error : backend signals

[Daniel_D]

Im evaluating the Intel C++ Studio XE for Windows on a Windows7 64-Bit machine. My solution contains 10 projects. Each time I compile the solution I get these messages:

10>C:\\Program Files (x86)\\MSBuild\\Microsoft.Cpp\\v4.0\\Platforms\\x64\\PlatformToolsets\\Intel C++ Compiler XE 12.1\\Microsoft.Cpp.x64.Intel C++ Compiler XE 12.1.targets(211,5): error : backend signals

10>icl : error #10298: problem during post processing of parallel object compilation

Sometimes it works and no error appears, but that is an exception. In general I have 1 to 4 of these messages if I build the whole solution. I already tried to chance the value of the Build and Run/maximum number of parallel. -settings to 1 but without any success. The compiler ALWAYS builds all files, even if only a single source is changed. I read the solution to uninstall some of the SECURITY-updates form MS that does solve this issue, but I read that it was address in one of the latest updates. How can I get these updates if I use any eval-copy? With these issues I cannot work with it ;(

Daniel

[Judith_W_Intel]

I found one openproblem report that looks like it might be related (cq #174339).

The problem report has to do with precompiled headers under these two circumstances as described
in the resolution note below. Does this look like it might be the same situation as yours?

Resolution note:

Two changes to the PCH memory management code.

First we get an internal error when trying to use a bad PCH. For instance if you tried to use a release build PCH in a debug compile you would get an internal error. This is represented with cfefailCpp/cq174339. To fix this I set a global (pch_restore_failed) when the pch cannot be used and then check it when allocating memory blocks. When the restore failed we will just use malloc blocks and continue.

Second we get an internal error with the Adobe test case which is a PCH that doesnt include anything. This is represented with cfefailCpp/cq174339a. The problem here is we allocate more blocks before reloading the PCH in the second use compile than in the first use compile when doing a multisrc compile. I didnt think that would happen and just reused the blocks from the first compile but when we needed another one it failed. To fix this I now check how many blocks I have to reuse and if I need more I use a malloc block. I have to keep this in a list and reuse it on subsequent compiles or it would be leaked. I think there will only be one of these malloc blocks but I implemented it as a list just in case since it is better safe than sorry with these pch problems.

Ifyou are experiencing the second problem then using the /Q_multisrc- switch todisable multisrc compilation (i.e. only one source file will be compiled at a time)should bea workaround.

Judy

[Om_S_Intel]

The evaluation allows you to download and use Intel compiler for one month. If you have registered your license in Intel registration center then yoy candown load the latetst compiler. You can login to registration center using link https://registrationcenter.intel.com/RegCenter/Register.aspx.

Om

[Daniel_D]

Hi Judy,

Thanks for your reply. I have deleted all files generated from the VC++ compiler, so the PCH header can only be on from the Intel-compiler. As I mentions in the OP I already tried to switch off multi-source compile by setting the value of Tools->Build & Run->maximum number of parallel. to 1. That doesnt make any difference. Im not sure what you suggest what I should do other than that??? I also get some more messages - now I get these in almost 2 of 3 runs, all of them can appear more than once ;(

This:

2>C:\Program Files (x86)\MSBuild\Microsoft.Cpp\v4.0\Platforms\x64\PlatformToolsets\Intel C++ Compiler XE 12.1\Microsoft.Cpp.x64.Intel C++ Compiler XE 12.1.targets(211,5): error : backend signals

2>icl : error #10298: problem during post processing of parallel object compilation

And this:

2>FileTracker : error FTK1011: could not create the new file tracking log file: D:\.. \x64\Debug-Win7\icl-mcpcom.write.1.tlog. The file exists.

Brand new this one:

1>C:\Program Files (x86)\Microsoft SDKs\Windows\v7.0A\include\imagehlp.h(1734): error : assertion failed: end_mangling_full: wrong number of leftover spaces (shared/cfe/edgcpfe/lower_name.c, line 992)

1>

1> typedef struct _IMAGEHLP_CBA_READ_MEMORY {

1> ^

1>

1>icl : error #10298: problem during post processing of parallel object compilation

And this one:

1> xilink: executing 'link'

1>libcpmt.lib(newaop.obj) : error LNK2038: mismatch detected for '_ITERATOR_DEBUG_LEVEL': value '0' doesn't match value '2' in MemoryManager.obj

1>x64\Debug-Win7\TeraDirect.exe : fatal error LNK1319: 1 mismatches detected

This makes it really unusable to work with I need about 10 tries until one compilation passes without any error ;(

Daniel

[Daniel_D]

Comes the latest release with the eval-package or is there something to update?

[Om_S_Intel]

The evaluation package may not be the latest release package. The "Intel Registration and download Center"contains latest release and older versions.

[Judith_W_Intel]

If you turn off precompiled headers do all the problems disappear?

Judy

[Daniel_D]

Turning precompiled headers off looks much better. Now I get new problems. First of all I need to link with a 3^rd party library which is compiled with the VC++. Are there any known issues if I try to link MS-libraries to an Intel compiled project???? I have my own libraries compiled as DEBUG and as RELEASE version. The sizes are very strange

Library1:
DEBUG: 5755 KB
RELEASE: 128573 KB

Library2:
DEBUG: 4941 KB
RELEASE: 27043 KB

Is there any typical mistake I have done?

Compiling one of my DLLs as release gives me this (DEBUG works OK):

1>ipo : warning #11072: ignoring invalid linker directive 'FAILIFMISMATCH:"_MSC_VER=1600"'

1>ipo : warning #11072: ignoring invalid linker directive 'FAILIFMISMATCH:"_ITERATOR_DEBUG_LEVEL=0"'

1>ipo : error #11018: Cannot open ipps

1>ipo : error #11018: Cannot open ippvm

1>ipo : error #11018: Cannot open ippcore

1>ipo : warning #11021: unresolved wsprintfW

1> Referenced in ipo_13685obj.obj

1>ipo : error #11023: Not all components required for linking are present on command line

1> xilink: executing 'link'

1>LINK : fatal error LNK1104: cannot open file 'ipps.lib'

Daniel

[JenniferJ · Accepted solution]

under the project property, "Configuration properties>Intel Performance Libaries", on the right, select "Use IPP", and set "Yes".

Build again.

Jennifer

[Daniel_D]

Thanks that works so far. The last issue now is using any 3rd party libraries that are compiled with vc++. If I link against the lib I get this (ONLY RELEASE-settings!!!!):

1>ipo : warning #11072: ignoring invalid linker directive 'FAILIFMISMATCH:"_MSC_VER=1600"'

1>ipo : warning #11072: ignoring invalid linker directive 'FAILIFMISMATCH:"_ITERATOR_DEBUG_LEVEL=0"'

3>ipo : warning #11072: ignoring invalid linker directive 'FAILIFMISMATCH:"_MSC_VER=1600"'

3>ipo : warning #11072: ignoring invalid linker directive 'FAILIFMISMATCH:"_ITERATOR_DEBUG_LEVEL=0"'

5>ipo : warning #11072: ignoring invalid linker directive 'FAILIFMISMATCH:"_MSC_VER=1600"'

5>ipo : warning #11072: ignoring invalid linker directive 'FAILIFMISMATCH:"_ITERATOR_DEBUG_LEVEL=0"'

Is it safe to ignore these messages? I get no message if I compile using the debug-settings.

Daniel

[TimP]

I've been ignoring those messages myself.

[Daniel_D]

Thanks. It is a lot of work to convert large existing solution from MS to Intel ;(

I read the compiler now supports IEEE754 BFP. I saw the Overview and a Getting Started, but no 'Documentation' for it. Is there any source for more detailed information and C++ samples?

I agree that the sample below from the help file is a nice start, but the is nothing I can use in real life.

double a64, b64;
a64 = 1.000000059604644775390625;
b64 = 1.1102230246251565404236316680908203125e-16;

I like to use it for a financial app, but I need more documentation how to use it.

Daniel

[Judith_W_Intel]

These warnings should be fixed in Update 7 (cq #174985). Resolution note says:

Front-end recently started recognizing and passing the detect_mismatch pragma. IPO object reader did not know how to handle the corresponding backend directive which led to the warning.