2@ji.dZddlZddlZddlZddlmZddlmZm Z m Z m Z m Z ddl ZddlmZddlmZddlmZddlmZmZmZmZddlmZdd lmZdd lm Z dd l!m"Z"m#Z#ej$%d Z&ej$%d Z'ej$%dZ(dZ)e)Z*dZ+Gdde j,Z-dZ.Gdde j/Z0dZ1eZ2ej$%dZ3Gdde4Z5da6dZ7da8da9dZ:dZ;dZe>Z?e"e?d Z@d!ZAe"eAd"ZBd#ZCe"eCd$ZDe#d%ZEd&ZFe"eFd'ZGe#d(ZHd)ejIvZJeKejILd)dZMeJr eMr e<d*ZNd+ZOe"eNd,ZPe"eOd-ZQdS).a This file implements the code-generator for parallel-vectorize. ParallelUFunc is the platform independent base class for generating the thread dispatcher. This thread dispatcher launches threads that execute the generated function of UFuncCore. UFuncCore is subclassed to specialize for the input/output types. The actual workload is invoked inside the function generated by UFuncCore. UFuncCore also defines a work-stealing mechanism that allows idle threads to steal works from other threads. N)RLock) CFUNCTYPEc_intCDLLPOINTERc_uint)ir)as_dtype)typescgutilsconfigerrors) signature) _wrapper_info) ufuncbuilder)overload intrinsicdarwinlinuxwin32cHtj}|dkrtd|S)z* Gets the available thread count. z(Number of threads specified must be > 0.)r NUMBA_NUM_THREADS ValueError)ts HC:\Projects\Historical\venv\Lib\site-packages\numba/np/ufunc/parallel.pyget_thread_countr$s)  A1uuCDDD Hc !t|tsJtjd}tj| tj }|t j!tj!}tjtj tj tj!tj! g}| d} | d} d t|j|j} tj| || } | d} tj| | j\}}}}|}|}|} fdt1|j}t|jt js|dz }tjtj gd z!gd zz}t7j| |d }tjtj ||| g}t7j| ||j}| |jt7jjtjtjt jj gd }!|g}"| }fd ||||fD}!||g|z!fd||fDz|gz|#||$|%| &| | |tO| | j|jS)aWrap the original CPU ufunc/gufunc with a parallel dispatcher. This function will wrap gufuncs and ufuncs something like. Args ---- ctx numba's codegen context info: (library, env, name) inner function info sig type signature of the gufunc inner_ndim inner dimension of the gufunc (this is len(sig.args) in the case of a ufunc) Returns ------- wrapper_info : (library, env, name) The info for the gufunc wrapper. Details ------- The kernel signature looks like this: void kernel(char **args, npy_intp *dimensions, npy_intp* steps, void* data) args - the input arrays + output arrays dimensions - the dimensions of the arrays steps - the step size for the array (this is like sizeof(type)) data - any additional data The parallel backend then stages multiple calls to this kernel concurrently across a number of threads. Practically, for each item of work, the backend duplicates `dimensions` and adjusts the first entry to reflect the size of the item of work, it also forms up an array of pointers into the args for offsets to read/write from/to with respect to its position in the items of work. This allows the same kernel to be used for each item of work, with simply adjusted reads/writes/domain sizes and is safe by virtue of the domain partitioning. NOTE: The execution backend is passed the requested thread count, but it can choose to ignore it (TBB)! parallelgufuncwrapperzparallel.gufunc.wrapperz .kernel.{}_{})namec0|SN)bitcast)argbuilder byte_ptr_ts r as_void_ptrz(build_gufunc_kernel..as_void_ptrssJ///rrnumba_parallel_forget_num_threadsc&g|] }|Sr0).0xr*s r z'build_gufunc_kernel..s-555AQ555rc&g|] }|Sr0r0)r1r2intp_ts rr3z'build_gufunc_kernel..s!???&&))???r)libraryr"env)( isinstancetupler IntType PointerTypeget_value_typer intp FunctionTypeVoidTypecodegencreate_librarycreate_ir_moduleformatidr7r"Functionappend_basic_block IRBuilderargsget_python_api gil_ensure save_threadlen return_typeNoneTyper get_or_insert_functionadd_linking_libraryr6modulebitwidthcallr&restore_thread gil_releaseret_void add_ir_moduler)"r6ctxinfosig inner_ndimbyte_tbyte_ptr_ptr_t intp_ptr_tfnty wrapperlibmod kernel_namelfuncbb_entryrH dimensionsstepsdatapyapi gil_state thread_state array_countparallel_for_ty parallel_forinnerfunc_fnty tmp_voidptrr. num_threadsfnptr innerargsr*r(r)r5s" @@@@rbuild_gufunc_kernelrs1s` dE " """ " Z]]F''J^J//N    + +F''J ?2;==2>*+E+E+->&+A+A+->&+A+A+5+7 8 8D--.EFFJ  % %&? @ @C!((DHtyAAK KT 4 4 4E''++Hl8$$G$)J!D*eT   w ' 'E  ""I$$&&L000000ch--K cou~ 6 6q obkmm'1lQ&6&a&GIIO1#2FHHL_ Z<N0n15==K""4<0004  5:#677<<O ,,33K OOK 4 4E5555:ud3555I LLw2????j+%>??? @CN- PQQQ &&& i    S!!!""7+++ %*$( K K KKrceZdZdZdS)ParallelUFuncBuilderc`t|j}|j}|j}|jj}t |||||}|j|j}d|j D} | tj |j jjd} | || fS)NcJg|] }tj|jj!Sr0)npdtyper"num)r1as rr3z.ParallelUFuncBuilder.build..s'BBBaRXaf%%)BBBrr0)_launch_threadstarget_contextrr6fndescllvm_func_namebuild_ufunc_wrapperget_pointer_to_functionr"rHappendrxryrMrz) selfcresrZrXrr6fnamerYptr dtypenums keepalives rbuildzParallelUFuncBuilder.builds!N , *"7C 4HHl2249==BB9>BBB )"7"<==ABBB #y((rN)__name__ __module__ __qualname__rr0rrrurus#)))))rruc tj||||d|}t||||t|j}|S)NF)objmoder)rrrsrLrH)r6rXrrr innerfuncrYs rrrsR0#u1:E6:<<22 4 4D Krc*eZdZ dfd ZdZxZS)ParallelGUFuncBuilderNFr0c|i}|tdtt|||||||dS)NT)nopython)py_funcridentitycache targetoptions writable_args)updatedictsuperr__init__)rrrrrrr __class__s rrzParallelGUFuncBuilder.__init__sw  MT4000111 !   ('' )))))rc~tt|j||j|j|jd}|j|j}|j }g}|j j D]M}t|tjr|j}n|}|t#|jN|||fS)zJ Returns (dtype numbers, function ptr, EnvironmentObject) Fr is_parfors)r|build_gufunc_wrapperrsinsoutrr6rr"r7rrHr8r Arrayryrr rz)rrrYrr7rr{tys rrzParallelGUFuncBuilder.builds $ L$$)4:   l2249==h $ / /A!U[)) W   Xb\\- . . . .#s""r)NFNr0)rrrrr __classcell__)rs@rrrsR@E35)))))) #######rrc|j}|j}|j}tj||||||} t d|D} t d|D} t | | z} t||| || } | S)zBuild gufunc wrapper for the given arguments. The *is_parfors* is a boolean indicating whether the gufunc is being built for use as a ParFors kernel. This changes codegen and caching behavior. rc3$K|] }|D]}|V dSr%r0r1termsyms r z'build_gufunc_wrapper.. s/55555555555rc3$K|] }|D]}|V dSr%r0rs rrz'build_gufunc_wrapper..s/77$$773#7777777r)r6r}rrrsetrLrs)rrrrrrr6rXr innerinfosym_insym_outr[rYs rrrs lG  CI1sD*I55555 5 5F77$77777GVg%&&J iJ  D KrceZdZdZdZdZdS)_nopzA no-op contextmanager cdSr%r0)rs r __enter__z_nop.__enter__" rcdSr%r0)rrHs r__exit__z _nop.__exit__%rrN)rrr__doc__rrr0rrrrs<        rrc t5ddl}d|vstr)|}|antaddddS#1swxYwYdS#t$rJ}d}tj |t|ztj taYd}~dSd}~wwxYw)NrforkaACould not obtain multiprocessing lock due to OS level error: %s A likely cause of this problem is '/dev/shm' is missing or read-only such that necessary semaphores cannot be written. *** The responsibility of ensuring multiprocessing safe access to this initialization sequence/module import is deferred to the user! *** )_backend_init_thread_lockmultiprocessingget_start_method_windows get_contextr_backend_init_process_lockrOSErrorwarningswarnstrrNumbaSystemWarning)rrXemsgs r_set_init_process_lockr,s/, ' 4 4 " " " "99;;;;x;%1133-0YY[[**-1VV*  4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4  , , ,    cCFFlF$=>>>%)VV""""""" ,s;A<AA/" A</A33A<6A37A<< C?C  CFc<ttdtS)zM Get the name of the threading layer in use for parallel CPU targets Nz#Threading layer is not initialized.)_threading_layerrr0rrthreading_layerrQs>???rc trd}n#trd}ntrd}ntdt |}|j}g|_t|_|}|dkr?d|z}tj |}tj |td|zdS#ttf$r}td|zd}~wwxYw) zF Checks that if TBB is present it is of a compatible version. z tbb12.dllzlibtbb.12.dylibz libtbb.so.12zUnknown operating systemi/zThe TBB threading layer requires TBB version 2021 update 6 or later i.e., TBB_INTERFACE_VERSION >= 12060. Found TBB_INTERFACE_VERSION = %s. The TBB threading layer is disabled.zProblem with TBB. Reason: %sN) _IS_WINDOWS_IS_OSX _IS_LINUXrrTBB_runtime_interface_versionargtypesrrestyper NumbaWarningrr ImportErrorr) libtbb_namelibtbb version_func tbb_iface_verrproblemrs r_check_tbb_version_compatibler[s >  9%KK  9+KK  9(KK788 8k""; " $ $  5 26C CC )#..G M' " " ">>81<===>sB%B))C:C  Cc tstt5t5tr dddddddSd fd}t t j}t j}t|dkrt|hdkstd|zd}t d d<d d <g} fd }||vrO |}|s>|d kr| dn"|d krtr| d |}n.|d vrd g}| d|dkrn|dkr2tr| d | d nh|dkrNts| d tr| d | dnd}t||z||\}}n\|dkrB||\}}|s1| dtr| d nd}t||z|s ||t!jd|jt!jd|jt!jd|jt!jd|jt!jd|jt/dt0|j} | t4t7||adadddn #1swxYwYddddS#1swxYwYdS)Nc:d}|dr& tddlm}nm#t$rYnawxYw|dr ddlm}n@#t$rYn4wxYw|drddlm}nd }t||z|S) zZ Loads a specific threading layer backend based on string Ntbbr)tbbpoolomp)omppool workqueue)rz/Unknown value specified for threading layer: %s) startswithrnumba.np.ufuncrrrrr)backendlibrs rselect_known_backendz-_launch_threads..select_known_backends%%e,,45777AAAAAAA&''.. 4AAAAAAA&'' 444???????KC$S7]333 s. ;;A A('A(c<d}|D]}|}|nd}||fS)z_ Selects from presented backends and returns the first working Nr#r0)backendsrrrs rselect_from_backendsz-_launch_threads..select_from_backendssI'!!G..w77C'!GG|#rr,>rrrz]THREADING_LAYER_PRIORITY invalid: %s. It must be a permutation of {'tbb', 'omp', 'workqueue'}z3Intel TBB is required, try: $ conda/pip install tbbTBBz?Intel OpenMP is required, try: $ conda/pip install intel-openmpOSX_OMPcd}d}t|dkrd}t|dkr||dz}t|dkr)dfd|D}|d|zz}t||z) Nz&No threading layer could be loaded. %szHINT: %srr#rz OR c g|] }| Sr0r0)r1r2 err_helperss rr3z<_launch_threads..raise_with_hint..s,N,N,N[^,N,N,Nrz One of: %s)rLjoinr)requirederrmsghintmsghintoptionsrs rraise_with_hintz(_launch_threads..raise_with_hintsB%x==A%%Dx==A%%"[!%==Dx==1$$&mm,N,N,N,NX,N,N,NOOG"mg&=>D $///rrr) threadsafeforksafesaferrrrz+No threading layer available for purpose %sdefaultz7The threading layer requested '%s' is unknown to Numba.r-do_scheduling_signeddo_scheduling_unsignedallocate_scheddeallocate_schedT)rrr_is_initializedrr THREADING_LAYERlowerTHREADING_LAYER_PRIORITYrLrrrrrrll add_symbolrmrrrrrrlaunch_threads NUM_THREADS_load_threading_functionsr) rr namedbackendsr requirementsrlibname availablerrrrs @@rr|r|}s %!   #R#R# &Q #Q # Q #Q #Q #Q #Q #Q #R#R#R#R#R#R#R#R#    6 $ $ $ $ $F*++1133A";M &&!++ &&*E*E*EEE 2$$C&&K#> WWi 43MBB W7 ''...7$++I666O q))) . --- M.0@ A A A M0#2J K K K M2C4N O O O M*C,> ? ? ? M,c.B C C C3YtU33C4FGGN N; ' ' ' %c * * * ' "OcQ #Q #Q #Q #Q #Q #Q #Q #Q #Q #Q #Q #Q #Q #Q #R#R#R#R#R#R#R#R#R#R#R#R#R#R#R#R#R#R#s@M M MK0M8 MM M M MM #M c ptjd|jtjd|jtjd|jt dt |jattt t |ja t t |ja tjd|j tjd|j tjd|j t tt|j at t|j at tttt#t t#t |j adS)Nr.set_num_threads get_thread_idset_parallel_chunksizeget_parallel_chunksizeget_sched_size)rrr.r rrr_set_num_threadsr_get_num_threads_get_thread_idrrrr_set_parallel_chunksize_get_parallel_chunksizer_get_sched_size)rs rrrsxM#S%8999M#S%8999M/3#4555.yu--c.ABB[!!!(y''(;<<&Yu%%c&788NM*C,FGGGM*C,FGGGM"C$67770i(.00030JLL0i//0JKK0i & & ' ' 00140B DDOOOrc,ddlmdzfd}|S)Nr)rz.The number of threads must be between 1 and %sc>|ks|dkrtdS)Nr)r)nrrs r snt_checkz gen_snt_check..snt_check<s*  AEES// !%*Er)numba.core.configr)rrrs @@r gen_snt_checkr8sG333333 :=N NC"""""" rctSr%)rrs r ol_snt_checkr Es rctt|ttjfst dt |t|dS)a Set the number of threads to use for parallel execution. By default, all :obj:`numba.config.NUMBA_NUM_THREADS` threads are used. This functionality works by masking out threads that are not used. Therefore, the number of threads *n* must be less than or equal to :obj:`~.NUMBA_NUM_THREADS`, the total number of threads that are launched. See its documentation for more details. This function can be used inside of a jitted function. Parameters ---------- n: The number of threads. Must be between 1 and NUMBA_NUM_THREADS. See Also -------- get_num_threads, numba.config.NUMBA_NUM_THREADS, numba.config.NUMBA_DEFAULT_NUM_THREADS, :envvar:`NUMBA_NUM_THREADS` 2The number of threads specified must be an integerN)r|r8intrxinteger TypeErrorrrrs rr r JsX. a#rz* + +NLMMM aLLLQrctt|tjsd}t j|d}|S)Nr"cBt|t|dSr%)rrrs rimplz ol_set_num_threads..implos!! rr|r8r Integerr TypingErrorrrr(s rol_set_num_threadsr-hsN a ' '&B %%% Krctt}|dkr"tdtd|d|S)aM Get the number of threads used for parallel execution. By default (if :func:`~.set_num_threads` is never called), all :obj:`numba.config.NUMBA_NUM_THREADS` threads are used. This number is less than or equal to the total number of threads that are launched, :obj:`numba.config.NUMBA_NUM_THREADS`. This function can be used inside of a jitted function. Returns ------- The number of threads. See Also -------- set_num_threads, numba.config.NUMBA_NUM_THREADS, numba.config.NUMBA_DEFAULT_NUM_THREADS, :envvar:`NUMBA_NUM_THREADS` rzLInvalid number of threads. This likely indicates a bug in Numba. (thread_id=z, num_threads=))r|r RuntimeErrorrrps rr.r.usZ,"$$Kal*OOOO[[[:;; ; rc(td}|S)Nct}|dkr;tdttd|td|S)NrzBroken thread_id: z num_threads: z@Invalid number of threads. This likely indicates a bug in Numba.)rprintrr0r1s rr(z ol_get_num_threads..impls^&(( !   &  8 8 8 /; / / / GHH Hrr|r(s rol_get_num_threadsr7s$ KrcZtd}ttj|fS)Nc|j}tjtjg}tj||d}||gS)Nr.rQr r>r r5rOrScontextr(rrHrar_fns rr@z"_iget_num_threads..codegensEnw~r22  +C7H I I||B###r)r|rr r= typingctxr@s r_iget_num_threadsr@s4$$$ UZ ' ))rc:ttS)zv Returns a unique ID for each thread in the range 0 (inclusive) to :func:`~.get_num_threads` (exclusive). )r|rr0rrrrs   rc(td}|S)NctSr%)_iget_thread_idr0rrr(zol_get_thread_id..impls   rr5r6s rol_get_thread_idrEs"!!! Krc>d}ttj|fS)Nc|j}tjtjg}tj||d}||gS)Nrr:r;s rr@z _iget_thread_id..codegensDnw~r22  +C G G||B###r)rr r=r>s rrDrDs($$$ UZ ' ))rNUMBA_DYLD_WORKAROUNDctt|ttjfst d|dkrt dt|S)N)The parallel chunksize must be an integerr/chunksize must be greater than or equal to zero)r|r8r#rxr$r%rrrs rrrs] a#rz* + +ECDDD1uuJKKK "1 % %%rc:ttSr%)r|rr0rrrrs " $ $$rctt|tjsd}t j|d}|S)NrJcJ|dkrtdt|S)NrrK)rrrs rr(z'ol_set_parallel_chunksize..impls( q55NOO O&q)))rr)r,s rol_set_parallel_chunksizerOsN a ' '&9 %%%*** Krc(td}|S)NctSr%)rr0rrr(z'ol_get_parallel_chunksize..impls&(((rr5r6s rol_get_parallel_chunksizerRs"))) Kr)Rrossysr threadingr threadRLockctypesrrrrrnumpyrxllvmlite.bindingbindingrllvmliter numba.np.numpy_supportr numba.corer r r rnumba.core.typingrnumba.np.ufunc.wrappersrrrnumba.extendingrrplatformrrrrrrrs UFuncBuilderrur GUFuncBuilderrrrrobjectrrrrrrrr|rrrr r r-r.r7r@rrErDenviron_DYLD_WORKAROUND_SETr#get_DYLD_WORKAROUND_VALrrrOrRr0rrris   ******::::::::::::::++++++555555555555''''''111111''''''//////// , ! !( + + L # #G , , l%%g..      LLLH)))))<4)))&(#(#(#(#(#L6(#(#(#\0(KMM < " "7 + +     6   ",,,>   >>>DV#V#V#rDDDD MOO  )< /   @ /    ** *    - ** */"*<s2:>>*A1EEFF0O&&&%%%   !!  "!   !!"!r