- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
I tried to compile a series of f90 files via Makefile with ifort (IFORT) 19.0.3.199 20190206/ifort (IFORT) 2021.2.0 20210228 and got
(col. 63) warning #13383: No instance of function "cos_sin" satisfies math function attributes "max-error=4.0 domain-exclusion=15 valid-status-bits=false fusa=false zmm-low=true"
04010002_0
fortcom: Severe: **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.
ifort: error #10014: problem during multi-file optimization compilation (code 3)
with some kind of debug flags
-O3 -qopenmp -c -fast -traceback -i8 -I${MKLROOT}/include/intel64/ilp64 -I${MKLROOT}/include -pg -g -mkl=sequential -O0 -i8 -warn all -check all -traceback -fp-stack-check
I am working on a non-open source project and cannot provide much information here. Basically the line triggers the issue is something like, dcos.. /dsin(x), 0<x<pi, and it's not about x slightly larger than zero still numerical issue in double precision
If I use
-O3 -qopenmp -c -fast -traceback -i8 -I${MKLROOT}/include/intel64/ilp64 -I${MKLROOT}/include -pg -g -mkl=sequential
the code can compile without any error message. If you need any additional information or there is any other channel I can ask than forum, please let me know.
Thank you very much
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Without you providing at least the exact line of code, and the variable definitions, it is impossible to provide you with any guidance.
Saying the code is something like ... isn't helpful, we really need to see the actual code (fragment).
Thanks,
David
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Thanks. Is there any email or inquiry form which I can use? than post in the open community? Since the code is in a non open source project.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
If it is a specific source file that is maybe "large", try removing say half the routines , you often can get down do a single routine that still shows the problem and then you can pair that to a small bare bones source file that can reproduce the error. You don't need to share any real production code.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Did you try a newer version of the compiler? ifort 2021.5.0 is available as part of the oneAPI HPC Toolkit. You'll also want to install the oneAPI Base Toolkit; MKL is in that Toolkit. You'll find more information and download instructions here.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I am experiencing the same warning, namely
Warning #13383: No instance of function "__svml_roundf4" satisfies math function attributes "max-error=4.0 domain-exclusion=0 valid-status-bits=true fusa=false zmm-low=true"
I am using the compiler version 2021.5.0 and the warning started to be issued after switching the compiler option Fortran -> Floating Point -> Floating-Point Exception Handling to /fpp:0 (from "Produce NaN, signed infinities, and denormal results").
In my code, the warning is issued for two locations.
Location 1:
if ( nint(electrode(1,electrodeNumber)) .ne. 0 ) then
where
integer(4), parameter :: MAX_ELECTRODE_ITEMS = 50
integer(4), parameter :: MSB = 10000
real*8 :: electrode(MAX_ELECTRODE_ITEMS,MSB)
integer(4) :: electrodeNumber
Location 2:
typeNew = nint(sReal(i,j,k))
where
real*4, allocatable, dimension(:,:,:) :: sReal
integer(4) :: typeNew
I tried reducing my project to a very simple example, however, it compiles without any warnings, even though I kept all compiler and linker settings and only did the change to /fpe:0, just like in the original project.
The warning seems to be related to using "nint", but the same like in locations 1 and 2, see above, is used throughout the code.

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page