topic Re:Can I use MKL routines when running on GPU? in Intel® oneAPI Math Kernel Library

Can I use MKL routines when running on GPU?

eliopoulos — Tue, 06 Feb 2024 23:52:22 GMT

I want to enable my Intel iris xe graphics GPU to run a Fortran program which includes MKL routines. Is this possible? I have created and attached a simple reproducer. When I build the solution, I get:

Rebuild started at 01:30...
1>------ Rebuild All started: Project: z2 (IFX), Configuration: Release x64 ------
Deleting intermediate files and output files for project 'z2', configuration 'Release|x64'.
Compiling with Intel® Fortran Compiler 2024.0.2 [Intel(R) 64]...
z2.for
ifx: warning #10148: option '/size-llp64' not supported
ifx: warning #10148: option '/size-llp64' not supported
Linking...
Intel(R) Fortran Compiler for applications running on Intel(R) 64, Version 2024.0.2 Build 20231213
Copyright (C) 1985-2023 Intel Corporation. All rights reserved.
13292354.o : warning LNK4078: multiple '__CLANG_OFFLOAD_BUNDLE__openmp-s' sections found with different attributes (40500040)
Embedding manifest...

Build log written to "file://C:/Work/Work/2023%20OPENZDM/Fortran%20folder/z2/z2/x64/Release/BuildLog.htm"
z2 - 0 error(s), 3 warning(s)
========== Rebuild All: 1 succeeded, 0 failed, 0 skipped ==========
========== Rebuild completed at 01:30 and took 01.366 seconds ==========

and when I run the program:

Libomptarget error: Unable to generate entries table for device id 0.
Libomptarget error: Failed to init globals on device 0
Libomptarget error: Run with
Libomptarget error: LIBOMPTARGET_DEBUG=1 to display basic debug information.
Libomptarget error: LIBOMPTARGET_DEBUG=2 to display calls to the compute runtime.
Libomptarget error: LIBOMPTARGET_INFO=4 to dump host-target pointer mappings.
Libomptarget error: Source location information not present. Compile with -g or -gline-tables-only.
Libomptarget fatal error 1: failure of target construct while offloading is mandatory.

Is there anything I can do to solve this issue?

Re: Can I use MKL routines when running on GPU?

eliopoulos — Fri, 09 Feb 2024 22:59:49 GMT

Can anyone tell me if it works on better Intel GPU?

Re:Can I use MKL routines when running on GPU?

Mahan — Tue, 13 Feb 2024 05:27:12 GMT

Could you please set these environment variables and run the application through command line

set OMP_TARGET_OFFLOAD=MANDATORY

set LIBOMPTARGET_PLUGIN=LEVEL_ZERO

set LIBOMPTARGET_DEVICETYPE=GPU

Re: Re:Can I use MKL routines when running on GPU?

eliopoulos — Tue, 13 Feb 2024 11:42:10 GMT

It does not work. It produces the same errors:

C:\Work\Work\2023 OPENZDM\Fortran folder\z2\z2\x64\Release>set OMP_TARGET_OFFLOAD=MANDATORY

C:\Work\Work\2023 OPENZDM\Fortran folder\z2\z2\x64\Release>set LIBOMPTARGET_PLUGIN=LEVEL_ZERO

C:\Work\Work\2023 OPENZDM\Fortran folder\z2\z2\x64\Release>set LIBOMPTARGET_DEVICETYPE=GPU

C:\Work\Work\2023 OPENZDM\Fortran folder\z2\z2\x64\Release>ifx /Qopenmp-targets:spir64 /Qopenmp /Qiopenmp /Qmkl z2.for
Intel(R) Fortran Compiler for applications running on Intel(R) 64, Version 2024.0.2 Build 20231213
Copyright (C) 1985-2023 Intel Corporation. All rights reserved.

ifx: warning #10148: option '/size-llp64' not supported
ifx: warning #10148: option '/size-llp64' not supported

C:\Work\Work\2023 OPENZDM\Fortran folder\z2\z2\x64\Release>z2
Libomptarget error: Unable to generate entries table for device id 0.
Libomptarget error: Failed to init globals on device 0
Libomptarget error: Run with
Libomptarget error: LIBOMPTARGET_DEBUG=1 to display basic debug information.
Libomptarget error: LIBOMPTARGET_DEBUG=2 to display calls to the compute runtime.
Libomptarget error: LIBOMPTARGET_INFO=4 to dump host-target pointer mappings.
Libomptarget error: Source location information not present. Compile with -g or -gline-tables-only.
Libomptarget fatal error 1: failure of target construct while offloading is mandatory

Re:Can I use MKL routines when running on GPU?

Mahan — Tue, 13 Feb 2024 16:34:10 GMT

If you use the following for compilation, does it help?

ifx /Qiopenmp /Qopenmp-targets:spir64 -fsycl mkl_sycl.lib mkl_intel_lp64.lib mkl_intel_thread.lib mkl_core.lib libiomp5md.lib OpenCL.lib

Re: Re:Can I use MKL routines when running on GPU?

eliopoulos — Tue, 13 Feb 2024 17:12:30 GMT

It works. Thank you.

Re:Can I use MKL routines when running on GPU?

Mahan — Tue, 13 Feb 2024 17:14:58 GMT

Thanks for letting me know.

Also remember to set those environment variables.