Unrecognized command line option '-mno-hw-div' when compiling AOCL kernel

[Altera_Forum]

Hello, 

 I am getting this error on every OpenCL kernel I am trying to compile, even those I have successfully compiled before on my machine. I have tried it on both AOCL 14.0 and 15.1.   2016.08.10.13:23:24 Error: s0: Error during execution of "/opt/altera/15.1/quartus/../nios2eds/nios2_command_shell.sh make all 2>> stderr.txt": error: unrecognized command line option '-mno-hw-div' 2016.08.10.13:23:24 Error: s0: Error during execution of "/opt/altera/15.1/quartus/../nios2eds/nios2_command_shell.sh make all 2>> stderr.txt": error: unrecognized command line option '-mno-hw-mul' 2016.08.10.13:23:24 Error: s0: Error during execution of "/opt/altera/15.1/quartus/../nios2eds/nios2_command_shell.sh make all 2>> stderr.txt": error: unrecognized command line option '-mno-hw-mulx'  I am coming back to AOCL after a few weeks, but I don't recall any change I made to the framework that led to this error. Can someone advise what might be the cause?  Cheers,  Waqar.

[Altera_Forum]

Hello, 

 have you ever solved this? I'm having the same problem trying to compile one specific kernel. When this happens, everything that I try to compile afterwards fails with the same error, even kernels that once compiled. I must reboot the computer to be able to recompile kernels that were once succesfully compiled.    --- Quote Start ---  Hello,  I am getting this error on every OpenCL kernel I am trying to compile, even those I have successfully compiled before on my machine. I have tried it on both AOCL 14.0 and 15.1.   2016.08.10.13:23:24 Error: s0: Error during execution of "/opt/altera/15.1/quartus/../nios2eds/nios2_command_shell.sh make all 2>> stderr.txt": error: unrecognized command line option '-mno-hw-div' 2016.08.10.13:23:24 Error: s0: Error during execution of "/opt/altera/15.1/quartus/../nios2eds/nios2_command_shell.sh make all 2>> stderr.txt": error: unrecognized command line option '-mno-hw-mul' 2016.08.10.13:23:24 Error: s0: Error during execution of "/opt/altera/15.1/quartus/../nios2eds/nios2_command_shell.sh make all 2>> stderr.txt": error: unrecognized command line option '-mno-hw-mulx'  I am coming back to AOCL after a few weeks, but I don't recall any change I made to the framework that led to this error. Can someone advise what might be the cause?  Cheers,  Waqar.  --- Quote End ---  

[Altera_Forum]

I remember encountering this once long ago; somehow the environmental variables got corrupted/broken and broke the compilation. I have not encountered this since v16.x; though. Are you using an old version of Quartus (v14.x or 15.x)?

[Altera_Forum]

Hello, 

 I remember that I had this problem once on old versions (14.x or 15.x), but now it happened again on version 16.1. I'm installing version 17.0 to see if it helps.  Thanks for the attention!    --- Quote Start ---  I remember encountering this once long ago; somehow the environmental variables got corrupted/broken and broke the compilation. I have not encountered this since v16.x; though. Are you using an old version of Quartus (v14.x or 15.x)?  --- Quote End ---  

[Altera_Forum]

I checked my old emails and the discussion I had with my colleague regarding this issue. I think the issue is the same as the one caused by X11 forwarding, just the error is different if you are targeting SoCs (you are targeting an SoC, right?). Try the suggestions from this topic (disable X11 forwarding and unset display): 

 https://www.alteraforum.com/forum/showthread.php?t=55328  Also avoid having multiple make files that call each other or export variables.

[Altera_Forum]

Hello, 

 I am compiling through ssh and screen as well, though X11 forwarding is off ($DISPLAY is empty). I'm trying to compile now using 15.1 and apparently it worked. I couldn't use version 17.0 since I don't have yet a licence for it. I'm using only 1 makefile and I'm targeting a Stratix V opencl board provided by bittware (s5-phq).  Thanks for the help anyway!    --- Quote Start ---  I checked my old emails and the discussion I had with my colleague regarding this issue. I think the issue is the same as the one caused by X11 forwarding, just the error is different if you are targeting SoCs (you are targeting an SoC, right?). Try the suggestions from this topic (disable X11 forwarding and unset display):  https://www.alteraforum.com/forum/showthread.php?t=55328  Also avoid having multiple make files that call each other or export variables.  --- Quote End ---  

[Altera_Forum]

Nope, still not working. Sometimes it compiles randomly, but then if I try again, it fails. I tried without using screen, directly on the PC, the problem persists. 

 I'm trying now on another computer that has Ubuntu installed (the first PC has CentOS), will inform if it works.    --- Quote Start ---  Hello,  I am compiling through ssh and screen as well, though X11 forwarding is off ($DISPLAY is empty). I'm trying to compile now using 15.1 and apparently it worked. I couldn't use version 17.0 since I don't have yet a licence for it. I'm using only 1 makefile and I'm targeting a Stratix V opencl board provided by bittware (s5-phq).  Thanks for the help anyway!  --- Quote End ---  

[Altera_Forum]

Compilations are running succesfully in the Ubuntu machine (version 16.1). Apart from the OS, the difference between both machines is the memory installed (the CentOS machine has 16GB, whilst the Ubuntu 32GB). 

   --- Quote Start ---  Nope, still not working. Sometimes it compiles randomly, but then if I try again, it fails. I tried without using screen, directly on the PC, the problem persists.  I'm trying now on another computer that has Ubuntu installed (the first PC has CentOS), will inform if it works.  --- Quote End ---  

[Altera_Forum]

I can assure you this has nothing to do with the OS since I have been using CentOS v6.x since the beginning and I encountered this issue only once when I was using Quartus v15 on a cyclone board. The compiler also gives clear messages in the log when it crashes due to running out of memory; this case back then seemed like some issue with environmental variables to me.

[Altera_Forum]

Agreed. This is something that Intel/Altera should really look into, since it is such an obscure error that gives us absolutely no hint where to start looking at. Perhaps they've fixed on 17.0, but I have no access to this version. Problem solved (for now). Thank you everyone for the help! 

   --- Quote Start ---  I can assure you this has nothing to do with the OS since I have been using CentOS v6.x since the beginning and I encountered this issue only once when I was using Quartus v15 on a cyclone board. The compiler also gives clear messages in the log when it crashes due to running out of memory; this case back then seemed like some issue with environmental variables to me.  --- Quote End ---