUDA_MODULE_LOADING set to   : 
GPU models and configuration : 
GPU 0: NVIDIA H800
GPU 1: NVIDIA H800
GPU 2: NVIDIA H800
GPU 3: NVIDIA H800
GPU 4: NVIDIA H800
GPU 5: NVIDIA H800
GPU 6: NVIDIA H800
GPU 7: NVIDIA H800

Nvidia driver version        : 580.95.05
cuDNN version                : Could not collect
HIP runtime version          : N/A
MIOpen runtime version       : N/A
Is XNNPACK available         : True

==============================
          CPU Info
==============================
Architecture:                         x86_64
CPU op-mode(s):                       32-bit, 64-bit
Address sizes:                        52 bits physical, 57 bits virtual
Byte Order:                           Little Endian
CPU(s):                               192
On-line CPU(s) list:                  0-191
Vendor ID:                            GenuineIntel
Model name:                           Intel(R) Xeon(R) Platinum 8468
CPU family:                           6
Model:                                143
Thread(s) per core:                   2
Core(s) per socket:                   48
Socket(s):                            2
Stepping:                             8
CPU max MHz:                          3800.0000
CPU min MHz:                          800.0000
BogoMIPS:                             4200.00
Flags:                                fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cat_l2 cdp_l3 invpcid_single intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities
Virtualization:                       VT-x
L1d cache:                            4.5 MiB (96 instances)
L1i cache:                            3 MiB (96 instances)
L2 cache:                             192 MiB (96 instances)
L3 cache:                             210 MiB (2 instances)
NUMA node(s):                         2
NUMA node0 CPU(s):                    0-47,96-143
NUMA node1 CPU(s):                    48-95,144-191
Vulnerability Gather data sampling:   Not affected
Vulnerability Itlb multihit:          Not affected
Vulnerability L1tf:                   Not affected
Vulnerability Mds:                    Not affected
Vulnerability Meltdown:               Not affected
Vulnerability Mmio stale data:        Not affected
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed:               Not affected
Vulnerability Spec rstack overflow:   Not affected
Vulnerability Spec store bypass:      Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1:             Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:             Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                  Not affected
Vulnerability Tsx async abort:        Not affected

==============================
Versions of relevant libraries
==============================
[pip3] flashinfer-python==0.6.7
[pip3] numpy==2.2.6
[pip3] nvidia-cublas==13.1.0.3
[pip3] nvidia-cuda-cupti==13.0.85
[pip3] nvidia-cuda-nvrtc==13.0.88
[pip3] nvidia-cuda-runtime==13.0.96
[pip3] nvidia-cudnn-cu13==9.19.0.56
[pip3] nvidia-cudnn-frontend==1.18.0
[pip3] nvidia-cufft==12.0.0.61
[pip3] nvidia-cufile==1.15.1.6
[pip3] nvidia-curand==10.4.0.35
[pip3] nvidia-cusolver==12.0.4.66
[pip3] nvidia-cusparse==12.6.3.3
[pip3] nvidia-cusparselt-cu13==0.8.0
[pip3] nvidia-cutlass-dsl==4.4.2
[pip3] nvidia-cutlass-dsl-libs-base==4.4.2
[pip3] nvidia-ml-py==13.595.45
[pip3] nvidia-nccl-cu13==2.28.9
[pip3] nvidia-nvjitlink==13.0.88
[pip3] nvidia-nvshmem-cu13==3.4.5
[pip3] nvidia-nvtx==13.0.85
[pip3] pyzmq==27.1.0
[pip3] torch==2.11.0+cu130
[pip3] torch_c_dlpack_ext==0.1.5
[pip3] torchaudio==2.11.0+cu130
[pip3] torchvision==0.26.0+cu130
[pip3] transformers==5.5.4
[pip3] triton==3.6.0
[conda] Could not collect

==============================
         vLLM Info
==============================
ROCM Version                 : Could not collect
vLLM Version                 : 0.19.1rc1.dev328+g18013df6a (git sha: 18013df6a)
vLLM Build Flags:
  CUDA Archs: 7.0 7.5 8.0 8.9 9.0 10.0 12.0; ROCm: Disabled; XPU: Disabled
GPU Topology:
  	GPU0	GPU1	GPU2	GPU3	GPU4	GPU5	GPU6	GPU7	NIC0	NIC1	NIC2	NIC3	NIC4	NIC5	NIC6	NIC7	CPU Affinity	NUMA Affinity	GPU NUMA ID
GPU0	 X 	NV8	NV8	NV8	NV8	NV8	NV8	NV8	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	0-47,96-143	0		N/A
GPU1	NV8	 X 	NV8	NV8	NV8	NV8	NV8	NV8	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	0-47,96-143	0		N/A
GPU2	NV8	NV8	 X 	NV8	NV8	NV8	NV8	NV8	PIX	PIX	NODE	NODE	SYS	SYS	SYS	SYS	0-47,96-143	0		N/A
GPU3	NV8	NV8	NV8	 X 	NV8	NV8	NV8	NV8	PXB	PXB	NODE	NODE	SYS	SYS	SYS	SYS	0-47,96-143	0		N/A
GPU4	NV8	NV8	NV8	NV8	 X 	NV8	NV8	NV8	SYS	SYS	SYS	SYS	PIX	PIX	NODE	NODE	48-95,144-191	1		N/A
GPU5	NV8	NV8	NV8	NV8	NV8	 X 	NV8	NV8	SYS	SYS	SYS	SYS	PXB	PXB	NODE	NODE	48-95,144-191	1		N/A
GPU6	NV8	NV8	NV8	NV8	NV8	NV8	 X 	NV8	SYS	SYS	SYS	SYS	NODE	NODE	NODE	NODE	48-95,144-191	1		N/A
GPU7	NV8	NV8	NV8	NV8	NV8	NV8	NV8	 X 	SYS	SYS	SYS	SYS	NODE	NODE	NODE	NODE	48-95,144-191	1		N/A
NIC0	NODE	NODE	PIX	PXB	SYS	SYS	SYS	SYS	 X 	PIX	NODE	NODE	SYS	SYS	SYS	SYS				
NIC1	NODE	NODE	PIX	PXB	SYS	SYS	SYS	SYS	PIX	 X 	NODE	NODE	SYS	SYS	SYS	SYS				
NIC2	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	NODE	NODE	 X 	PIX	SYS	SYS	SYS	SYS				
NIC3	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	NODE	NODE	PIX	 X 	SYS	SYS	SYS	SYS				
NIC4	SYS	SYS	SYS	SYS	PIX	PXB	NODE	NODE	SYS	SYS	SYS	SYS	 X 	PIX	NODE	NODE				
NIC5	SYS	SYS	SYS	SYS	PIX	PXB	NODE	NODE	SYS	SYS	SYS	SYS	PIX	 X 	NODE	NODE				
NIC6	SYS	SYS	SYS	SYS	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	NODE	NODE	 X 	PIX				
NIC7	SYS	SYS	SYS	SYS	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	NODE	NODE	PIX	 X 				

Legend:

  X    = Self
  SYS  = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
  NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
  PHB  = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
  PXB  = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
  PIX  = Connection traversing at most a single PCIe bridge
  NV#  = Connection traversing a bonded set of # NVLinks

NIC Legend:

  NIC0: mlx5_0
  NIC1: mlx5_1
  NIC2: mlx5_2
  NIC3: mlx5_3
  NIC4: mlx5_4
  NIC5: mlx5_5
  NIC6: mlx5_6
  NIC7: mlx5_7

==============================
     Environment Variables
==============================
NVIDIA_VISIBLE_DEVICES=GPU-750f387f-40be-cef6-531e-8c754db35274
NVIDIA_REQUIRE_CUDA=cuda>=13.0 brand=unknown,driver>=535,driver<536 brand=grid,driver>=535,driver<536 brand=tesla,driver>=535,driver<536 brand=nvidia,driver>=535,driver<536 brand=quadro,driver>=535,driver<536 brand=quadrortx,driver>=535,driver<536 brand=nvidiartx,driver>=535,driver<536 brand=vapps,driver>=535,driver<536 brand=vpc,driver>=535,driver<536 brand=vcs,driver>=535,driver<536 brand=vws,driver>=535,driver<536 brand=cloudgaming,driver>=535,driver<536 brand=unknown,driver>=550,driver<551 brand=grid,driver>=550,driver<551 brand=tesla,driver>=550,driver<551 brand=nvidia,driver>=550,driver<551 brand=quadro,driver>=550,driver<551 brand=quadrortx,driver>=550,driver<551 brand=nvidiartx,driver>=550,driver<551 brand=vapps,driver>=550,driver<551 brand=vpc,driver>=550,driver<551 brand=vcs,driver>=550,driver<551 brand=vws,driver>=550,driver<551 brand=cloudgaming,driver>=550,driver<551 brand=unknown,driver>=565,driver<566 brand=grid,driver>=565,driver<566 brand=tesla,driver>=565,driver<566 brand=nvidia,driver>=565,driver<566 brand=quadro,driver>=565,driver<566 brand=quadrortx,driver>=565,driver<566 brand=nvidiartx,driver>=565,driver<566 brand=vapps,driver>=565,driver<566 brand=vpc,driver>=565,driver<566 brand=vcs,driver>=565,driver<566 brand=vws,driver>=565,driver<566 brand=cloudgaming,driver>=565,driver<566 brand=unknown,driver>=570,driver<571 brand=grid,driver>=570,driver<571 brand=tesla,driver>=570,driver<571 brand=nvidia,driver>=570,driver<571 brand=quadro,driver>=570,driver<571 brand=quadrortx,driver>=570,driver<571 brand=nvidiartx,driver>=570,driver<571 brand=vapps,driver>=570,driver<571 brand=vpc,driver>=570,driver<571 brand=vcs,driver>=570,driver<571 brand=vws,driver>=570,driver<571 brand=cloudgaming,driver>=570,driver<571 brand=unknown,driver>=575,driver<576 brand=grid,driver>=575,driver<576 brand=tesla,driver>=575,driver<576 brand=nvidia,driver>=575,driver<576 brand=quadro,driver>=575,driver<576 brand=quadrortx,driver>=575,driver<576 brand=nvidiartx,driver>=575,driver<576 brand=vapps,driver>=575,driver<576 brand=vpc,driver>=575,driver<576 brand=vcs,driver>=575,driver<576 brand=vws,driver>=575,driver<576 brand=cloudgaming,driver>=575,driver<576
TORCH_CUDA_ARCH_LIST=7.0 7.5 8.0 8.9 9.0 10.0 12.0
CUDA_DEVICE_SM_LIMIT=0
NVIDIA_DRIVER_CAPABILITIES=compute,utility
VLLM_USAGE_SOURCE=production-docker-image
CUDA_VERSION=13.0.1
VLLM_ENABLE_CUDA_COMPATIBILITY=0
CUDA_DEVICE_MEMORY_LIMIT_0=81559m
CUDA_DEVICE_MEMORY_SHARED_CACHE=/usr/local/vgpu/3f21f157-aeb6-49e0-9f8e-a3bad69b5407.cache
LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib64
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
TORCHINDUCTOR_CACHE_DIR=/tmp/torchinductor_root
RAW_BUFFERClick to expand / collapse

Your current environment

<details> <summary>The output of <code>python collect_env.py</code></summary> 
Collecting environment information...
==============================
        System Info
==============================
OS                           : Ubuntu 22.04.5 LTS (x86_64)
GCC version                  : (Ubuntu 11.4.0-1ubuntu1~22.04.3) 11.4.0
Clang version                : Could not collect
CMake version                : Could not collect
Libc version                 : glibc-2.35

==============================
       PyTorch Info
==============================
PyTorch version              : 2.11.0+cu130
Is debug build               : False
CUDA used to build PyTorch   : 13.0
ROCM used to build PyTorch   : N/A
XPU used to build PyTorch    : N/A

==============================
      Python Environment
==============================
Python version               : 3.12.13 (main, Mar  4 2026, 09:23:07) [GCC 11.4.0] (64-bit runtime)
Python platform              : Linux-5.15.0-119-generic-x86_64-with-glibc2.35
    
==============================
       CUDA / GPU Info
==============================
Is CUDA available            : True
CUDA runtime version         : 13.0.88
CUDA_MODULE_LOADING set to   : 
GPU models and configuration : 
GPU 0: NVIDIA H800
GPU 1: NVIDIA H800
GPU 2: NVIDIA H800
GPU 3: NVIDIA H800
GPU 4: NVIDIA H800
GPU 5: NVIDIA H800
GPU 6: NVIDIA H800
GPU 7: NVIDIA H800

Nvidia driver version        : 580.95.05
cuDNN version                : Could not collect
HIP runtime version          : N/A
MIOpen runtime version       : N/A
Is XNNPACK available         : True

==============================
          CPU Info
==============================
Architecture:                         x86_64
CPU op-mode(s):                       32-bit, 64-bit
Address sizes:                        52 bits physical, 57 bits virtual
Byte Order:                           Little Endian
CPU(s):                               192
On-line CPU(s) list:                  0-191
Vendor ID:                            GenuineIntel
Model name:                           Intel(R) Xeon(R) Platinum 8468
CPU family:                           6
Model:                                143
Thread(s) per core:                   2
Core(s) per socket:                   48
Socket(s):                            2
Stepping:                             8
CPU max MHz:                          3800.0000
CPU min MHz:                          800.0000
BogoMIPS:                             4200.00
Flags:                                fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cat_l2 cdp_l3 invpcid_single intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities
Virtualization:                       VT-x
L1d cache:                            4.5 MiB (96 instances)
L1i cache:                            3 MiB (96 instances)
L2 cache:                             192 MiB (96 instances)
L3 cache:                             210 MiB (2 instances)
NUMA node(s):                         2
NUMA node0 CPU(s):                    0-47,96-143
NUMA node1 CPU(s):                    48-95,144-191
Vulnerability Gather data sampling:   Not affected
Vulnerability Itlb multihit:          Not affected
Vulnerability L1tf:                   Not affected
Vulnerability Mds:                    Not affected
Vulnerability Meltdown:               Not affected
Vulnerability Mmio stale data:        Not affected
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed:               Not affected
Vulnerability Spec rstack overflow:   Not affected
Vulnerability Spec store bypass:      Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1:             Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:             Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                  Not affected
Vulnerability Tsx async abort:        Not affected

==============================
Versions of relevant libraries
==============================
[pip3] flashinfer-python==0.6.7
[pip3] numpy==2.2.6
[pip3] nvidia-cublas==13.1.0.3
[pip3] nvidia-cuda-cupti==13.0.85
[pip3] nvidia-cuda-nvrtc==13.0.88
[pip3] nvidia-cuda-runtime==13.0.96
[pip3] nvidia-cudnn-cu13==9.19.0.56
[pip3] nvidia-cudnn-frontend==1.18.0
[pip3] nvidia-cufft==12.0.0.61
[pip3] nvidia-cufile==1.15.1.6
[pip3] nvidia-curand==10.4.0.35
[pip3] nvidia-cusolver==12.0.4.66
[pip3] nvidia-cusparse==12.6.3.3
[pip3] nvidia-cusparselt-cu13==0.8.0
[pip3] nvidia-cutlass-dsl==4.4.2
[pip3] nvidia-cutlass-dsl-libs-base==4.4.2
[pip3] nvidia-ml-py==13.595.45
[pip3] nvidia-nccl-cu13==2.28.9
[pip3] nvidia-nvjitlink==13.0.88
[pip3] nvidia-nvshmem-cu13==3.4.5
[pip3] nvidia-nvtx==13.0.85
[pip3] pyzmq==27.1.0
[pip3] torch==2.11.0+cu130
[pip3] torch_c_dlpack_ext==0.1.5
[pip3] torchaudio==2.11.0+cu130
[pip3] torchvision==0.26.0+cu130
[pip3] transformers==5.5.4
[pip3] triton==3.6.0
[conda] Could not collect

==============================
         vLLM Info
==============================
ROCM Version                 : Could not collect
vLLM Version                 : 0.19.1rc1.dev328+g18013df6a (git sha: 18013df6a)
vLLM Build Flags:
  CUDA Archs: 7.0 7.5 8.0 8.9 9.0 10.0 12.0; ROCm: Disabled; XPU: Disabled
GPU Topology:
  	GPU0	GPU1	GPU2	GPU3	GPU4	GPU5	GPU6	GPU7	NIC0	NIC1	NIC2	NIC3	NIC4	NIC5	NIC6	NIC7	CPU Affinity	NUMA Affinity	GPU NUMA ID
GPU0	 X 	NV8	NV8	NV8	NV8	NV8	NV8	NV8	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	0-47,96-143	0		N/A
GPU1	NV8	 X 	NV8	NV8	NV8	NV8	NV8	NV8	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	0-47,96-143	0		N/A
GPU2	NV8	NV8	 X 	NV8	NV8	NV8	NV8	NV8	PIX	PIX	NODE	NODE	SYS	SYS	SYS	SYS	0-47,96-143	0		N/A
GPU3	NV8	NV8	NV8	 X 	NV8	NV8	NV8	NV8	PXB	PXB	NODE	NODE	SYS	SYS	SYS	SYS	0-47,96-143	0		N/A
GPU4	NV8	NV8	NV8	NV8	 X 	NV8	NV8	NV8	SYS	SYS	SYS	SYS	PIX	PIX	NODE	NODE	48-95,144-191	1		N/A
GPU5	NV8	NV8	NV8	NV8	NV8	 X 	NV8	NV8	SYS	SYS	SYS	SYS	PXB	PXB	NODE	NODE	48-95,144-191	1		N/A
GPU6	NV8	NV8	NV8	NV8	NV8	NV8	 X 	NV8	SYS	SYS	SYS	SYS	NODE	NODE	NODE	NODE	48-95,144-191	1		N/A
GPU7	NV8	NV8	NV8	NV8	NV8	NV8	NV8	 X 	SYS	SYS	SYS	SYS	NODE	NODE	NODE	NODE	48-95,144-191	1		N/A
NIC0	NODE	NODE	PIX	PXB	SYS	SYS	SYS	SYS	 X 	PIX	NODE	NODE	SYS	SYS	SYS	SYS				
NIC1	NODE	NODE	PIX	PXB	SYS	SYS	SYS	SYS	PIX	 X 	NODE	NODE	SYS	SYS	SYS	SYS				
NIC2	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	NODE	NODE	 X 	PIX	SYS	SYS	SYS	SYS				
NIC3	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	NODE	NODE	PIX	 X 	SYS	SYS	SYS	SYS				
NIC4	SYS	SYS	SYS	SYS	PIX	PXB	NODE	NODE	SYS	SYS	SYS	SYS	 X 	PIX	NODE	NODE				
NIC5	SYS	SYS	SYS	SYS	PIX	PXB	NODE	NODE	SYS	SYS	SYS	SYS	PIX	 X 	NODE	NODE				
NIC6	SYS	SYS	SYS	SYS	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	NODE	NODE	 X 	PIX				
NIC7	SYS	SYS	SYS	SYS	NODE	NODE	NODE	NODE	SYS	SYS	SYS	SYS	NODE	NODE	PIX	 X 				

Legend:

  X    = Self
  SYS  = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
  NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
  PHB  = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
  PXB  = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
  PIX  = Connection traversing at most a single PCIe bridge
  NV#  = Connection traversing a bonded set of # NVLinks

NIC Legend:

  NIC0: mlx5_0
  NIC1: mlx5_1
  NIC2: mlx5_2
  NIC3: mlx5_3
  NIC4: mlx5_4
  NIC5: mlx5_5
  NIC6: mlx5_6
  NIC7: mlx5_7

==============================
     Environment Variables
==============================
NVIDIA_VISIBLE_DEVICES=GPU-750f387f-40be-cef6-531e-8c754db35274
NVIDIA_REQUIRE_CUDA=cuda>=13.0 brand=unknown,driver>=535,driver<536 brand=grid,driver>=535,driver<536 brand=tesla,driver>=535,driver<536 brand=nvidia,driver>=535,driver<536 brand=quadro,driver>=535,driver<536 brand=quadrortx,driver>=535,driver<536 brand=nvidiartx,driver>=535,driver<536 brand=vapps,driver>=535,driver<536 brand=vpc,driver>=535,driver<536 brand=vcs,driver>=535,driver<536 brand=vws,driver>=535,driver<536 brand=cloudgaming,driver>=535,driver<536 brand=unknown,driver>=550,driver<551 brand=grid,driver>=550,driver<551 brand=tesla,driver>=550,driver<551 brand=nvidia,driver>=550,driver<551 brand=quadro,driver>=550,driver<551 brand=quadrortx,driver>=550,driver<551 brand=nvidiartx,driver>=550,driver<551 brand=vapps,driver>=550,driver<551 brand=vpc,driver>=550,driver<551 brand=vcs,driver>=550,driver<551 brand=vws,driver>=550,driver<551 brand=cloudgaming,driver>=550,driver<551 brand=unknown,driver>=565,driver<566 brand=grid,driver>=565,driver<566 brand=tesla,driver>=565,driver<566 brand=nvidia,driver>=565,driver<566 brand=quadro,driver>=565,driver<566 brand=quadrortx,driver>=565,driver<566 brand=nvidiartx,driver>=565,driver<566 brand=vapps,driver>=565,driver<566 brand=vpc,driver>=565,driver<566 brand=vcs,driver>=565,driver<566 brand=vws,driver>=565,driver<566 brand=cloudgaming,driver>=565,driver<566 brand=unknown,driver>=570,driver<571 brand=grid,driver>=570,driver<571 brand=tesla,driver>=570,driver<571 brand=nvidia,driver>=570,driver<571 brand=quadro,driver>=570,driver<571 brand=quadrortx,driver>=570,driver<571 brand=nvidiartx,driver>=570,driver<571 brand=vapps,driver>=570,driver<571 brand=vpc,driver>=570,driver<571 brand=vcs,driver>=570,driver<571 brand=vws,driver>=570,driver<571 brand=cloudgaming,driver>=570,driver<571 brand=unknown,driver>=575,driver<576 brand=grid,driver>=575,driver<576 brand=tesla,driver>=575,driver<576 brand=nvidia,driver>=575,driver<576 brand=quadro,driver>=575,driver<576 brand=quadrortx,driver>=575,driver<576 brand=nvidiartx,driver>=575,driver<576 brand=vapps,driver>=575,driver<576 brand=vpc,driver>=575,driver<576 brand=vcs,driver>=575,driver<576 brand=vws,driver>=575,driver<576 brand=cloudgaming,driver>=575,driver<576
TORCH_CUDA_ARCH_LIST=7.0 7.5 8.0 8.9 9.0 10.0 12.0
CUDA_DEVICE_SM_LIMIT=0
NVIDIA_DRIVER_CAPABILITIES=compute,utility
VLLM_USAGE_SOURCE=production-docker-image
CUDA_VERSION=13.0.1
VLLM_ENABLE_CUDA_COMPATIBILITY=0
CUDA_DEVICE_MEMORY_LIMIT_0=81559m
CUDA_DEVICE_MEMORY_SHARED_CACHE=/usr/local/vgpu/3f21f157-aeb6-49e0-9f8e-a3bad69b5407.cache
LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib64
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
TORCHINDUCTOR_CACHE_DIR=/tmp/torchinductor_root
</details>

🐛 Describe the bug

I have tested the performance of Model Runner V2 with Eagle3 enabled on H800. The result is inferior to that of Model Runner V1. Current analysis identifies that the excessive latency of the _gumbel_sample_kernel is the root cause, as shown in the figure below. <img width="2484" height="1034" alt="Image" src="https://github.com/user-attachments/assets/0e2227a4-40e6-4e40-a501-beffc2b7a083" />

I conducted the test with the following command and test script. command:

VLLM_USE_V2_MODEL_RUNNER=1 vllm serve /nas/disk1/Qwen3-8B --max-num-batched-tokens 32768 --no-enable-prefix-caching --host 0.0.0.0 --port 8898

test script

vllm bench serve --backend openai-chat --model /nas/disk1/Qwen3-8B --tokenizer /nas/disk1/Qwen3-8B --served-model-name /nas/disk1/Qwen3-8B --dataset-name custom --dataset-path /model/lyc/esData/esdata5daysago.jsonl --custom-output-len 100 --num-prompts 400 --max-concurrency 40 --endpoint /v1/chat/completions --ignore-eos --percentile-metrics ttft,tpot,itl,e2el --host 55.122.12.106 --port 30004 --seed 42

The _gumbel_sample_kernel needs to process logits with the shape of [num_tokens, vocab_size].

def gumbel_sample( logits: torch.Tensor, # [num_tokens, vocab_size] ...... ) -> torch.Tensor: num_tokens, vocab_size = logits.shape BLOCK_SIZE = 1024 num_blocks = triton.cdiv(vocab_size, BLOCK_SIZE) local_argmax = logits.new_empty(num_tokens, num_blocks, dtype=torch.int64) local_max = logits.new_empty(num_tokens, num_blocks, dtype=torch.float64) _gumbel_sample_kernel[(num_tokens, num_blocks)](

When Eagle3 is disabled, each request generates one token. Under 40 concurrency, num_tokens equals 40, and the latency of this operator on the H800 is approximately 800 μs. When Eagle3 is enabled and num_draft_token is 5, each request need to deal with 6 tokens. So under 40 concurrency, num_tokens equals 240, the latency of this operator on the H800 is approximately 4ms.

I tested the performance of PR #34854. The comparison with the current performance data is shown below.

40 concurrency

num_tokens#34854 pr#37798 pr
40(eagle3 disable)81us800us
240(eagle3 enable)374us4ms

I would like to confirm whether this behavior is normal. @WoosukKwon, could you please take a look at this issue?

Before submitting a new issue...

  • Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

extent analysis

TL;DR

The excessive latency of the _gumbel_sample_kernel is likely due to the increased num_tokens when Eagle3 is enabled, causing a significant performance degradation.

Guidance

  • Investigate the _gumbel_sample_kernel implementation to understand its performance characteristics and potential bottlenecks.
  • Consider optimizing the kernel for larger num_tokens values, potentially by improving parallelization or reducing memory access patterns.
  • Evaluate the impact of num_draft_token on performance and explore alternatives to reduce the number of tokens processed by the _gumbel_sample_kernel.
  • Compare the performance of PR #34854 with the current implementation to identify potential improvements.

Example

No code snippet is provided as the issue is more related to performance analysis and optimization rather than a specific code fix.

Notes

The issue seems to be related to the performance of a specific kernel (_gumbel_sample_kernel) when dealing with larger input sizes (num_tokens). The provided data suggests a significant increase in latency when Eagle3 is enabled, which may indicate a need for optimization or alternative approaches.

Recommendation

Apply workaround: Investigate and optimize the _gumbel_sample_kernel for larger num_tokens values to reduce latency and improve performance when Eagle3 is enabled.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#api#dependency conflict#environment setup#docker error#environment variable

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

vllm - ✅(Solved) Fix [Model Runner V2][Bug]: The _gumbel_sample_kernel exhibits poor performance on H800. [1 pull requests, 1 participants]