how to report compiler bugs

[Martin1]

I just tried oneAPI, came across two compiler bugs and thought I might report those. But where?

"Request support" in the support center is the wrong place to go, as I do not seek any support. Nevertheless I tried, but as I do not have any support contract, the two tickets I opened were immediately closed.

Is intel interested in bug reports at all? And if yes, how should bugs be reported?

[Martin1]

After searching recent posts, I found that the forum is now used as bug tracking tool.

[Ron_Green · Accepted solution]

Customers with active support contracts use the Online Service Center to enter bug reports.

Without active support you can post bug reports here on this Forum.  We'd like to see them and can grab it from here and put it into our bug tracking system.

[accowa]

Bug report: 

Array  shape checking fails to detect runtime errors in all but the simplest assignment cases. Any operations on the RHS seem to disable checking even though the operations are returning mis-matched arrays. All these demonstator cases are detected correctly by gfortran with the -bounds-check option. Failure to detect these errors slows down development and discourages use of the intel compiler.

I've tried reporting this through an active support contract but having to go through our IT department may lose something in translation. Here are the details and a working demonstrator:

ifort -V

Intel(R) Fortran Intel(R) 64 Compiler Classic for applications running on Intel(R) 64, Version 2021.4.0 Build 20210910_000000

Copyright (C) 1985-2021 Intel Corporation. All rights reserved.

(and earlier releases)

Demonstration:

ifort -o trya2d_c -check shape trya2d_c.f90

./trya2d_c

  SHAPE(htau) 100 100

  L/UBOUND(htau) 2 2 101 101

  SHAPE(gphit) 102 102

  L/UBOUND(gphit) 1 1 102 102

This line is not detected as a shape mismatch even with -check shape at compilation:

htau(:,:) = MAX( 0.5, MIN( 30., 45.* ABS( SIN( 3.142/180. * gphit(:,:) ) ) ))

Even though the shapes do not match:

  100 100 102 102

  Note: this error is detected correctly by gfortran with -bounds-check

In fact, a simple operation with the RHS prevents shape checking:

 A: htau(:,:) = 1.0 * gphit(:,:) 102 102

 B: htau(:,) = gphit(:,:) * 1.0 102 102

Only a simple assignment is trapped correctly:

 C: htau(:,:) = gphit(:,:) 102 102

forrtl: severe (408): fort: (33): Shape mismatch: The extent of dimension 2 of array HTAU is 100 and the corresponding extent of array GPHIT is 102

Image PC Routine Line Source

trya2d_c 0000000000406F8F Unknown Unknown Unknown

trya2d_c 0000000000404370 Unknown Unknown Unknown

trya2d_c 0000000000403852 Unknown Unknown Unknown

libc-2.17.so 00002ADBAFD7E555 __libc_start_main Unknown Unknown

trya2d_c 0000000000403769 Unknown Unknown Unknown

Demonstrator program:

cat trya2d_c.f90

REAL, ALLOCATABLE, DIMENSION(:,:) :: htau, gphit

ALLOCATE(htau(2:101,2:101), gphit(102,102))

WRITE(*,*) ' SHAPE(htau)     ', SHAPE(htau)

WRITE(*,*) ' L/UBOUND(htau)  ', LBOUND(htau), UBOUND(htau)

WRITE(*,*) ' SHAPE(gphit)    ', SHAPE(gphit)

WRITE(*,*) ' L/UBOUND(gphit) ', LBOUND(gphit), UBOUND(gphit)

!

write(*,'(a)') 'This line is not detected as a shape mismatch even with -check shape at compilation: '

write(*,'(a)') 'htau(:,:) =  MAX(  0.5, MIN( 30., 45.* ABS( SIN( 3.142/180. * gphit(:,:) ) ) ))'

htau(:,:) =  MAX(  0.5, MIN( 30., 45.* ABS( SIN( 3.142/180. * gphit(:,:) ) ) ))

write(*,'(a)') 'Even though the shapes do not match: '

write(*,*) SHAPE(htau), SHAPE( MAX(  0.5, MIN( 30., 45.* ABS( SIN( 3.142/180. * gphit(:,:) ) ) )))

write(*,'(a)') '  Note: this error is detected correctly by gfortran with -bounds-check '

!

write(*,'(a)') 'In fact, a simple operation with the RHS prevents shape checking: '

write(*,*) 'A: htau(:,:) = 1.0 * gphit(:,:) ', SHAPE(1.0 * gphit(:,:))

htau(:,:) =   1.0 * gphit(:,:)

!

write(*,*) 'B: htau(:,) = gphit(:,:)  * 1.0 ',SHAPE(gphit(:,:)  * 1.0)

htau(:,:) =   gphit(:,:)  * 1.0

!

write(*,'(a)') 'Only a simple assignment is trapped correctly: '

write(*,*) 'C: htau(:,:) = gphit(:,:) ',SHAPE(gphit(:,:))

htau(:,:) =   gphit(:,:)

!

write(*,*) 'D'

end

[JNorw]

I think this is more of an dpc++ library bug.

dpc++ allows you to create multiple queues on a device.

dpc++ allows you to use depends_on(event1,event2,...) to synchronize execution.

dpc++ accepts events returned from different queues in the depends_on statement, however it appears that depends_on fails to work correctly if the current queue context is not the same as  the queue returning the event.

I'll provide example code modified from the dpc++ book to use multiple queues.  With event.wait() it works as expected.  Attempting to use the commented out depends_on(event) statements fails to execute the kernels.

Sorry, I found there was a more appropriate place for this in the dpc++ forum.  Please ignore this.

[Barbara_P_Intel]

dpc++ has its own forum for asking questions and posting bugs.  Can you post this there?