connorgoggins opened a new issue #17716: [Large Tensor] linalg ops fail w/input dim >= 2**32 URL: https://github.com/apache/incubator-mxnet/issues/17716 ## Description While testing the `linalg_*` ops on large tensor (dimension >= 2**32) data, I found that all of these ops fail with a segmentation fault on large tensor data. In a test run with the `linalg_det` op, I traced the error to line 485 of `src/operator/tensor/la_op-inl.h`. This line lies within the `Map` void function which takes in several parameters, namely an `int` i, an `int` N, and an `int*` pivot. The error is thrown in the iteration portion of the function, where an `int` j is incremented up to the value of `int` N. This error occurred irrespective of which BLAS engine MXNet was built with (MKL or OpenBLAS). ## Environment ``` ----------Python Info---------- Version : 3.6.6 Compiler : GCC 7.2.0 Build : ('default', 'Jun 28 2018 17:14:51') Arch : ('64bit', '') ------------Pip Info----------- Version : 19.3.1 Directory : /home/ubuntu/anaconda3/lib/python3.6/site-packages/pip ----------MXNet Info----------- Version : 1.6.0 Directory : /home/ubuntu/forked-mxnet/python/mxnet Num GPUs : 0 Hashtag not found. Not installed from pre-built package. ----------System Info---------- Platform : Linux-4.4.0-1102-aws-x86_64-with-debian-stretch-sid system : Linux node : ip-172-31-41-238 release : 4.4.0-1102-aws version : #113-Ubuntu SMP Wed Jan 29 14:54:54 UTC 2020 ----------Hardware Info---------- machine : x86_64 processor : x86_64 Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Byte Order: Little Endian CPU(s): 96 On-line CPU(s) list: 0-95 Thread(s) per core: 2 Core(s) per socket: 24 Socket(s): 2 NUMA node(s): 2 Vendor ID: GenuineIntel CPU family: 6 Model: 85 Model name: Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz Stepping: 7 CPU MHz: 2499.998 BogoMIPS: 4999.99 Hypervisor vendor: KVM Virtualization type: full L1d cache: 32K L1i cache: 32K L2 cache: 1024K L3 cache: 36608K NUMA node0 CPU(s): 0-23,48-71 NUMA node1 CPU(s): 24-47,72-95 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm $ onstant_tsc arch_perfmon rep_good nopl xtopology nonstop_tsc aperfmperf tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe $ opcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single kaiser fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms in vpcid mpx avx512f rdseed adx smap clflushopt clwb avx512cd xsaveopt xsavec xgetbv1 ida arat pku ``` ### MXNet build flags #### BLAS = MKL ✖ CUDA, ✖ CUDNN, ✖ NCCL, ✖ CUDA_RTC, ✖ TENSORRT, ✔ CPU_SSE, ✔ CPU_SSE2, ✔ CPU_SSE3, ✔ CPU_SSE4_1, ✔ CPU_SSE4_2, ✖ CPU_SSE4A, ✔ CPU_AVX, ✖ CPU_AVX2, ✔ OPENMP, ✖ SSE, ✔ F16C, ✖ JEMALLOC, ✖ BLAS_OPEN, ✖ BLAS_ATLAS, ✔ BLAS_MKL, ✖ BLAS_APPLE, ✔ LAPACK, ✖ MKLDNN, ✖ OPENCV, ✖ CAFFE, ✖ PROFILER, ✖ DIST_KVSTORE, ✖ CXX14, ✔ INT64_TENSOR_SIZE, ✖ SIGNAL_HANDLER, ✔ DEBUG, ✖ TVM_OP #### BLAS = OpenBLAS ✖ CUDA, ✖ CUDNN, ✖ NCCL, ✖ CUDA_RTC, ✖ TENSORRT, ✔ CPU_SSE, ✔ CPU_SSE2, ✔ CPU_SSE3, ✔ CPU_SSE4_1, ✔ CPU_SSE4_2, ✖ CPU_SSE4A, ✔ CPU_AVX, ✖ CPU_AVX2, ✔ OPENMP, ✖ SSE, ✔ F16C, ✖ JEMALLOC, ✔ BLAS_OPEN, ✖ BLAS_ATLAS, ✖ BLAS_MKL, ✖ BLAS_APPLE, ✔ LAPACK, ✖ MKLDNN, ✖ OPENCV, ✖ CAFFE, ✖ PROFILER, ✖ DIST_KVSTORE, ✖ CXX14, ✔ INT64_TENSOR_SIZE, ✖ SIGNAL_HANDLER, ✔ DEBUG, ✖ TVM_OP ## Steps to reproduce ### Script Create a Python script with the following content: ``` from mxnet import nd print(nd.linalg_det(A=nd.random_normal(shape=(2**16, 2**16)))) ``` and run it with Python3. ### Error With both BLAS engines, the error is the same: ``` Segmentation fault (core dumped) ``` ## Additional Information The `linalg` ops do not throw errors on data with dimension <= 2**32. See the following example script and output: ### Script ``` from mxnet import nd print(nd.linalg_det(A=nd.random_normal(shape=(2**15, 2**15)))) ``` ### Output ``` [inf] <NDArray 1 @cpu(0)> ```
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
