ó \ K]c@`sWddlmZmZmZddlZddlZddlmZm Z m Z ddl m Z m Z mZddlmZd„Zd„Zd„Zd „Zd „Zd „Zd „Zd „Zd„Zd„Zd„Zd„Zd„Zd„Zd„Zd„Z d„Z!d„Z"de e j#fd„ƒYZ$e%dkrSe j&ƒndS(i(tprint_functiontabsolute_importtdivisionN(tcudatint32tfloat32(tunittestt SerialMixintskip_on_cudasim(tIS_PY3cC`stjj}||di( RR?RR@RARtarangeRBtall(RCRDtNR texp((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_fill_threadidx{s c`skd\‰‰‰‡‡‡fd†}‡‡‡fd†}|ƒ}|ƒ}|jtj||kƒƒdS(Niiic`sWtjdƒtƒ}tjˆˆˆfdtjƒ}|dˆˆˆff|ƒ|S(Nsvoid(int32[:,:,::1])R>i(RR?RR@tzerosR(RDR (tXtYtZ(sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt c_contigous†s!c`s`tjdƒtƒ}tjtjˆˆˆfdtjƒƒ}|dˆˆˆff|ƒ|S(Nsvoid(int32[::1,:,:])R>i(RR?RR@tasfortranarrayRKR(RDR (RLRMRN(sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt f_contigousŒs*(iii(RBR@RG(RCRORQtc_restf_res((RLRMRNsFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_fill3d_threadidxƒs   cC`s€tjdƒtƒ}d\}}||}tj|dtjƒ}|||f|ƒ|jtj|tj|ƒkƒƒdS(Nsvoid(int32[::1])iiR>(ii( RR?RR@temptyRRBRGRF(RCRDtntidtnctaidtnelemR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_simple_grid1d–s   c C`sítjdƒtƒ}d }d }|d|d|d|df}tj|dtjƒ}|jƒ}|||f|ƒxMt|jdƒD]8}x/t|jdƒD]}|||||f(ii(ii( RR?RR@RURtcopyRRRBRG( RCRDRVRWRR RIR R((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_simple_grid2džs$ cC`sitjdƒtƒ}d\}}tjddtjƒ}|||f|ƒ|j|d||ƒdS(Nsvoid(int32[::1])iiiR>i(ii(RR?RR@RKRt assertEqual(RCRDRVRWR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_simple_gridsize1d­s  sTests PTX emissionc C`sctjdƒtƒ}tjdƒtƒ}d}d}tjdddddtjƒ}|jƒ}d|d *tj|dtjƒ}||d f|||ƒ|j ƒ}|j d t t j d |ƒƒƒtjj||d dƒtj|dtjƒ}||d f|||ƒ|j ƒ}|j dt t j d |ƒƒƒtjj||d dƒdS(Nsvoid(i8[:], i8, i8[:])i iRt fill_valueiR>iiiis \s+bra\s+terr_msgt branchingiR6(RR?R5R8R@tfulltint64RZRFt inspect_asmR\tlentretfindallttestingtassert_array_equal( RCtcu_branching_with_ifstcu_branching_with_selpstnR+RtexpectedR*tptx((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_selp´s"!   " "cC`s‘tjdƒtƒ}d }d }tjddtjƒ}|||f|ƒ|j|d|d|dƒ|j|d |d |d ƒdS( Nsvoid(int32[::1])iiiiiR>ii(ii(ii(RR?RR@RKRR\(RCRDRVRWR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_simple_gridsize2dÌs cC`sKtjdƒtƒ}d }d }|d|d|d|df}tj|dtjƒ}|||f|ƒ|\}}|j\}} x¸tt|dƒt|dƒƒD]“\} } || || } } xot| | |ƒD][}xRt| ||ƒD]>}|j |||f||k|||f||fƒqýWqäWq°WdS( Nsvoid(float32[:,::1])iiiiiiR>(ii(ii( RR?R&R@RURRtzipRRB(RCRDRVRWRR R"R#R$R%R RR R!R R((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_intrinsic_forloop_stepÖs$ 0cC`sdtjd„ƒ}tjddtjƒjdddƒ}|ddf|ƒtjj|dƒdS( NcS`sOtjdƒ\}}}tjdƒ\}}}|||||||fiÙ(iii(iiiiÙ(RR?R@RKRtreshapeRgt assert_equal(RCRstarr((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_3dgridçs'cC`sƒtjd„ƒ}ddd}}}tj|||dtjƒj|||ƒ}|d d f|ƒ|jtj|ƒƒdS( Nc S`stjdƒ\}}}tjdƒ\}}}|tjjtjjtjjkoŸ|tjjtjjtjjkoŸ|tjjtjjtjjk}|tjjtj jkoó|tjjtj jkoó|tjjtj jk}|oÿ|||||fi ii(iii(iii(RR?R@RKtboolRtRBRG(RCRsR RRRv((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_3dgrid_2ós /cC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(Nsvoid(int32[:], uint32)iR>iðii(RR?R(R@RKRt assertEquals(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_popc_u4s cC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(Nsvoid(int32[:], uint64)iR>Iðii(RR?R(R@RKRR}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_popc_u8s cC`s[tjdƒtƒ}tjddtjƒ}||dddƒtjj|dd ƒdS( Nsvoid(f4[:], f4, f4, f4)iR>g@g@g@iiiiii (RR?R,R@RKRRgtassert_allclose(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_fma_f4scC`s[tjdƒtƒ}tjddtjƒ}||dddƒtjj|dd ƒdS( Nsvoid(f8[:], f8, f8, f8)iR>g@g@g@iiiiii (RR?R,R@RKtfloat64RgR€(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_fma_f8scC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(Nsvoid(uint32[:], uint32)iR>ið0ii (RR?R.R@RKtuint32R}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_brev_u4 s s.only get given a Python "int", assumes 32 bitscC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(Nsvoid(uint64[:], uint64)iR>Ið0ð0iI  (RR?R.R@RKtuint64R}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_brev_u8&s cC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(Nsvoid(int32[:], int32)iR>iii (RR?R0R@RKRR}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_clz_i4-s cC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(s² Although the CUDA Math API (http://docs.nvidia.com/cuda/cuda-math-api/group__CUDA__MATH__INTRINSIC__INT.html) only says int32 & int64 arguments are supported in C code, the LLVM IR input supports i8, i16, i32 & i64 (LLVM doesn't have a concept of unsigned integers, just unsigned operations on integers). http://docs.nvidia.com/cuda/nvvm-ir-spec/index.html#bit-manipulations-intrinics svoid(int32[:], uint32)iR>iii N(RR?R0R@RKR„R}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_clz_u43s cC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(Nsvoid(int32[:], int32)iR>Iÿÿÿÿi(RR?R0R@RKRR}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_clz_i4_1s@s cC`sUtjdƒtƒ}tjddtjƒ}||dƒ|j|dddƒdS(Nsvoid(int32[:], int32)iR>ii sCUDA semantics(RR?R0R@RKRR}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_clz_i4_0sFs cC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(Nsvoid(int32[:], int64)iR>iii/(RR?R0R@RKRR}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_clz_i8Ls cC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(Nsvoid(int32[:], int32)iR>iii(RR?R2R@RKRR}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_ffs_i4Ss cC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(Nsvoid(int32[:], uint32)iR>iii(RR?R2R@RKR„R}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_ffs_u4Ys cC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(Nsvoid(int32[:], int32)iR>Iÿÿÿÿi(RR?R2R@RKRR}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_ffs_i4_1s_s cC`sUtjdƒtƒ}tjddtjƒ}||dƒ|j|dddƒdS(Nsvoid(int32[:], int32)iR>ii sCUDA semantics(RR?R2R@RKRR}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_ffs_i4_0ses cC`sRtjdƒtƒ}tjddtjƒ}||dƒ|j|ddƒdS(Nsvoid(int32[:], int64)iR>iii(RR?R2R@RKRR}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_ffs_i8ks cC`s“tjdƒtƒ}d}tj|ddtjƒ}tjtjddtjƒ|ƒ}|d|df|ƒ|jtj ||kƒƒdS(Nsvoid(int32[:])ii R>i( RR?R:R@RKRttileRFRBRG(RCRDtcountR RI((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_simple_laneidrs $cC`sRtjdƒtƒ}tjddtjƒ}||ƒ|j|dddƒdS(Nsvoid(int32[:])iR>ii sCUDA semantics(RR?R<R@RKRR}(RCRDR ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyttest_simple_warpsizezs sround() returns float on Py2cC`stjdƒtƒ}tjddtjƒ}xMdddddd d d gD]-}|||ƒ|j|d t|ƒƒqLWdS( Nsvoid(int64[:], float32)iR>gÀgÀgÀgø¿gø?g@g@g@i(RR?R4R@RKRR}R3(RCRDR R ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_round_f4€s % cC`stjdƒtƒ}tjddtjƒ}xMdddddd d d gD]-}|||ƒ|j|d t|ƒƒqLWdS( Nsvoid(int64[:], float64)iR>gÀgÀgÀgø¿gø?g@g@g@i(RR?R4R@RKRR}R3(RCRDR R ((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyt test_round_f8‰s % (%t__name__t __module__RERJRTRYR[R]RRnRoRqRwR|R~RRRƒR…R‡RˆR‰RŠR‹RŒRRŽRRR‘R”R•Rt skipUnlessR R–R—(((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyR=ts>                       t__main__('t __future__RRRtnumpyR@RetnumbaRRRtnumba.cuda.testingRRRt numba.utilsR RRRRRRRR&R(R,R.R0R2R4R5R8R:R<tTestCaseR=R˜tmain(((sFlib/python2.7/site-packages/numba/cuda/tests/cudapy/test_intrinsics.pyts6                  ÿ