error #5623: **Internal compiler error: internal abort**

[frs]

Hi,

Today I have started to use ifx & icx for a large accelerator project MAD-X. Up to recently we have been using the ifort & icc combi. The C modules work fine while for one single Fortran module scater.f90 the execution of ifx end with a "internal compiler error":

ifx -m64 -O3 -warn all -warn nounused -warn nointerfaces -c -mp1 -fp-model strict -heap-arrays 65536 -assume noold_unit_star -module build/Linux64 -fexceptions -D_
STATIC -D_ICC -D_GXX -D_IFX -D_LINUX -D_NTPSA scater.f90 -o scater.o
#0 0x0000000002394d07 (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x2394d07)
#1 0x0000000002394cd6 (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x2394cd6)
#2 0x0000000002313696 (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x2313696)
#3 0x0000000002361420 (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x2361420)
#4 0x00000000022af5bc (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x22af5bc)
#5 0x00000000022afdad (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x22afdad)
#6 0x0000000002365a33 (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x2365a33)
#7 0x0000000002364901 (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x2364901)
#8 0x00000000022703db (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x22703db)
#9 0x000000000226fcd0 (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x226fcd0)
#10 0x0000000002452dbe (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x2452dbe)
#11 0x00007f720b64614a __libc_start_call_main (/usr/lib64/libc.so.6+0x2814a)
#12 0x00007f720b64620b __libc_start_main@GLIBC_2.2.5 (/usr/lib64/libc.so.6+0x2820b)
#13 0x00000000020ab129 (/opt/intel/oneapi/compiler/2024.0/bin/compiler/xfortcom+0x20ab129)

scater.f90(1035): error #5623: **Internal compiler error: internal abort** Please report this error along with the circumstances in which it occurred in a Software Problem Report.  Note
: File and line given may not be explicit cause of this error.
     DOUBLE PRECISION FUNCTION RNDM5 (RDUMMY)                                    
--------------------------------^
compilation aborted for scater.f90 (code 3)
*******************************************************************************

I am adding the fortran code below.

Thanks for your attention!

[Barbara_P_Intel]

Thank you for reporting this ICE (internal compiler error). How unfortunate that you got this during your initial testing.

I duplicated the ICE with the latest internal nightly build of ifx. I filed a bug report, CMPLRLLVM-56733. We'll let you know when it's fixed.

Do you know that you can mix and match ifx and ifort .o and .mod files as long as you are not using -ipo or -flto?

Have you seen the ifort to ifx Porting Guide?

[frs]

Hi Barbara,

I have not seen any updates to this request nor has the software

being updated. Please let me know about any solution and Thanks

in advance!

Most likely I will need another ticket for the following prettyy serious

limitation. But let me explain"

- The same code with the scater.f90 issue, very complex indeed(!), I

find the following problem:

a) When compiling with the duo gcc/ifort and using ifort for linking

and going for full DEBUG and using gdb for finding run-time errors

all goes very smoothly.

b) Going to icx/ifx the nominal executables work perfectly fine. However,

in DEBUG mode and "ifx" in linking I am getting a tremendous amount of missing

packages. Consulting the Internet that issues should be due to missing

"check -all" link flag. But these are the identical link flags as in the functional a)

and certainly this check flag has been in place. For the time being Debugging

is impossible for "ifx".

Maybe you could launch this second issue for the same case. I would be happy

to provide any additional information!

Cheers, Frank Schmidt

[Barbara_P_Intel]

It looks like the first issue you reported will be fixed in the release planned for mid-2024.

Please open a new thread for your new issue. I think it will be easier for both of us to track that way. Thank you!