o `ZhPN@sddlZddlZddlZddlZddlZddlZddlmZddlm Z m Z m Z ddl m Zddl mZddZd d Zd d Zd*ddZd+ddZd,ddZGdddZGdddZddZd-ddZd-d d!Zd"d#Zed.d&d'Zd-d(d)ZdS)/N)contextmanager)AnyDictList)language)runtimecCsLd|}dddd|dg}t|}|tjjd}dd|D}|S) N, nvidia-smi-i0z --query-gpu=z--format=csv,noheader,nounitscSsg|]}t|qS)int.0xr r =/var/www/auris/lib/python3.10/site-packages/triton/testing.py znvsmi..)join subprocess check_outputdecodesysstdoutencodingsplit)attrscmdoutretr r rnvsmi s  r!cs0ttfddfdd|DS)Ncshd|kr dkstdtd|d}t|}t|}||}d||||S)Nrrz%Quantiles must be in the range [0, 1]) ValueErrormathfloorceil)qpointloweruppert)anr r get_quantiles   z_quantile..get_quantilecsg|]}|qSr r )rr&)r-r rr'rz_quantile..)lensorted)r+r&r )r+r-r,r _quantiles r0cCs~|durt||}t|dkr|d}|S|dkr|S|dkr#t|S|dkr+t|S|dkr4t|S|dkr=t|SdS)Nrrallminmaxmeanmedian)r0r.r2r3 statisticsr4r5)times quantiles return_moder r r r_summarize_statistics*s     r:r4c Csddl}|dvs J|j|j||dur,|D]}||dd|_q|jjdd}|jjdd}|t dD]} |qB||j | |d} t dt || } |j} |j| t | D]} |dur|D]}d|_qy|qqWdn1swY|j g} d}t |D]+} |jjdd}|jjdd}|| ||j | | || g7} qt| ||WdS1swYdS) a Benchmark the runtime of the provided function. :param fn: Function to benchmark :type fn: Callable :param rep: Repetition time (in ms) :type rep: int :param grad_to_none: Reset the gradient of the provided tensor to None :type grad_to_none: torch.tensor, optional :param return_mode: The statistical measure to return. Options are "min", "max", "mean", "median", or "all". Default is "mean". :type return_mode: str rNr2r3r4r5r1TZ enable_timingr )torchcudastreamZStreamZdetach_Zrequires_grad_gradEventrecordrange synchronize elapsed_timer3rZ CUDAGraphgraphZreplayr:)fnrep grad_to_noner8r9r@r start_event end_event_ estimate_msn_repeatgr Z n_retriesr r rdo_bench_cudagraph<sP          $rSdcsp|dvsJtjj|tjj}jdd}jdd}|tdD] } tjj ||q-|| |d} t dt || } t dt || } fddt| D}fddt| D}t| D]} |qut| D]$} |d ur|D]}d |_ qtjj ||| ||| qd dt||D}t|||S) a Benchmark the runtime of the provided function. By default, return the median runtime of :code:`fn` along with the 20-th and 80-th performance percentile. :param fn: Function to benchmark :type fn: Callable :param warmup: Warmup time (in ms) :type warmup: int :param rep: Repetition time (in ms) :type rep: int :param grad_to_none: Reset the gradient of the provided tensor to None :type grad_to_none: torch.tensor, optional :param quantiles: Performance percentile to return in addition to the median. :type quantiles: list[float], optional :param return_mode: The statistical measure to return. Options are "min", "max", "mean", "median", or "all". Default is "mean". :type return_mode: str r<Tr=r>rcg|]}jddqSTr=rDriZdir rrzdo_bench..crVrWrXrYr[r rrr\NcSsg|] \}}||qSr )rH)rser r rrs)rdriveractiveZget_device_interfacerGZget_empty_cache_for_benchmarkrDrErF clear_cacherHr3rrCzipr:)rJZwarmuprKrLr8r9cacherMrNrOrPZn_warmuprQrZrr7r r[rdo_bench{s>          rdc CsJddl}ddl}t||js||}t||js||}|dur$d}t|r-||jn|}|dur5d}t|r>||jn|}t||jrX|j|jkrP|}| }t||jrp|j|jkrh|}| }|j dksz|j dkr|j j ||||dddS|j||||dst|d |d |d |d |d dS)a Asserts that two inputs are close within a certain tolerance. :param x: The first input. :type x: scala, list, numpy.ndarray, or torch.Tensor :param y: The second input. :type y: scala, list, numpy.ndarray, or torch.Tensor :param atol: The absolute tolerance. Default value is 1e-2. :type atol: float, optional :param rtol: The relative tolerance. Default value is 0. :type rtol: float, optional :param err_msg: The error message to use if the assertion fails. :type err_msg: str rNg{Gz?grT)atolrtolZ equal_nan)rfrg z is not close to z (atol=z, rtol=))numpyr@ isinstanceZTensorZtensorcallabledtypebfloat16floatcpudetachsizetestingZassert_allcloseZallcloseAssertionError)ryrfrgerr_msgnpr@r r r assert_closes4        &rxc@sjeZdZdZ     ddeedeededeed eed ed eeefd ed ededefddZ dS) Benchmarkzk This class is used by the :code:`perf_report` function to generate line plots with a concise API. reFNx_namesx_valsline_arg line_vals line_names plot_nameargsxlabelylabelx_logy_logc CsL||_||_| |_||_||_||_| |_| |_||_| |_ ||_ ||_ dS)aq Constructor. x_vals can be a list of scalars or a list of tuples/lists. If x_vals is a list of scalars and there are multiple x_names, all arguments will have the same value. If x_vals is a list of tuples/lists, each element should have the same length as x_names. :param x_names: Name of the arguments that should appear on the x axis of the plot. :type x_names: List[str] :param x_vals: List of values to use for the arguments in :code:`x_names`. :type x_vals: List[Any] :param line_arg: Argument name for which different values correspond to different lines in the plot. :type line_arg: str :param line_vals: List of values to use for the arguments in :code:`line_arg`. :type line_vals: List[Any] :param line_names: Label names for the different lines. :type line_names: List[str] :param plot_name: Name of the plot. :type plot_name: str :param args: Dictionary of keyword arguments to remain fixed throughout the benchmark. :type args: Dict[str, Any] :param xlabel: Label for the x axis of the plot. :type xlabel: str, optional :param ylabel: Label for the y axis of the plot. :type ylabel: str, optional :param x_log: Whether the x axis should be log scale. :type x_log: bool, optional :param y_log: Whether the y axis should be log scale. :type y_log: bool, optional :param styles: A list of tuples, where each tuple contains two elements: a color and a linestyle. :type styles: list[tuple[str, str]] N) rzr{rr|r}r~rstylesrrrr) selfrzr{r|r}r~rrrrrrrr r r__init__s/ zBenchmark.__init__)rereFFN) __name__ __module__ __qualname____doc__rstrrrboolrr r r rrys<      ryc @s>eZdZddZ  ddedededefd d Zdd d ZdS)MarkcCs||_||_dSN)rJ benchmarks)rrJrr r rr5s z Mark.__init__Fbench save_path show_plots print_datac srddl}ddlm} ddl} |j} dd|jD} dd|jD} t|j}| j|| | | d}|jD]t tt fsGfdd|Dt t |kr[t dt |dt t|}ggg}}}|jD]<}|jdi||j|i|j|}z|\} } } Wnty|dd} } } Ynw|| g7}|| g7}|| g7}qmt||||jt |<q5|jrl| | }|d}t|jD][\}}||d ||d } } |jr|j|dnd}|jr|j|d nd}|j|||||||d | s*| s*| t} | t} |j||| | d |dq| |!|j"p6||#|j$|%|j&rFdnd|'|j(rQdnd|r[| )|rl| *|j+,||jd|||j}|r|j-d dkr|j./\}}|||||d<|rt0|jdt0|1|r|j2|j+,||jdd|ddd|S)NrcSg|]}|dqS)-minr rr r rr@zMark._run..cSr)-maxr rr r rrAr)columnscsg|]}qSr r )rrOrr rrGsz Expected z values, got rrr)labelcolorZlsg333333?)alpharlogZlinearz.pngDiff:z.csvz%.fF)Z float_formatindexr )3osZmatplotlib.pyplotZpyplotZpandasr~listrzZ DataFramer{rktupler.r"dictrbr}rJr|r TypeErrorlocrZfigureZsubplot enumeraterZplotZisnullr1ZastyperoZ fill_betweenZlegendZ set_xlabelrZ set_ylabelrZ set_xscalerZ set_yscalershowZsavefigpathrshapertolistprintZ to_stringZto_csv)rrrrrZdiff_colZsave_precisionZkwragsrZpltpdZy_meanZy_minZy_maxrzZdfZx_argsZrow_meanZrow_minZrow_maxrur axZfirst_xrZcolZstyZcol0Zcol1r rr_run9sz    $         "z Mark._runrec Kst|jt}|r |jgn|j}g}|r)tj|ddttj|dd} | d|D]} | |j | |||fi||rH| d| j dq+|rT| d| |r^|r\|d S|SdS) NT)exist_okz results.htmlwz z z r) rkrryrmakedirsopenrrwriteappendrrclose) rrrrZ return_dfkwargsZhas_single_benchrZ result_dfshtmlrr r rrun~s(   zMark.runN)Fr)FFreF) rrrrryrrrrr r r rr3s Ercfdd}|S)z Mark a function for benchmarking. The benchmark can then be executed by using the :code:`.run` method on the return value. :param benchmarks: Benchmarking configurations. :type benchmarks: List of :class:`Benchmark` cs t|Sr)r)rJrr rs zperf_report..r )rwrapperr rr perf_reports rcCs^ddl}ddlm}|s|j}|jj|d}|jj|d}||ddd }|S) z return DRAM bandwidth in GB/s rNrr_Zmem_clock_rateZ mem_bus_widthrg.A)r@rr_rAcurrent_devicer`utilsget_device_properties)devicer@r_Z mem_clock_khzZ bus_widthZbw_gbpsr r r get_dram_gbpss  rc Csddl}ddlm}|s|j}|jj|dd}|j|}|ddkr2||j ks/Jd}n+||j |j fvr=d}n ||j |j |j fvrJd}n||jtjtjtjfvrYd }ntd |||d }|S) Nrrrmultiprocessor_countriidtype not supported& .>)r@rr_rArr`rrget_device_capabilityfloat16float32Zint32rnZint16Zint8tlZ float8e4nvZ float8e4b15Zfloat8e5 RuntimeError rmZ clock_raterr@r_Z num_subcoresZ capabilityZops_per_sub_coretflopsr r rget_max_tensorcore_tflopss$    rc r)Ncstfdd}|S)Nc sddl}|t}|k}|rg|dkrgtjjd}tj ddd}d|vs4Jd|dj j j }|d j d |d }tjdd d |gd|d} | jdks\Jddt| jvseJdS|i|dS)Nrz cuda-memcheck__file__PATH1)rZPYTORCH_NO_CUDA_MEMORY_CACHINGrequestz@memcheck'ed test must have a (possibly unused) `request` fixturez::[]Zpytestz-vsT)capture_outputenvz7cuda-memcheck returned an error: bounds checking failedzERROR SUMMARY: 0 errors)psutilProcessrgetppidnameitemsrrealpath __globals__environnodeZcallspecidrrr returncoderr) rrrZ ppid_nameZrun_cuda_memcheckrrZtest_idrr) target_kwargstest_fnr rrs z1cuda_memcheck..decorator..wrapper) functoolswraps)rrr)rr decoratorsz cuda_memcheck..decoratorr )rrr rr cuda_memchecks rFc cs$zztgdtdddd|d|gtdddd|d|gtdgd }td gd }t||d ksEJd |d t||d ksUJd |d d|}d|d}||fVWtgdtgdtgddStgdtgdtgdw)N)r r r -pmrr r r z--lock-gpu-clocks=r z--lock-memory-clocks=zclocks.current.smrzclocks.current.memoryr?zGPU SMs must run at z MHzg3O?igMbP?)r r r rr )r r r z-rgc)r r r z-rmc)rrr!abs)Z ref_sm_clockZ ref_mem_clockZ cur_sm_clockZ cur_mem_clockrZgbpsr r r set_gpu_clocks8    rc Csddl}ddlm}|s|j}|jj|dd}|j}|ddkr;||j kr/d}n#||j kr7d}nt d ||j krCd}n||j |j fvrNd}nt d |||d }|S) Nrrrrrr @rr) r@rr_rArr`rrrrrrrnrr r rget_max_simd_tflopss&       r)r;NNr4)rTrUNNr4)NNrer)rr)rr#rr6rr contextlibrtypingrrrrerrrr!r0r:rSrdrxryrrrrrrrr r r rs2      ? B3Cc