How compilate CRYSTAL23 on old intell processors?

QMQDCHEM

Hi!

I am trying to compile Crystal 23 on Xeon E-2695v2 processors, but they are not supported by the Intel Toolkit. I am missing the instruction vfmadd231sd, giving GDB output:

Program received signal SIGILL, Illegal instruction.
0x0000000000e915a2 in timer_module_mp_init_timer_ ()
(gdb) bt full
#0 0x0000000000e915a2 in timer_module_mp_init_timer_ ()
No symbol table info available.
#1 0x0000000006828517 in initpp_ ()
No symbol table info available.
#2 0x00000000004073bd in MAIN__ ()
No symbol table info available.
(gdb) info registers
Missing register name
(gdb) x/i $rip
=> 0xe915a2 <timer_module_mp_init_timer_+82>:
vfmadd231sd 0x73c02ed(%rip),%xmm0,%xmm3 # 0x8251898

Could you please suggest a compilation method, such as using GCC or any other compiler? Is there a way to compile it using CMake?

Thank you!

GiacomoAmbrogio

Hi,
The intel oneAPI toolkit should in principle be compatible with any Xeon processor, have you compiled the code from the object files or have you used directly the executables distributed?

In the first case, could you provide the .inc file that you used for the compilation?

QMQDCHEM

I probably won’t be able to share the .inc file at the moment, as I’ve modified it multiple times. The situation is that the compilation completes successfully and crystal23 starts, but it crashes because my processor’s AVX version is too old for the Intel toolkit.

I mentioned the processor model for a reason: https://www.intel.com/content/www/us/en/products/sku/75281/intel-xeon-processor-e52695-v2-30m-cache-2-40-ghz/specifications.html

Support for this CPU was effectively discontinued even before the first versions of the Intel toolkit were released.

I wasn’t able to disable the use of this instruction during compilation. However, if you know how to do it, I can locate one of the working .inc files I still have.

Given this, it seems that using a different compiler would be necessary.

QMQDCHEM

First — my apologies. My previous post was written with an automatic translator and was confusing. Sorry for that. Below is a corrected, concise description of the problem and the exact output I have. I would appreciate any advice.

Short summary

Machine: dual-socket server node with two Intel Xeon E5-2697 v2 (Ivy Bridge). These CPUs support AVX but do not support AVX2.
Compiler/linker: Intel Fortran (I use mpiifx as the Fortran driver and icc/icpc for C/C++).
Build: I compiled using the included .inc (snippet below) with -march=ivybridge. Compilation completed successfully.
4)Run: When running the produced binary with the CRYSTAL input (included below) the program crashes immediately with:

forrtl: severe (168): Program Exception - illegal instruction
Image PC Routine Line Source
libc.so.6 0000718B3FC45330 Unknown Unknown Unknown
PcrystalOMP 0000000000E915A2 Unknown Unknown Unknown
PcrystalOMP 0000000006828517 Unknown Unknown Unknown
PcrystalOMP 00000000004073BD Unknown Unknown Unknown
PcrystalOMP 000000000040735D Unknown Unknown Unknown
libc.so.6 0000718B3FC2A1CA Unknown Unknown Unknown
libc.so.6 0000718B3FC2A28B __libc_start_main Unknown Unknown
PcrystalOMP 0000000000407275 Unknown Unknown Unknown

My .inc file:

For Linux based on x86_64 arch, using Intel Fortran Compiler

F90 = mpiifx
LD = $(F90)
PLD = mpiifx

F90COMMON = -diag-disable 7373 -diag-error 5198,6182,6893,6916,6919,7374,7416,7423,8089,8586
-align -static-intel -cxxlib -qopenmp
F90FLAGS = $(F90COMMON) -O3 -march=ivybridge
F90BASIS = $(F90COMMON) -O0
F90GORB = $(F90COMMON) -O2
F90DENS = $(F90COMMON) -O2
F90FIXED = -FI
F90FREE = -FR
SAVEMOD = -module $(MODDIR)
INCMOD = -I$(MODDIR)
LDFLAGS = $(F90FLAGS)
EIGENV = $(OBJDIR)/diag_lapack.o
MATMULT = $(OBJDIR)/mult_blas.o

MKLPATH = $(MKLROOT)/lib/intel64

LDLIBS = $(LIBXCFUN) -Wl,--start-group
$(MKLPATH)/libmkl_intel_lp64.a $(MKLPATH)/libmkl_intel_thread.a
$(MKLPATH)/libmkl_core.a $(MKLPATH)/libmkl_blacs_openmpi_lp64.a -Wl,--end-group -liomp5 -lpthread -lm -ldl
#LDLIBS =

MXMB = $(OBJDIR)/libmxm.o

MACHINE_C=mach_linux

CC = icc
CFLAGS = -O2 -vec-report0 -Wall -diag-disable 177,279,383,869,981,1418,1419,1572 -DNDEBUG
CXX = icpc
CXXFLAGS = $(CFLAGS) -fno-rtti -fno-exceptions

MPI harness

HARNESS = $(MPI)

https://www.intel.com/content/www/us/en/developer/tools/oneapi/onemkl-link-line-advisor.html

MKL=$(MKLROOT)/lib/intel64
MPPLIB=-L$(MKL) $(MKL)/libmkl_scalapack_lp64.a -Wl,--start-group
$(MKL)/libmkl_intel_lp64.a $(MKL)/libmkl_sequential.a
$(MKL)/libmkl_core.a $(MKL)/libmkl_blacs_openmpi_lp64.a -Wl,--end-group -liomp5 -lpthread -lm -ldl

Thank you in advance for any guidance. I’m happy to attach any of the files or run recommended diagnostics — please tell me which outputs would be most helpful.

Again, sorry for the confusion from the earlier (poorly translated) message.

GiacomoAmbrogio

The AVX (or AVX2/AVX512) option can indeed cause compatibility issues with old processors. Since the object files you are using for compilation were built with AVX2 enabled, there is no way to remove that option and make those files work on processors without AVX support.

However, for this exact reason, we also provide object files compiled without AVX enabled. You should download the appropriate version of the distribution from the CRYSTAL Solutions website (should be named CRYSTAL23 - Precompiled object files for parallel version no avx2) and compile it using the same .inc file you provided.

Let me know if you manage to do it.