#863 Crash in matrix multiplication

Stable_(v3.10.x)
closed-fixed
5
2014-07-09
2012-10-17
No

On Itanium2 (iras on skynet), atlas-10.0 crashes in a Sage doctest:

sage: for i in range(150):
m = random_matrix(CDF, i)
m*m

Output follows:

...
101 x 101 dense matrix over Complex Double Field
102 x 102 dense matrix over Complex Double Field
103 x 103 dense matrix over Complex Double Field
104 x 104 dense matrix over Complex Double Field
105 x 105 dense matrix over Complex Double Field
106 x 106 dense matrix over Complex Double Field

Program received signal SIGSEGV, Segmentation fault.
0x2000000003ca2d00 in ATL_zupNBmm0_1_0_b0 ()
from /home/vbraun/opt/iras/sage-5.3.beta0/local/lib/libatlas.so.2
(gdb) bt
#0 0x2000000003ca2d00 in ATL_zupNBmm0_1_0_b0 ()
from /home/vbraun/opt/iras/sage-5.3.beta0/local/lib/libatlas.so.2
#1 0x2000000003cc9c60 in ATL_zpNBmm_b0 ()
from /home/vbraun/opt/iras/sage-5.3.beta0/local/lib/libatlas.so.2
#2 0x2000000003b64150 in ATL_zMBJBmm ()
from /home/vbraun/opt/iras/sage-5.3.beta0/local/lib/libatlas.so.2
#3 0x2000000003b65eb0 in ATL_zmmJIK2 ()
from /home/vbraun/opt/iras/sage-5.3.beta0/local/lib/libatlas.so.2
#4 0x2000000003b66ab0 in ATL_zmmJIK ()
from /home/vbraun/opt/iras/sage-5.3.beta0/local/lib/libatlas.so.2
#5 0x2000000003b1d990 in ATL_zgemm ()
from /home/vbraun/opt/iras/sage-5.3.beta0/local/lib/libatlas.so.2
#6 0x2000000003c659d0 in ATL_ztgemm ()
from /home/vbraun/opt/iras/sage-5.3.beta0/local/lib/libatlas.so.2
#7 0x2000000010df2a60 in cblas_zgemm ()
from /home/vbraun/opt/iras/sage-5.3.beta0/local/lib/libptcblas.so.2
#8 0x2000000010d668b0 in dotblas_matrixproduct (__NPY_UNUSED_TAGGEDdummy=0x0, args=0x20000000121e35f0)
at numpy/core/blasdot/_dotblas.c:780
#9 0x200000000017ec50 in PyCFunction_Call (func=<value optimized out>, arg=0x20000000121e35f0, kw=0x0)
at Objects/methodobject.c:90
#10 0x20000000000c83d0 in PyObject_Call (func=0x20000000109106c8, arg=0x20000000121e35f0, kw=0x0)
at Objects/abstract.c:2529
#11 0x20000000109a6fa0 in __pyx_f_4sage_6matrix_19matrix_double_dense_19Matrix_double_dense__matrix_times_matrix_ (__pyx_v_self=0x20000000121ddeb0, __pyx_v_right=0x20000000121ddeb0)
at sage/matrix/matrix_double_dense.c:4835
#12 0x2000000002c3e870 in __pyx_pw_4sage_9structure_7element_6Matrix_5__mul__ (
__pyx_v_left=0x20000000121ddeb0, __pyx_v_right=0x20000000121ddeb0)
at sage/structure/element.c:19152

Something is wrong in ATL_zupNBmm0_1_0_b0 at matrix size 107 and higher.

Discussion

  • R. Clint Whaley

    R. Clint Whaley - 2012-10-30
    • assigned_to: nobody --> rwhaley
     
  • R. Clint Whaley

    R. Clint Whaley - 2013-03-28

    I've been told this bug is fixed. Can someone confirm that true/false and I will close out if true?

     
  • R. Clint Whaley

    R. Clint Whaley - 2014-07-09
    • status: open --> closed-fixed
     
  • R. Clint Whaley

    R. Clint Whaley - 2014-07-09

    No response, closing as fixed.

     

Log in to post a comment.