If one was to write a C function that allocated the memory and performed the offload operation, and it was called from within a Fortran code (so, ostensibly compiled with icc then ifort), could it utilize Implicit Copy, which is only set up for C/C++?
After applying a new technology (a new processor, a hardware accelerator, a new instruction, etc) besides measuring the immediate performance delta one requires a method to verify that this technology has been applied correctly and efficiently. Intel® Transactional Synchronization Extensions (Intel® TSX - instructions for speculative execution of critical sections protected by locks) are not an exception here.
I am trying to compile Seismic Unix Pakage from CWP but while compiling it generates mic executable which is not executed on Xeon. Do anyone have have experience for Seismic Unix compilation for MIC. Any help will be appriciated.
Several announcements were made on or during the ISC'13 conference. Of note are a couple of announcements from Acceleware and Bright Computing:
I have compiled WRF v3.5 in (dm+sm) par mode with all necessary libraries for Intel MIC.
I am getting the error "APPLICATION TERMINATED WITH THE EXIT STRING: Segmentation fault (signal 11)" when executed with mpirun as "mpirun -host localhost-mic0 -n 1 ./wrf.exe"
but when run the executable (./wrf.exe) by logging into the card, the executable runs without any problem provided ulimit is set to unlimited.
So, I tried changing the environment variable KMP_STACKSIZE on the host, to different values and i still have the same seg fault issue.
I've got an issue when compiling Mozilla (64-Bit) where it cannot find a specific header file even though it is present and all the other header files are being found in that directory:
c:/Users/Alex/Downloads/mozilla-beta/media/webrtc/trunk/webrtc/system_wrappers/source/critical_section_win.h(15): catastrophic error: cannot open source file "webrtc/system_wrappers/interface/critical_section_wrapper.h"
I get a curious crash on a vpgatherdd call and was trying to debug it. But I'm stuck because I can't figure out how idbc works. I'm used to gdb, so I thought this wouldn't be too hard. But:
- disasm doesn't work
- print %zmm0 (and all other zmm registers) doesn't work: prints "void"
- there is no register view (like "layout regs" in gdb)
- there is no asm view (like "layout asm" in gdb)
Is this a known limitation? How do you debug your MIC programs?
simple question: what constraint does one use for mask registers in inline-asm? "k" does not work and "r" would be wrong.