Invalid memory access of location on OS X?

181 views
Skip to first unread message

Sean Owen

unread,
Apr 18, 2013, 6:21:32 PM4/18/13
to jblas...@googlegroups.com
Hello all, I've just gotten jblas working nicely on my OS X 10.7 machine -- really my application is just computing a matrix inverse with Solve.pinv(). It works at small scale.

But when I run at larger scale, I get a consistent core dump. On java 6 (update 45) the message is simply like "Invalid memory access of location 0x7f829a8228d0 rip=0x7fff93c2ae26"

On Java 7 (update 21) I get a nice big SIGSEGV. I can provide the whole hs_err_pid log file; the key stack trace seems to be

Stack: [0x000000010bda4000,0x000000010bea4000],  sp=0x000000010bea32b0,  free space=1020k
Native frames: (J=compiled Java code, j=interpreted, Vv=VM code, C=native code)
C  [libsystem_c.dylib+0x27e03]  memmove$VARIANT$sse42+0x146
C  [jblas2824452995535255166libjblas.dylib+0x148129]  ilaenv_+0xe3
C  [jblas2824452995535255166libjblas.dylib+0x1af96]  Java_org_jblas_NativeBlas_ilaenv+0x106
...

I'm not sure if this points to something I'm doing wrong -- but it is working in some cases. This is with the latest jblas 1.2.3.

mikiobraun

unread,
Apr 19, 2013, 5:37:21 AM4/19/13
to jblas...@googlegroups.com
Hi Sean,

sounds odd. Can you attach some example code? Then I'll have a look.

Best,

-Mikio

mikiobraun

unread,
May 14, 2013, 8:46:50 AM5/14/13
to jblas...@googlegroups.com
Hi Sean,

I did a quick fix. I think something was wrong with the sizes of the working arrays.

I tried it run with random matrices for two hours now, everything seems to be ok.

If you have the time, give it a try and tell me whether it works or not, that'll be great!

-M
Reply all
Reply to author
Forward
0 new messages