Question

我在不同机器上都遇到python可再现性问题。我有两台计算机：装有SSE4_2的旧服务器和装有AVX的新台式机。在两台机器上运行相同的python脚本会导致不同的输出（我确保两个virtualenv都完全相同）。

我实际上找到了一个解决方案：由于这些帖子，我认为问题是由于指令集的差异造成的：

他们建议我通过设置os.environ ['MKL_CBWR'] ='SSE4_2'来使用MKL强制使用哪个指令集。我卸载了numpy并安装了intel-numpy（使用mkl构建的numpy），以便可以设置os.environ ['MKL_CBWR'] ='SSE4_2'。现在，两台计算机的脚本输出相同。

但是，即使结果匹配，它们也不匹配结果pre mkl（使用纯numpy）。能够使用intel-numpy从服务器重现旧结果将对我有很大帮助；我进行了许多实验，耗时一周以上，并且我希望能够使用这些结果而无需重新运行它们。

编辑：我进行了更多研究，我认为问题可能出在Lapack和Blas实施中。我认为MKL和OpenBlas（这是我的默认numpy所使用的）具有不同的实现，导致不同的操作顺序。因此，也许我的解决方案是继续使用默认的numpy，但将指令集强制为台式机上的SSE4_2。 是否可以不使用MKL强制使用“ SSE4_2”？

我是否需要从源代码编译numpy并从源代码编译openblas，并确保已禁用avx？如果是这样，有人能指出我正确的入门方向吗？我找到了一些资源，但是似乎有很多方法可以从源代码进行编译。

如果我的问题不清楚，请告诉我，我会尽力澄清。

其他详细信息：当我使用默认的numpy（而不是intel-numpy）运行np.__config__.show()时，我得到：

lapack_opt_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/usr/local/lib']
    define_macros = [('HAVE_CBLAS', None)]
    language = c
blas_opt_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/usr/local/lib']
    define_macros = [('HAVE_CBLAS', None)]
    language = c
openblas_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/usr/local/lib']
    define_macros = [('HAVE_CBLAS', None)]
    language = c
blis_info:
  NOT AVAILABLE
openblas_lapack_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/usr/local/lib']
    define_macros = [('HAVE_CBLAS', None)]
    language = c
lapack_mkl_info:
  NOT AVAILABLE
blas_mkl_info:
  NOT AVAILABLE

当我使用intel-numpy运行np.__config__.show()时，我得到：

lapack_opt_info:
    libraries = ['mkl_rt', 'pthread']
    library_dirs = ['/opt/anaconda1anaconda2anaconda3/lib']
    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]
    include_dirs = ['/opt/anaconda1anaconda2anaconda3/include']
blas_opt_info:
    libraries = ['mkl_rt', 'pthread']
    library_dirs = ['/opt/anaconda1anaconda2anaconda3/lib']
    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]
    include_dirs = ['/opt/anaconda1anaconda2anaconda3/include']
lapack_mkl_info:
    libraries = ['mkl_rt', 'pthread']
    library_dirs = ['/opt/anaconda1anaconda2anaconda3/lib']
    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]
    include_dirs = ['/opt/anaconda1anaconda2anaconda3/include']
blas_mkl_info:
    libraries = ['mkl_rt', 'pthread']
    library_dirs = ['/opt/anaconda1anaconda2anaconda3/lib']
    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]
    include_dirs = ['/opt/anaconda1anaconda2anaconda3/include']
mkl_info:
    libraries = ['mkl_rt', 'pthread']
    library_dirs = ['/opt/anaconda1anaconda2anaconda3/lib']
    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]
    include_dirs = ['/opt/anaconda1anaconda2anaconda3/include']

在每台计算机上运行lshw会导致：

旧服务器：

    description: Computer

        width: 64 bits

        capabilities: vsyscall32

    *-core
       description: Motherboard
       physical id: 0
     *-memory
          description: System memory
          physical id: 0
          size: 47GiB
     *-cpu:0
          product: Intel(R) Xeon(R) CPU           X5650  @ 2.67GHz
          vendor: Intel Corp.
          physical id: 1
          bus info: cpu@0
          width: 64 bits
          capabilities: fpu fpu_exception wp vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp x86-64 constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm pcid dca sse4_1 sse4_2 popcnt aes lahf_lm pti tpr_shadow vnmi flexpriority ept vpid dtherm ida arat
     *-cpu:1
          product: Intel(R) Xeon(R) CPU           X5650  @ 2.67GHz
          vendor: Intel Corp.
          physical id: 2
          bus info: cpu@1
          width: 64 bits
          capabilities: fpu fpu_exception wp vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp x86-64 constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm pcid dca sse4_1 sse4_2 popcnt aes lahf_lm pti tpr_shadow vnmi flexpriority ept vpid dtherm ida arat
     *-pci:0
          description: Host bridge
          product: 5500 I/O Hub to ESI Port
          vendor: Intel Corporation
          physical id: 100
          bus info: pci@0000:00:00.0
          version: 13
          width: 32 bits
          clock: 33MHz

新桌面：

    width: 64 bits

    capabilities: vsyscall32

    *-core
       description: Motherboard
       physical id: 0
     *-memory
          description: System memory
          physical id: 0
          size: 31GiB
     *-cpu
          product: Intel(R) Core(TM) i5-6600 CPU @ 3.30GHz
          vendor: Intel Corp.
          physical id: 1
          bus info: cpu@0
          size: 1250MHz
          capacity: 3900MHz
          width: 64 bits
          capabilities: fpu fpu_exception wp vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp x86-64 constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb invpcid_single pti ssbd ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx rdseed adx smap clflushopt intel_pt xsaveopt xsavec xgetbv1 xsaves dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp flush_l1d cpufreq
     *-pci
          description: Host bridge
          product: Sky Lake Host Bridge/DRAM Registers
          vendor: Intel Corporation
          physical id: 100
          bus info: pci@0000:00:00.0
          version: 07
          width: 32 bits
          clock: 33MHz
          configuration: driver=skl_uncore
          resources: irq:0

如何使numpy使用SSE4_2而不是AVX？

0 个答案: