https://community.intel.com/t5/Intel-Integrated-Performance/Segmentation-violation-with-Matlab-R2006b-IPP-5-1-1/m-p/903017#M13023 <P>Hi Koen,</P> <P>thank you a lot for providing us with that detailed information, it definetely should help in uderstanding the reasons. I also recommend you to submit your issue report to both <A href="http://premier.intel.com">Intel Premier Support</A> and Matlab support. At least for Intel side, it will give you opportunity to be informed about investigation results.</P> <P>Regards,<BR /> Vladimir</P> Wed, 01 Nov 2006 22:36:08 GMT Vladimir_Dudnik 2006-11-01T22:36:08Z https://community.intel.com/t5/Intel-Integrated-Performance/Segmentation-violation-with-Matlab-R2006b-IPP-5-1-1/m-p/903016#M13022 Hi,<BR /><BR />I've been encountering various problems with Matlab and IPP on Linux. Somehow, when calling IPP routines from Matlab, something gets corrupted. Sometimes Matlab just crashes, but even if it doesn't, it cannot exit gracefully. It doesn't happen with all IPP functions, but the filters/convolutions result in these problems (ippiFilter* and ippiConv*).<BR /><BR />When I stop Matlab and restart it, I get a screen which may point you experts into the right direction. The most important part is:<BR /><BR /><FONT size="2"> Operating System: Linux 2.6.18koen #1 SMP Mon Oct 23 12:14:46 CEST 2006 i686<BR /> GNU C Library: 2.3.6 stable<BR /> Processor ID: x86 Family 15 Model 4 Stepping 7, GenuineIntel<BR /> Virtual Machine: Java 1.5.0_08 with Sun Microsystems Inc. Java HotSpot Client VM mixed mode, sharing<BR /> Default Charset: UTF-8<BR /><BR />Register State:<BR /> eax = 00000000 ebx = 0000036c<BR /> ecx = 000003da edx = 0000000b<BR /> esi = b5a1f100 edi = b75d6ff4<BR /> ebp = a9c3d39c esp = a9c3d384<BR /> eip = b7f68410 flg = 00000206<BR /><BR />Stack Trace:<BR /> [0] 0xb7f68410(11, 0, 0, 0x454710ef)<BR /> [1] libguide.so:0xb5a0d12f(0xb5a1f100, 0xa9c3d490 ", 0xa9c3d490 ", 0xa9c3d490 ")<BR /> [2] libpthread.so.0:0xb75cd240(0xa9c3dbb0, 0, 0, 0)<BR /></FONT><BR />Just to see which numbers are reproducible, I did it again, resulting in:<BR /><BR /><FONT size="2">Register State:<BR /> eax = 00000000 ebx = 000003e0<BR /> ecx = 0000052f edx = 0000000b<BR /> esi = b5a7e100 edi = b7635ff4<BR /> ebp = aa13e39c esp = aa13e384<BR /> eip = b7fc7410 flg = 00000202<BR /><BR />Stack Trace:<BR /> [0] 0xb7fc7410(11, 0, 0, 0x45471318)<BR /> [1] libguide.so:0xb5a6c12f(0xb5a7e100, 0xaa13e490 ", 0xaa13e490 ", 0xaa13e490 ")<BR /> [2] libpthread.so.0:0xb762c240(0xaa13ebb0, 0, 0, 0)<BR /></FONT><BR />With small images, it sometimes seems to work OK. For now, I think I'll stop using these routines, because I've spend way too much time on this already. If someone has got some great idea how to solve this, please let me know!<BR /><BR />Thanks!<BR /><BR />Koen<BR /> Tue, 31 Oct 2006 17:15:11 GMT https://community.intel.com/t5/Intel-Integrated-Performance/Segmentation-violation-with-Matlab-R2006b-IPP-5-1-1/m-p/903016#M13022 kavermeer 2006-10-31T17:15:11Z https://community.intel.com/t5/Intel-Integrated-Performance/Segmentation-violation-with-Matlab-R2006b-IPP-5-1-1/m-p/903017#M13023 <P>Hi Koen,</P> <P>thank you a lot for providing us with that detailed information, it definetely should help in uderstanding the reasons. I also recommend you to submit your issue report to both <A href="http://premier.intel.com">Intel Premier Support</A> and Matlab support. At least for Intel side, it will give you opportunity to be informed about investigation results.</P> <P>Regards,<BR /> Vladimir</P> Wed, 01 Nov 2006 22:36:08 GMT https://community.intel.com/t5/Intel-Integrated-Performance/Segmentation-violation-with-Matlab-R2006b-IPP-5-1-1/m-p/903017#M13023 Vladimir_Dudnik 2006-11-01T22:36:08Z