framework,version,device,op_name,kernel_source,mla_dtype,kv_cache_dtype,num_heads,batch_size,isl,tp_size,step,latency
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,1,0.3107146620750427
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,1,0.3265226682027181
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,1,0.3187466661135356
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,1,0.3062453269958496
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,1,0.34001068274180096
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,1,0.3134079972902934
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,1,0.3187626600265503
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,1,0.31518399715423584
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,1,0.3695146640141805
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,1,0.3154826760292053
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,1,0.32387200991312665
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,1,0.34890135129292804
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,1,0.349834680557251
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,1,0.3150399923324585
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,1,0.3489280144373576
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,1,0.8841439882914225
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,1,0.3147413333257039
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,1,0.343669335047404
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,1,0.3314560055732727
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,1,0.30932267506917316
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,1,0.3198666572570801
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,1,0.3718453248341878
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,1,0.34909331798553467
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,1,0.3423360188802083
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,1,0.34470399220784503
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,1,0.3360746701558431
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,1,0.31112533807754517
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,1,0.36795198917388916
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,1,0.31651200850804645
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,1,0.32819199562072754
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,1,0.34411199887593585
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,1,0.3427893320719401
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,1,0.31539199749628705
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,1,0.31505600611368817
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,1,0.33430933952331543
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,1,0.34206398328145343
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,1,0.3419253428777059
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,1,0.3396480083465576
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,1,0.32713067531585693
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,1,0.34433066844940186
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,1,0.343509316444397
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,1,0.32048000892003375
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,1,0.3362186749776204
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,1,0.3458026647567749
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,1,0.34220266342163086
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,1,0.3489333391189575
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,1,0.31005332867304486
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,1,0.343669335047404
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,1,0.3137600024541219
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,1,0.34652264912923175
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,1,0.3399999936421712
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,1,0.30451732873916626
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,1,0.3381386597951253
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,1,0.377402663230896
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,1,0.33869866530100506
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,1,0.3151786724726359
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,1,0.3439626693725586
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,1,0.3492906490961711
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,1,0.3771786689758301
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,1,0.37645868460337323
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,1,0.35128533840179443
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,1,0.33858664830525714
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,1,0.31542932987213135
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,1,0.32556267579396564
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,1,0.33956265449523926
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,1,0.34964799880981445
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,1,0.3483733336130778
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,1,0.3507893482844035
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,1,0.3153226574261983
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,1,0.6716852982838949
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,1,0.31863999366760254
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,1,0.3362933397293091
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,1,0.3474453290303548
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,1,0.3433653513590495
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,1,0.3524959882100423
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,1,0.63919464747111
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,1,1.6297866503397624
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,1,0.3203893303871155
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,1,0.33985598882039386
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,1,0.3367946545283
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,1,0.6310240030288696
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,1,0.34945066769917804
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,1,1.6129919687906902
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,1,0.35232532024383545
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,1,0.3453119993209839
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,1,0.3431359926859538
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,1,0.33448533217112225
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,1,3.5293334325154624
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,3,0.3361813227335612
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,3,0.34065067768096924
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,3,0.33931732177734375
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,3,0.3378880023956299
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,3,0.3127359946568807
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,3,0.3473546504974365
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,3,0.3572746515274048
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,3,0.34673066933949787
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,3,0.34810133775075275
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,3,0.3521706660588582
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,3,0.3638399839401245
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,3,0.3476746479670207
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,3,0.34893866380055744
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,3,0.4062986771265666
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,3,0.7688746452331543
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,3,0.34565865993499756
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,3,0.3629386822382609
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,3,0.34993600845336914
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,3,0.3553599913914998
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,3,0.35920000076293945
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,3,0.3140053351720174
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,3,0.3736853202184041
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,3,0.3447999954223633
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,3,2.2084959348042807
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,3,0.6627839803695679
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,3,0.3173813422520955
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,3,0.3332479993502299
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,3,0.31983999411265057
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,3,0.5410399834314982
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,3,0.3387146790822347
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,3,4.8100159962972
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,3,0.3484373490015666
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,3,0.3513013521830241
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,3,0.35364266236623126
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,3,0.31414933999379474
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,3,1.4503733317057292
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,3,0.4039626518885295
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,3,0.39770134290059406
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,3,0.3628693421681722
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,3,0.5409013430277506
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,3,0.3383573293685913
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,3,0.3343840042750041
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,3,0.3558666706085205
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,3,0.3538026809692383
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,3,1.1619359652201335
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,3,0.34437334537506104
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,3,0.3501226504643758
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,3,0.3349866469701131
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,3,0.3129226764043172
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,3,0.3481706778208415
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,3,0.3505866527557373
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,3,1.2976000308990479
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,3,0.3526666561762492
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,3,0.3604746659596761
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,3,0.3975573380788167
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,3,0.3160799940427144
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,3,0.48449599742889404
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,3,0.35367465019226074
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,3,0.38390934467315674
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,3,0.34912534554799396
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,3,0.3589013417561849
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,3,0.3542773326237996
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,3,0.38053866227467853
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,3,0.31309332450230914
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,3,0.35097599029541016
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,3,1.315344015757243
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,3,0.34890135129292804
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,3,0.35332798957824707
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,3,1.2293492952982585
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,3,0.35680532455444336
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,3,0.31863999366760254
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,3,0.33208000659942627
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,3,0.4601279894510905
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,3,0.3455573320388794
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,3,0.353279987970988
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,3,1.2737226486206055
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,3,0.32452799876530963
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,3,2.8349278767903647
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,3,0.32656000057856244
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,3,0.3334346612294515
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,3,0.4517600138982137
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,3,0.3455573320388794
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,3,1.270522673924764
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,3,2.804431915283203
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,3,5.633306503295898
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,7,0.3376213312149048
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,7,0.3555413484573364
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,3,0.35356799761454266
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,7,0.37225600083669025
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,7,0.3482453425725301
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,7,0.3487253189086914
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,7,0.34252798557281494
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,7,0.5103253523508707
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,7,0.3452800114949544
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,7,0.311626672744751
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,7,0.3490399916966756
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,7,1.3820106188456218
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,7,0.3229759931564331
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,7,0.35207998752593994
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,7,0.343664010365804
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,7,0.3479573329289754
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,3,0.3519146839777629
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,7,0.33898667494455975
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,7,0.34325865904490155
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,7,0.35321064790089923
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,7,0.35472532113393146
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,7,0.3522133429845174
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,7,0.35471999645233154
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,7,0.3558453321456909
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,7,0.3573226531346639
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,7,0.35440532366434735
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,7,0.35761066277821857
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,7,0.35312533378601074
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,7,0.3619626760482788
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,7,0.3588373263676961
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,7,0.353279987970988
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,7,0.3423893451690674
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,7,1.3783733050028484
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,3,1.1411680380503337
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,7,0.32015466690063477
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,7,0.3262773354848226
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,7,0.31601067384084064
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,7,0.37854401270548504
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,7,0.36405332883199054
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,7,0.3568106492360433
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,7,0.34621334075927734
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,7,0.38386666774749756
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,7,0.35416531562805176
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,7,0.34857598940531415
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,7,0.35074134667714435
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,7,0.35100265343983966
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,7,0.35319467385609943
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,7,0.3516266743342082
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,7,0.349727988243103
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,7,0.35266133149464923
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,7,0.36181867122650146
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,7,0.3472586472829183
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,7,0.32499732573827106
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,7,0.34888001283009845
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,7,0.345360000928243
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,7,0.3433813254038493
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,7,0.34832533200581867
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,7,0.32098666826883954
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,7,0.6153759956359863
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,7,0.3351680040359497
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,7,0.35442133744557697
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,7,0.40008000532786053
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,7,0.33787735303243
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,7,0.34940266609191895
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,7,0.3499679962793986
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,7,0.34358398119608563
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,7,0.3216586709022522
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,7,0.5989173253377279
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,7,0.34481600920359295
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,7,1.5747307141621907
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,7,0.3462933301925659
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,7,0.3617759943008423
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,7,0.3550026814142863
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,7,0.3185386657714844
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,7,0.3047306736310323
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,7,0.34980801741282147
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,7,0.34931735197703045
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,7,0.5969386498133341
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,7,1.5453440348307292
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,7,3.366191864013672
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,7,0.3139413396517436
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,7,0.34600532054901123
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,7,0.33450134595235187
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,7,0.35361599922180176
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,7,0.5935253302256266
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,7,3.3493760426839194
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,7,1.5418933232625325
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,15,0.3203999996185303
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,15,0.3074880043665568
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,15,0.337285319964091
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,15,0.33816532293955487
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,15,0.33568533261617023
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,15,0.948357343673706
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,15,0.34358398119608563
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,15,0.3481066624323527
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,7,6.7061812082926435
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,15,0.34463465213775635
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,15,0.3431306680043538
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,15,0.32175467411677044
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,15,0.34113065401713055
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,7,0.31914132833480835
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,15,0.3471466700236003
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,15,0.35266133149464923
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,15,0.3474400043487549
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,15,0.4848586718241374
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,7,1.0371200243632
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,15,1.2552320162455242
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,15,0.3611413240432739
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,7,0.3649226824442546
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,15,0.6128213405609131
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,15,0.3535573482513428
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,15,0.3461013237635295
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,15,0.3820799986521403
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,15,0.3495413462320964
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,15,0.43653865655263263
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,15,0.346837321917216
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,15,0.3322826623916626
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,15,0.3537973165512085
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,15,0.35013866424560547
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,15,0.3510506550470988
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,15,0.3474026521046956
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,15,0.35277867317199707
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,15,0.3478399912516276
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,15,0.3544906775156657
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,15,0.37517865498860675
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,15,0.3546186685562134
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,15,0.35049601395924884
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,15,2.563210646311442
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,15,0.32180800040562946
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,15,0.35313065846761066
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,15,0.3399893442789714
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,15,0.3428853352864583
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,15,0.3526986837387085
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,15,0.6430986722310384
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,15,0.3614506721496582
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,15,0.34479467074076336
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,15,1.4000426928202312
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,15,0.3496853510538737
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,15,0.3207893371582031
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,15,0.34834134578704834
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,15,0.37809598445892334
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,15,0.38070933024088544
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,15,0.35420799255371094
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,15,0.3824053208033244
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,15,1.4832480748494465
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,15,0.32038400570551556
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,15,0.32499200105667114
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,15,0.35132265090942383
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,15,0.35574932893117267
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,15,0.9109866619110107
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,15,0.3389333486557007
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,15,0.35687466462453205
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,15,0.38598934809366864
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,15,0.6567733287811279
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,15,3.820277214050293
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,15,0.35539201895395917
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,15,0.34568532307942706
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,15,0.6026240189870199
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,15,0.35923198858896893
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,15,0.37508265177408856
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,15,0.356330672899882
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,15,1.5923093159993489
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,15,0.34853867689768475
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,15,0.348469336827596
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,15,0.34991466999053955
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,15,0.3471626838048299
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,15,0.36953067779541016
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,15,0.3349333206812541
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,15,0.6058453321456909
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,15,1.6687307357788086
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,15,3.371413230895996
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,15,0.35197333494822186
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,15,3.346362749735514
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,15,0.34829866886138916
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,15,1.5763413111368816
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,15,0.641813317934672
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,15,0.35204799969991046
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,31,0.344490647315979
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,31,0.3434613148371379
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,31,0.3420480092366536
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,31,0.3344959815343221
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,15,0.330458660920461
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,31,0.33925867080688477
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,15,6.710266749064128
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,31,0.34333332379659015
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,31,0.34521599610646564
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,31,0.33470932642618817
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,31,0.3218826651573181
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,31,0.38012266159057617
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,31,0.35153599580128986
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,31,0.3452213207880656
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,31,0.3492746750513713
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,31,0.36716798941294354
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,31,0.3447573184967041
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,31,0.3516586621602376
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,31,0.35186131795247394
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,31,0.352730671564738
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,31,0.6488320032755533
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,31,0.37810667355855304
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,31,0.34594134489695233
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,31,0.3407040039698283
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,31,0.35051735242207843
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,31,0.34971733887990314
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,31,0.3524053494135539
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,31,1.9391412734985352
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,31,0.36841599146525067
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,31,0.34835199515024823
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,31,0.35092798868815106
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,31,0.34753600756327313
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,31,0.3561439911524455
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,31,0.3693866729736328
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,31,0.40065598487854004
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,31,0.3378293514251709
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,31,0.34958934783935547
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,31,0.3516106605529785
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,31,0.3506346543629964
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,31,0.35256000359853107
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,31,0.34407468636830646
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,31,1.3332373301188152
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,31,0.35230934619903564
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,31,0.35099200407663983
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,31,0.34644798437754315
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,31,0.3510719935099284
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,31,0.34838934739430744
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,31,0.3172266681989034
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,31,0.3492266734441121
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,31,0.3491946856180827
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,31,0.3501439889272054
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,31,0.44143998622894287
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,31,0.33366934458414715
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,31,0.34886399904886883
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,31,0.3510986566543579
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,31,0.34518933296203613
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,31,0.3514133294423421
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,31,0.33697064717610675
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,31,0.3195466597874959
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,31,0.6237759987513224
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,31,0.3497120141983032
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,31,0.3521386782328288
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,31,0.35440532366434735
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,31,0.3503733476003011
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,31,0.3811253309249878
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,31,0.34753600756327313
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,31,0.35232532024383545
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,31,2.118117332458496
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,31,0.337007999420166
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,31,0.3506506681442261
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,31,0.7827146848042806
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,31,0.35522667566935223
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,31,1.5933279991149902
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,31,0.34278400739034015
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,31,0.3609066804250081
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,31,0.34738131364186603
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,31,0.4345173438390096
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,31,0.35102399190266925
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,31,0.6264853477478027
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,31,1.5920532544453938
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,31,3.377141316731771
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,31,0.3185439904530843
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,31,1.634768009185791
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,31,0.36135466893513996
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,31,0.35205332438151044
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,31,0.651317318280538
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,31,3.3952906926472983
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,31,0.3579519987106323
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,63,0.33952534198760986
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,63,0.33769599596659344
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,63,0.34357333183288574
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,63,0.3397066593170166
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,31,6.755568186442058
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,31,0.35499199231465656
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,63,0.3427040179570516
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,63,0.3350293238957723
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,63,0.3559360106786092
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,63,0.3513600031534831
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,63,0.312063992023468
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,63,0.3424533208211263
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,63,0.34065600236256915
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,63,1.8686399459838867
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,63,0.3186720013618469
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,63,0.3521600166956584
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,63,0.3646986484527588
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,63,0.3459626833597819
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,63,0.346127986907959
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,63,0.34726933638254803
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,63,0.34734400113423664
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,63,0.3516480127970378
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,63,0.34466131528218585
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,63,0.350053350130717
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,63,0.3126453359921773
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,63,0.34941331545511883
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,63,0.35072000821431476
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,63,0.3487306833267212
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,63,0.3444960117340088
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,63,0.3548053503036499
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,63,0.3487733205159505
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,63,0.3102826674779256
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,63,1.9554239908854167
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,63,0.3515626589457194
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,63,0.3574399948120117
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,63,0.8188959757486979
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,63,0.8538453578948975
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,63,0.3150506615638733
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,63,0.6468000014623007
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,63,0.3463199933369954
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,63,0.33367466926574707
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,63,0.34626134236653644
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,63,0.38140801588694256
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,63,0.3790239890416463
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,63,0.3450026512145996
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,63,0.3580319881439209
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,63,1.3217013676961262
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,63,0.3507466713587443
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,63,0.3269813259442647
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,63,0.31150933106740314
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,63,0.34830931822458905
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,63,0.3567093213399251
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,63,0.3575093348821004
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,63,0.3582133452097575
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,63,0.3512959877649943
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,63,0.3450346787770589
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,63,0.35951467355092365
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,63,0.3779093424479167
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,63,0.348965326944987
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,63,0.3813120126724243
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,63,0.37936532497406006
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,63,0.34966933727264404
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,63,0.35002132256825763
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,63,0.6358186801274618
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,63,0.35258666674296063
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,63,0.3396000067392985
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,63,0.63373335202535
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,63,0.3789173364639282
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,63,0.37644267082214355
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,63,0.3797066609064738
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,63,0.35438398520151776
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,63,0.3471733331680298
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,63,1.6186879475911458
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,63,0.3169813354810079
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,63,0.36426134904225665
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,63,0.666320006052653
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,63,0.34884798526763916
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,63,0.3489760160446167
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,63,1.6262720425923665
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,63,3.397695859273275
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,63,0.3149813413619995
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,63,0.32818132638931274
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,63,2.347599983215332
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,63,0.7269706726074219
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,63,0.35201064745585126
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,63,3.531754811604818
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,63,0.37034134070078534
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,63,0.3485333522160848
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,63,7.0276533762613935
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,63,0.3264320095380147
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,127,0.5116373300552368
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,127,0.5000799894332886
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,127,0.7280213038126627
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,127,0.3337706724802653
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,127,0.5040800174077352
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,127,0.5324533383051554
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,127,0.5374453465143839
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,127,0.6619146664937338
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,127,0.3567519982655843
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,127,0.5424533287684122
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,127,0.3474773168563843
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,127,0.3421493371327718
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,127,2.6627680460611978
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,127,0.3452533483505249
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,127,0.34413333733876544
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,127,0.3375946680704753
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,127,0.3526719808578491
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,127,0.3484319845835368
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,127,0.35793066024780273
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,127,0.3506506681442261
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,127,1.8923039436340332
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,127,0.3456053336461385
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,127,0.33395198980967206
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,127,0.33182400465011597
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,127,0.3415253162384033
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,127,0.34331198533376056
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,127,0.3538293441136678
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,127,0.3577973445256551
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,127,0.34402668476104736
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,127,0.3240213394165039
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,127,0.31701866785685223
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,127,0.3497226635615031
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,127,0.3375946680704753
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,127,0.3694560130437215
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,127,0.3412160078684489
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,127,0.32525867223739624
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,127,0.3086400032043457
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,127,0.3336053291956584
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,127,0.33746135234832764
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,127,0.3336533308029175
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,127,0.33768534660339355
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,127,0.3342826763788859
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,127,0.31252266963322956
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,127,0.30400000015894574
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,127,0.34008534749348956
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,127,0.33429865042368573
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,127,0.33670934041341144
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,127,0.33768534660339355
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,127,0.33501867453257245
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,127,0.6536586682001749
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,127,0.3352533181508382
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,127,0.3325120011965434
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,127,0.33418134848276776
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,127,0.33157867193222046
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,127,0.3351893424987793
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,127,0.33221866687138873
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,127,0.6865812937418619
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,127,1.6668799718221028
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,127,0.7092693646748861
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,127,0.3226026693979899
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,127,0.30664000908533734
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,127,0.30874667565027875
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,127,0.30844799677530926
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,127,0.3078560034434001
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,127,0.3010666569073995
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,127,3.5063467025756836
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,127,0.30348799626032513
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,127,0.29873599608739215
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,127,0.40411198139190674
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,127,1.7343840599060059
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,127,0.7575626373291016
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,127,0.31226134300231934
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,127,0.36341333389282227
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,127,0.31592533985773724
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,127,0.3369813362757365
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,127,0.34306132793426514
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,127,0.30742400884628296
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,127,0.30053865909576416
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,127,0.3072213331858317
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,127,0.8829867045084635
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,127,1.975279966990153
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,127,0.5072266658147176
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,127,3.9236427942911782
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,127,7.7280426025390625
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,255,0.3239786624908447
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,255,0.35117332140604657
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,255,0.31003199021021527
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,255,0.33748265107472736
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,255,0.3400426705678304
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,255,0.34223465124766034
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,255,0.331493337949117
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,255,0.5347093343734741
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,255,0.3366933266321818
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,127,0.3871786594390869
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,255,0.38843198617299396
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,255,0.34489067395528156
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,255,0.31830400228500366
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,255,0.3256213267644246
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,255,0.3489973147710164
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,255,0.3642880121866862
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,255,1.2917386690775554
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,127,0.37001601854960126
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,255,0.3434986670811971
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,255,0.3508053223292033
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,255,0.49857600529988605
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,255,0.32762666543324787
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,255,0.3463253180185954
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,255,0.3420533339182536
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,255,0.354853351910909
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,255,0.34838398297627765
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,255,0.35978134473164874
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,255,0.3494773308436076
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,255,2.0218666394551597
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,255,0.525386651357015
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,255,0.35091733932495117
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,255,0.6785386403401693
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,255,0.34749865531921387
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,255,0.34111467997233075
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,255,0.3448479970296224
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,255,0.3492799997329712
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,255,0.3506346543629964
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,127,0.3901333411534627
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,127,0.3449973265329997
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,255,0.3454879919687907
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,255,0.3487573464711507
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,255,0.3494666814804077
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,255,0.35072533289591473
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,255,0.35070931911468506
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,255,0.35231467088063556
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,255,0.34359999497731525
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,255,0.34989333152770996
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,255,1.3571732838948567
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,255,0.3842346668243408
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,255,0.34251733620961505
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,255,0.31617599725723267
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,255,0.3160746693611145
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,255,0.3447893460591634
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,255,2.438485304514567
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,255,0.3502133289972941
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,255,0.3510133425394694
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,255,0.3495573202768962
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,255,0.3559733231862386
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,255,0.31835200389226276
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,255,0.31270933151245117
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,255,0.36686933040618896
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,255,0.3438933293024699
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,255,0.34941331545511883
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,255,0.3442666530609131
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,255,0.3272320032119751
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,255,0.6986613273620605
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,255,0.3471946716308594
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,255,0.3436160087585449
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,255,0.3945759932200114
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,255,0.3485226631164551
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,255,0.35321064790089923
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,255,0.38363198439280194
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,255,0.7707306543986002
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,255,1.7455466588338215
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,255,0.3503199815750122
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,255,0.333679993947347
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,255,1.9875359535217285
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,255,0.5056373278299967
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,255,0.9664106369018555
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,255,0.3779999812444051
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,255,0.37452268600463867
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,255,0.3436746597290039
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,255,0.3669546842575073
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,255,3.868000030517578
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,255,0.6069546540578207
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,255,0.5411680142084757
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,255,0.5629066626230875
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,255,1.1375093460083008
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,255,0.7419093449910482
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,255,2.4020959536234536
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,255,4.810240109761556
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,511,0.3389279842376709
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,511,0.31173866987228394
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,255,9.547749201456705
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,511,0.33796266714731854
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,511,0.7184906800587972
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,511,0.3366026480992635
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,511,0.3486986557642619
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,511,0.3486666679382324
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,511,2.4094932874043784
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,511,0.3477226495742798
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,511,0.3487893342971802
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,511,0.34694401423136395
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,511,0.3457119862238566
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,511,0.3532586495081584
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,511,0.31961599985758465
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,511,0.3115839958190918
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,511,0.34829866886138916
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,511,0.3314506610234578
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,511,0.4304853280385335
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,511,0.3511679967244466
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,511,0.34908799330393475
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,511,0.31404266754786175
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,511,0.3092213273048401
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,511,0.34595731894175213
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,511,1.3928106625874836
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,511,0.41236265500386554
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,511,0.40458667278289795
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,511,0.38206934928894043
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,511,0.34541865189870197
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,511,0.4190026521682739
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,511,0.351093331972758
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,511,0.3464053471883138
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,511,0.36128532886505127
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,511,0.35284801324208576
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,511,0.3452106714248657
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,511,0.4715786774953206
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,511,0.3463786840438843
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,511,0.34940799077351886
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,511,3.1121066411336265
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,511,0.34595731894175213
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,511,0.34798399607340497
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,511,0.34200533231099445
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,511,0.34200533231099445
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,511,0.34169065952301025
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,511,0.3481866518656413
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,511,0.3439146677652995
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,511,0.3528906504313151
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,511,0.3431626558303833
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,511,0.4354506731033325
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,511,0.38093332449595135
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,511,0.3431466817855835
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,511,0.3596160014470418
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,511,0.3442933162053426
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,511,0.33580267429351807
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,511,0.34381333986918133
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,511,0.3471626838048299
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,511,0.32522666454315186
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,511,0.4338239828745524
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,511,0.34117333094278973
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,511,0.3479679822921753
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,511,0.857151985168457
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,511,1.8792959849039714
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,511,0.48023998737335205
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,511,0.34747199217478436
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,511,0.31224532922108966
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,511,0.4259200096130371
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,511,0.3777546485265096
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,511,0.5468213160832723
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,511,0.779584010442098
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,511,0.3673226833343506
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,511,0.9901386896769205
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,511,1.9322080612182617
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,511,0.3517013390858968
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,511,1.4196640650431316
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,511,0.8637226422627767
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,511,0.7409706910451254
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,511,0.5371946493784586
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,511,2.7216641108194985
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,511,0.5522240002950033
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,511,5.38261349995931
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,511,0.5024426778157552
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,511,1.5516907374064128
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,511,3.1560586293538413
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,511,2.099717299143473
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,511,1.4274986584981282
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,511,0.9684000015258789
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,511,1.0138026873270671
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,511,6.299045562744141
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,511,12.670661926269531
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,1023,0.32150934139887494
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,1023,0.307584007581075
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,1023,0.3139093319574992
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,1023,0.30722665786743164
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,1023,0.33744001388549805
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,1023,0.3044533332188924
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,1023,0.30614932378133136
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,1023,0.31146132946014404
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,1023,0.31382934252421063
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,1023,0.3104533354441325
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,1023,0.3317333261171977
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,1023,0.3224853277206421
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,1023,0.3405439853668213
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,1023,0.3397279977798462
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,1023,0.3510560194651286
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,1023,0.32790400584538776
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,1023,0.34310932954152423
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,1023,0.3508533239364624
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,1023,0.34384000301361084
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,1023,0.3110239903132121
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,1023,0.3166666626930237
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,1023,0.34025601545969647
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,1023,0.3144479990005493
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,1023,0.3103626569112142
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,1023,0.34004799524943036
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,1023,0.3030453324317932
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,1023,0.30717867612838745
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,1023,0.30853867530822754
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,1023,0.30451732873916626
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,1023,0.33955732981363934
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,1023,0.3211573362350464
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,1023,0.34443199634552
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,1023,0.3108160098393758
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,1023,0.32706133524576825
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,1023,0.30534933010737103
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,1023,0.30669333537419635
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,1023,0.30804266532262164
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,1023,0.3096853295962016
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,1023,0.31016000111897785
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,1023,0.3072800040245056
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,1023,0.32151999076207477
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,1023,0.4523093303044637
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,1023,0.5707093477249146
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,1023,0.38279465834299725
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,1023,0.5603679815928141
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,1023,0.333957314491272
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,1023,0.34121068318684894
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,1023,0.3425706624984741
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,1023,0.3382879892985026
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,1023,1.250655968983968
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,1023,0.6766506830851237
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,1023,0.4068693319956462
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,1023,0.35809067885080975
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,1023,0.35442666212717694
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,1023,0.35607465108235675
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,1023,0.3084106643994649
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,1023,0.3543039957682292
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,1023,0.3630826473236084
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,1023,2.9485813776652017
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,1023,1.5925013224283855
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,1023,0.9193173249562582
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,1023,0.6057813167572021
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,1023,0.5416906674702963
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,1023,0.5091573397318522
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,1023,0.488810658454895
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,1023,0.5446826616923014
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,1023,4.727786699930827
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,1023,9.510784149169922
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,1023,2.368394692738851
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,1023,0.3086613416671753
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,1023,1.192581335703532
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,1023,0.3267893393834432
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,1023,0.31727999448776245
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,1023,0.9658346970876058
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,1023,1.0435946782430012
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,1023,0.9288907051086426
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,1023,1.012069304784139
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,1023,0.309386670589447
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,1023,0.3043733239173889
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,1023,19.018863677978516
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,1023,4.735967953999837
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,1023,9.457647959391275
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,1023,2.3630453745524087
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,2047,0.30633066097895306
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,1023,2.046570618947347
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,2047,0.343450665473938
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,2047,0.34117333094278973
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,2047,0.3405919869740804
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,2047,0.34891200065612793
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,2047,0.34466131528218585
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,2047,0.34326398372650146
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,2047,0.3487199942270915
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,2047,0.31546666224797565
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,2047,0.3197279969851176
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,1023,1.8885706265767415
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,2047,0.3507466713587443
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,2047,0.3497653404871623
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,2047,0.34592000643412274
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,2047,0.3298506736755371
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,2047,0.367578665415446
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,2047,0.35689600308736164
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,2047,0.34942400455474854
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,2047,0.3538080056508382
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,2047,0.3229546745618184
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,2047,0.35079999764760333
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,2047,0.3515946865081787
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,2047,0.3492533365885417
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,2047,0.34458665053049725
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,2047,0.36716798941294354
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,2047,0.31246399879455566
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,1023,1.7952586809794109
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,2047,0.35233068466186523
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,2047,0.34956268469492596
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,2047,0.7137066523234049
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,2047,0.5056320031483968
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,2047,0.35659201939900714
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,2047,0.3464053471883138
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,2047,0.3548266490300496
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,1023,0.34721068541208905
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,2047,0.35649065176645917
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,2047,0.35968534151713055
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,2047,0.314794659614563
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,1023,1.8936373392740886
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,2047,0.3503520091374715
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,2047,0.3461386760075887
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,2047,0.3299199938774109
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,2047,0.3558666706085205
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,2047,0.39314667383829754
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,2047,0.49399999777475995
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,2047,0.3508960008621216
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,2047,0.3744373321533203
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,2047,0.35579200585683185
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,2047,0.3440479834874471
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,2047,0.34614400068918866
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,2047,0.34195733070373535
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,2047,0.3135040005048116
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,2047,0.36986132462819415
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,2047,0.928330659866333
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,2047,0.5240106582641602
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,2047,0.3998719851175944
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,2047,0.3914560079574585
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,2047,0.3763306538263957
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,2047,0.34122133255004883
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,2047,0.3404266834259033
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,2047,0.5138399998346964
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,2047,1.1893173058827717
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,2047,0.5367946624755859
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,2047,0.49216000239054364
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,2047,1.1571839650472004
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,2047,0.4795253276824951
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,2047,2.0623040199279785
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,2047,0.5358773469924927
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,2047,1.060634692509969
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,2047,3.9255892435709634
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,2047,0.9774293104807535
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,2047,0.9354560375213623
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,2047,1.9930507342020671
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,2047,0.9144746462504069
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,2047,7.972501118977864
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,2047,0.9992799758911133
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,2047,4.043802579243978
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,2047,2.041114648183187
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,2047,1.7954346338907878
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,2047,8.077834447224935
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,2047,1.8809386889139812
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,2047,1.7466506958007812
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,2047,16.199637095133465
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,2047,1.9138933817545574
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,2047,16.22410710652669
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,4095,3.1108531951904297
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,2047,32.395477294921875
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,4095,0.3629759947458903
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,4095,0.3558453321456909
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,4095,0.36553064982096356
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,4095,0.34963734944661456
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,4095,0.353434681892395
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,4095,0.355461319287618
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,2047,3.7052745819091797
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,4095,0.33930134773254395
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,4095,0.3558613459269206
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,4095,0.36323734124501544
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,4095,0.3296320041020711
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,4095,0.36566932996114093
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,4095,0.3156426747639974
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,4095,0.3601706822713216
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,4095,0.35095465183258057
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,4095,0.35760001341501874
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,4095,0.3545759916305542
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,4095,0.36238400141398114
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,4095,0.3537919918696086
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,4095,0.3725706736246745
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,4095,0.38594667116800946
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,4095,0.5850239992141724
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,2047,8.077605565388998
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,4095,0.35949333508809406
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,4095,1.422554651896159
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,2047,3.529168128967285
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,2047,4.053274790445964
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,4095,0.3496319850285848
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,2047,3.4196694691975913
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,2047,3.630645434061686
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,4095,1.3975733121236165
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,4095,0.3803360064824422
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,4095,0.3502826690673828
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,4095,0.3779573440551758
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,4095,0.9123679796854655
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,4095,0.35339732964833576
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,4095,0.3333280086517334
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,4095,0.36216533184051514
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,4095,0.3670933246612549
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,4095,0.3686666488647461
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,4095,0.47910932699839276
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,4095,0.3549013137817383
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,4095,0.35704533259073895
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,4095,0.33342933654785156
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,4095,0.7637813091278076
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,4095,1.3503680229187012
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,4095,0.348410685857137
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,4095,0.8578879833221436
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,4095,0.47494399547576904
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,4095,0.8461066881815592
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,4095,0.3664640188217163
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,4095,0.4076373179753621
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,4095,0.3665599822998047
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,4095,1.067951997121175
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,4095,1.7625439961751301
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,4095,0.6444746653238932
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,4095,0.48895466327667236
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,4095,0.48479998111724854
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,4095,1.52292267481486
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,4095,0.4756213426589966
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,4095,0.5330506563186646
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,4095,1.1989920139312744
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,4095,0.9011039733886719
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,4095,1.7969333330790203
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,4095,0.9182666937510172
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,4095,3.415973345438639
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,4095,0.8838559786478678
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,4095,0.9794399738311768
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,4095,6.680538813273112
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,4095,3.701125462849935
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,4095,1.906442642211914
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,4095,7.331941604614258
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,4095,1.7642079989115398
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,4095,1.8103574117024739
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,4095,14.692607879638672
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,4095,1.893349329630534
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,4095,1.7364320755004883
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,4095,7.332096099853516
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,4095,14.658111572265625
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,4095,3.71014404296875
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,4095,29.40442657470703
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,4095,3.4911200205485025
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,4095,3.529029210408529
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,4095,3.3627894719441733
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,4095,3.716554641723633
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,4095,14.648671468098959
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,4095,7.321952184041341
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,8191,0.3580640157063802
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,8191,0.35334400335947674
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,8191,0.3526666561762492
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,8191,0.3698986768722534
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,4095,6.958149592081706
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,8191,0.34962666034698486
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,4095,29.3245112101237
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,4095,6.62667719523112
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,8191,0.32682132720947266
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,8191,0.34645334879557294
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,4095,6.751989364624023
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,8191,0.48692798614501953
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,8191,0.3682773510615031
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,8191,0.3871413469314575
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,8191,0.33460799853007
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,8191,0.3521973292032878
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,8191,0.33590400218963623
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,8191,0.356389323870341
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,8191,0.39722665150960285
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,8191,0.4866986672083537
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,8191,0.3607413371404012
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,8191,0.39188265800476074
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,8191,0.3640640179316203
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,8191,0.3713173468907674
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,8191,0.3500906626383464
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,8191,0.36100268363952637
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,8191,0.37138132254282635
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,8191,0.46492799123128253
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,8191,1.1223573684692383
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,8191,0.3461120128631592
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,8191,0.3568640152613322
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,8191,1.3194986979166667
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,8191,0.3588213523228963
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,8191,0.3618239959081014
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,8191,0.36265599727630615
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,8191,0.48892800013224286
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,8191,0.8329599698384603
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,8191,0.3748319943745931
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,4095,58.7524159749349
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,8191,0.36952535311381024
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,8191,1.3546773592631023
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,8191,1.8086133003234863
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,8191,0.37043734391530353
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,8191,0.3689013322194417
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,8191,0.7095999717712402
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,8191,0.47519465287526447
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,4095,7.114522933959961
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,8191,0.47210665543874103
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,8191,0.48580801486968994
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,8191,0.8641546567281088
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,8191,0.552069346110026
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,8191,1.6101172765096028
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,8191,0.4703253507614136
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,8191,0.7612373034159342
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,8191,1.2392906347910564
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,8191,0.8858986695607504
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,8191,2.263530731201172
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,8191,0.914085308710734
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,8191,0.8919093608856201
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,8191,3.7745014826456704
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,8191,0.88155730565389
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,8191,0.9741866588592529
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,8191,1.770426591237386
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,8191,1.72544527053833
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,8191,3.432453473409017
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,8191,1.688170591990153
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,8191,1.7140480677286785
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,8191,6.686368306477864
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,8191,1.8641653060913086
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,8191,13.30190404256185
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,8191,27.654922485351562
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,8191,4.184117317199707
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,8191,3.3969812393188477
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,8191,6.94862429300944
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,8191,3.4522507985432944
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,8191,13.781572977701822
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,8191,3.6581013997395835
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,8191,3.3737332026163735
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,8191,6.953114827473958
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,8191,13.769524892171225
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,8191,27.572293599446613
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,8191,55.36060078938802
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,8191,6.645834604899089
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,8191,7.321882883707683
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,8191,6.751263936360677
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,8191,6.588464101155599
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1024,1,4,8191,27.607940673828125
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,16383,0.3858506679534912
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1024,1,8,8191,13.852223714192709
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,16383,0.41627200444539386
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,16383,0.38737066586812335
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,16383,0.3822773297627767
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,16383,0.3758133252461751
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,16383,0.36885865529378253
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,16383,0.3725706736246745
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,16383,0.912384033203125
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,16383,0.41969064871470135
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,16383,0.3765706618626912
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,16383,0.38388800621032715
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,16383,0.3837226629257202
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,16383,0.3487360080083211
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,16383,0.37590932846069336
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,16383,0.3731306791305542
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1024,1,2,8191,55.22184753417969
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,16383,0.918341318766276
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,16383,0.3853973150253296
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,16383,0.4575839837392171
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,16383,0.38676265875498456
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1024,1,32,8191,13.135322570800781
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,16383,0.3826826810836792
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,16383,0.3744159936904907
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1024,1,16,8191,13.346506754557291
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,16383,0.3764053185780843
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,16383,0.3826773166656494
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,16383,0.9191093444824219
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1024,1,64,8191,12.999652862548828
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,16383,0.4583786725997925
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,16383,0.813653310139974
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1024,1,128,8191,14.017120361328125
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,16383,0.42683732509613037
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,16383,0.42977599302927655
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,16383,0.381877342859904
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,16383,0.38497598965962726
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,16383,0.390613317489624
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,16383,0.8464799722035726
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,16383,1.3362080256144206
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,16383,0.48602132002512616
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,16383,0.4734293222427368
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,16383,1.5772746404012044
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,16383,0.4690186580022176
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,16383,0.4666026830673218
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,16383,0.465557336807251
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,16383,1.353984038035075
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,16383,0.8974080085754395
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,16383,1.8903627395629883
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,16383,0.8895413080851237
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,16383,0.9986026287078857
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,16383,0.8832106590270996
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,16383,3.1387411753336587
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,16383,0.9448586304982504
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,16383,1.4108692804972331
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,16383,1.72490660349528
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,16383,2.3566080729166665
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,16383,1.7075573603312175
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,16383,4.405786514282227
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,16383,1.692815939585368
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,16383,7.244656244913737
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,16383,1.6881066958109539
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,16383,1.855855941772461
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1024,1,1,8191,110.72032674153645
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,16383,13.175050099690756
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,16383,26.38360595703125
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,16383,3.308442751566569
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,16383,6.717221577962239
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,16383,3.3830881118774414
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,16383,3.3329865137736
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,16383,3.3491039276123047
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,16383,3.6226933797200522
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,16383,26.611600240071613
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,16383,13.40786616007487
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,16383,6.780431747436523
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,16383,6.870047887166341
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,256,1,1,16383,53.77765401204427
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,256,1,32,16383,6.680746714274089
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,256,1,64,16383,6.654938379923503
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,256,1,128,16383,7.222319920857747
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,512,1,8,16383,13.384197235107422
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,512,1,4,16383,26.615440368652344
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,512,1,16,16383,13.169551849365234
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,512,1,32,16383,13.055231730143229
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,512,1,64,16383,12.980480194091797
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,512,1,128,16383,14.456080118815104
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,512,1,2,16383,53.18908182779948
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,32767,0.6770719687143961
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,512,1,1,16383,106.5957539876302
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,32767,0.6533546845118204
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,32767,0.6679413318634033
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,32767,0.5802186727523804
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,32767,0.6551839907964071
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,32767,0.5779146750768026
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,32767,0.5768959919611613
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,32767,1.7885600725809734
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,32767,0.6622666517893473
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,32767,0.7406187057495117
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,32767,0.6568640073140463
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,32767,0.6603413422902426
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,32767,0.5828479925791422
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,32767,0.5820533434549967
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,32767,0.5810933510462443
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,32767,1.7940425872802734
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,32767,0.7274133364359537
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,32767,0.7917226950327555
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,32767,1.8802720705668132
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,32767,0.716597318649292
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,32767,0.6347306569417318
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,32767,0.6308426856994629
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,32767,0.6313706636428833
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,32767,2.6124000549316406
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,32767,1.5092585881551106
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,32767,0.8499786853790283
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,32767,0.7242240111033121
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,32767,0.7221706708272299
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,32767,0.6398346821467081
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,32767,0.6372106472651163
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,32767,0.6344480117162069
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,32767,2.6129706700642905
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,32767,0.8827040195465088
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,32767,1.5755093892415364
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,32767,3.0165065129597983
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,32767,0.8676426410675049
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,32767,0.8836800257364908
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,32767,1.5585120519002278
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,32767,0.87009064356486
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,32767,2.6441866556803384
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,32767,1.8901599248250325
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,32767,3.0803359349568686
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,32767,1.6864426930745442
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,32767,1.779322624206543
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,32767,1.6976213455200195
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,32767,6.064821243286133
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,32767,1.6792213122049968
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,32767,2.739210764567057
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,32767,4.81220277150472
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,32767,3.355855941772461
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,32767,3.320613225301107
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,32767,3.385024070739746
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,32767,3.3654187520345054
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,32767,8.866554896036783
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,32767,3.619488080342611
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,32767,15.043851216634115
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,32767,26.15416971842448
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,128,1,1,32767,52.036051432291664
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,32767,6.531967798868815
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,32767,6.571584065755208
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,32767,6.5029652913411455
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,32767,13.334378560384115
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,32767,7.131765365600586
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,32767,6.485626856486003
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,256,1,4,32767,26.32098642985026
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,256,1,2,32767,52.1307118733724
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,256,1,16,32767,13.26425552368164
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,256,1,8,32767,13.436271667480469
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,65535,1.2578453222910564
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,65535,1.286298672358195
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,65535,1.2481226921081543
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,65535,1.263978640238444
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,65535,1.1089973449707031
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,65535,1.1064799626668294
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,65535,1.1015893618265789
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,65535,1.3590985933939617
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,65535,3.568256060282389
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,65535,1.3905332883199055
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,65535,1.3576854070027669
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,65535,1.3598453203837078
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,65535,1.2074933052062988
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,65535,1.1952266693115234
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,65535,1.1921333471934001
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,65535,1.3726240793863933
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,65535,1.359605312347412
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,65535,1.4779574076334636
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,65535,5.118608156840007
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,65535,1.3682293891906738
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,65535,1.204095999399821
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,65535,1.1991093158721924
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,65535,1.1989280382792156
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,65535,1.3739573160807292
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,65535,1.5540587107340496
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,65535,5.138544082641602
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,65535,2.894511858622233
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,65535,1.375285307566325
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,65535,1.209546645482381
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,65535,1.2154773076375325
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,65535,1.203269322713216
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,65535,1.6749653816223145
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,65535,5.160319964090983
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,65535,3.0487305323282876
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,65535,5.8750559488932295
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,65535,1.6583093007405598
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,65535,1.660037358601888
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,65535,1.6499627431233723
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,65535,1.6514719327290852
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,65535,5.214693387349446
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,65535,6.04852294921875
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,65535,3.668405214945475
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,65535,11.966325124104818
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,65535,3.300784111022949
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,65535,3.289823849995931
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,65535,3.278287887573242
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,65535,3.283029238382975
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,65535,5.400058746337891
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,65535,9.598586400349935
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,65535,18.19748306274414
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,65535,6.613728205362956
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,65535,31.251429239908855
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,65535,6.524485270182292
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,65535,6.52122688293457
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,65535,6.527392069498698
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,65535,7.1514237721761065
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,128,1,4,65535,26.25367482503255
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,128,1,8,65535,12.946576436360678
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,128,1,2,65535,51.599212646484375
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,128,1,16,65535,12.897808074951172
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,128,1,32,65535,12.877263387044271
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,128,1,64,65535,12.85248057047526
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,128,1,128,65535,14.118970235188803
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,1,1,1,131071,2.6442346572875977
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,1,1,2,131071,2.6491519610087075
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,1,1,4,131071,2.6714347203572593
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,1,1,8,131071,2.649818738301595
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,1,1,16,131071,2.3367786407470703
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,1,1,32,131071,2.322970708211263
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,1,1,64,131071,2.3132853507995605
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,2,1,1,131071,2.6745761235555015
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,2,1,2,131071,2.6342612902323403
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,2,1,4,131071,2.6511893272399902
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,1,1,128,131071,10.200234731038412
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,2,1,8,131071,2.6523946126302085
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,2,1,16,131071,2.3474666277567544
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,2,1,32,131071,2.3339573542277017
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,2,1,64,131071,2.320687929789225
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,4,1,1,131071,2.8329652150472007
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,4,1,2,131071,2.6520214080810547
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,2,1,128,131071,10.229674657185873
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,4,1,8,131071,2.6589013735453286
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,4,1,4,131071,2.6436586380004883
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,4,1,16,131071,2.339178721110026
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,4,1,32,131071,2.326650619506836
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,4,1,64,131071,2.326746622721354
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,4,1,128,131071,10.215226491292318
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,8,1,1,131071,5.678576151529948
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,8,1,2,131071,3.002293268839518
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,8,1,4,131071,2.6711788177490234
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,8,1,16,131071,2.3568533261617026
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,8,1,8,131071,2.6742560068766275
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,8,1,32,131071,2.3485973676045737
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,8,1,64,131071,2.344501336415609
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,8,1,128,131071,10.260821024576822
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,16,1,1,131071,11.689983367919922
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,16,1,2,131071,6.006250381469727
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,16,1,8,131071,3.2395893732706704
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,16,1,4,131071,3.6750399271647134
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,16,1,16,131071,3.2445812225341797
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,16,1,32,131071,3.2380479176839194
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,16,1,64,131071,3.6273441314697266
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,16,1,128,131071,10.369743982950846
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,32,1,4,131071,8.01687494913737
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,32,1,1,131071,28.1125005086263
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,32,1,2,131071,13.312180836995443
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,32,1,8,131071,6.49073600769043
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,32,1,16,131071,6.496623992919922
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,32,1,32,131071,7.283258438110352
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,32,1,64,131071,6.476912180582683
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,32,1,128,131071,10.756287892659506
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,32,64,1,4,131071,20.562943776448567
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,64,64,1,2,131071,45.52254740397135
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,16,64,1,8,131071,12.959696451822916
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,8,64,1,16,131071,12.995333353678385
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,4,64,1,32,131071,12.960192362467447
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,128,64,1,1,131071,76.74679565429688
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,2,64,1,64,131071,12.949775695800781
VLLM,0.12.0,NVIDIA L40S,generation_mla,vllm_triton_mla,float16,float16,1,64,1,128,131071,14.243637084960938
