Crash on Matlab 2016b Ubuntu

```
------------------------------------------------------------------------
Segmentation violation detected at Mon Jul 10 16:19:57 2017
------------------------------------------------------------------------

Configuration:
Crash Decoding : Disabled - No sandbox or build area path
Crash Mode : continue (default)
Current Graphics Driver: Unknown software
Current Visual : None
Default Encoding : UTF-8
Deployed : false
GNU C Library : 2.19 stable
Host Name : xx
MATLAB Architecture : glnxa64
MATLAB Entitlement ID: Unknown
MATLAB Root : /usr/local/MATLAB/R2016b
MATLAB Version : 9.1.0.441655 (R2016b)
OpenGL : software
Operating System : Linux 3.13.0-113-generic #160-Ubuntu SMP Thu Mar 9 09:27:29 UTC 2017 x86_64
Processor ID : x86 Family 6 Model 79 Stepping 1, GenuineIntel
Virtual Machine : Java 1.7.0_60-b19 with Oracle Corporation Java HotSpot(TM) 64-Bit Server VM mixed mode
Window System : No active display

Fault Count: 10

Abnormal termination:
Segmentation violation

Register State (from fault):
RAX = 00007fa60f2b36e0 RBX = 00007fcec961ea54
RCX = 00000000000000c0 RDX = 0000002800000d74
RSP = 00007fa72f926be8 RBP = 00007fa72f926dd0
RSI = 0000000000000000 RDI = 00007fa72f926e00

R8 = 0000000000002718 R9 = 0000000000001200
R10 = 00007fa72f926ff0 R11 = 00007fa72f926ff0
R12 = 00007fa72f926e00 R13 = 0000000000002718
R14 = 00000000000000bf R15 = 00007fa72f927388

RIP = 00007fa643b48402 EFL = 0000000000010212

CS = 0033 FS = 0000 GS = 0000

Stack Trace (from fault):
[ 0] 0x00007fa643b48402 /usr/local/MATLAB/R2016b/bin/glnxa64/mkl.so+45290498 mkl_blas_avx2_sgemm_scopy_down24_no+00000258
[ 1] 0x00007fa6424aee51 /usr/local/MATLAB/R2016b/bin/glnxa64/mkl.so+21593681 mkl_blas_avx2_sgemm_copyan+00000097
[ 2] 0x00007fa72f926e00 +00000000
[ 3] 0x000000030000035d +00000000
```

No source file, test, or reproduction command is provided. Start by reproducing the crash on MATLAB R2016b on Ubuntu/Linux, then investigate the reported mkl.so stack trace and system configuration to isolate the trigger. Done means matrix multiplication no longer causes a segmentation violation on the affected setup, with a regression check if a reproducible case can be established.

Written by the indexing model from the issue text.