I met a bug when I added a loop into vis_xc. codes:
! non-collinear case
! modified by Xiaoyu Zhang, Peking University, China
! 2024.04.04
IF (noco%l_noco) THEN
! set sample points for Lebedev grids
! options: 6(for test),146, 170, 194, 230, 266, 302, 350, 434, 590, 770, 974, 1202, 1454, 1730, 2030, 2354, 2702, 3074, 3470, 3890, 4334, 4802, 5294, 5810
a=1202
! prepare for the sum of lebedev sampling
ALLOCATE(v_xc_tmp,mold=rho)
ALLOCATE(v_xc2_tmp,mold=rho)
ALLOCATE(v_x_tmp,mold=rho)
IF (ALLOCATED(exc%pw_w)) THEN
ALLOCATE ( e_xc_tmp(SIZE(rho,1),1) ); e_xc_tmp=0.0
ENDIF
v_xc_tmp=0
v_xc2_tmp=0
v_x_tmp=0
allocate(x(a), y(a), z(a), w(a))
! store directions and weights
call ld_by_order(a, x, y, z, w)
! precautions for wroung sampling points setting or failed calling
IF (any(w >= 1.0) .or. any(w <= 0)) then
CALL juDFT_error("lebedev grids insertion fails",calledby="vis_xc")
END IF

l_libxc=.FALSE.

perform_MetaGGA = ALLOCATED(EnergyDen%mt) &
.AND. (xcpot%exc_is_MetaGGA() .or. xcpot%vx_is_MetaGGA())

call timestart("init_pw_grid")
CALL init_pw_grid(stars,sym,cell,xcpot)
call timestop("init_pw_grid")


! start Lebedev grids sampling loop
DO m=1, a

!Put the charge on the grid, in GGA case also calculate gradients
call timestart("pw_to_grid")
CALL pw_to_grid(xcpot%needs_grad(),input%jspins,noco%l_noco,stars,cell,den%pw,grad,xcpot,rho, x_omega=x(m), y_omega=y(m), z_omega=z(m))
call timestop("pw_to_grid")

ALLOCATE(v_xc,mold=rho)
ALLOCATE(v_xc2,mold=rho)
ALLOCATE(v_x,mold=rho)


call timestart("apply_cutoffs")
CALL xcpot%apply_cutoffs(1.E-6,rho,grad)
call timestop("apply_cutoffs")
#ifdef CPP_LIBXC
if(perform_MetaGGA .and. kinED%set) then
CALL xcpot%get_vxc(input%jspins,rho,v_xc, v_x,grad, kinEnergyDen_KS=kinED%is)
else
CALL xcpot%get_vxc(input%jspins,rho,v_xc,v_x,grad)
endif
#else
call timestart("get_vxc")
CALL xcpot%get_vxc(input%jspins,rho,v_xc,v_x,grad)
call timestop("get_vxc")
#endif

SELECT TYPE(xcpot)
TYPE IS (t_xcpot_libxc)
l_libxc=.TRUE.
IF (xcpot%needs_grad()) THEN
CALL libxc_postprocess_gga_pw(xcpot,stars,cell,v_xc,grad)
CALL libxc_postprocess_gga_pw(xcpot,stars,cell,v_x,grad)
END IF
END SELECT
!IF (l_libxc.AND.xcpot%needs_grad()) THEN
! CALL save_npy('vxc_gga_ir_libxc.npy',v_xc)
!ELSE IF (l_libxc.AND.(.NOT.xcpot%needs_grad())) THEN
! CALL save_npy('vxc_lda_ir_libxc.npy',v_xc)
!ELSE IF ((.NOT.l_libxc).AND.xcpot%needs_grad()) THEN
! CALL save_npy('vxc_gga_ir_inbuild.npy',v_xc)
!ELSE
! CALL save_npy('vxc_lda_ir_inbuild.npy',v_xc)
!END IF
v_xc2=v_xc
! do sum of v_xc , v_xc2 and v_x
v_xc_tmp=v_xc_tmp+w(m)*v_xc
v_xc2_tmp=v_xc2_tmp+w(m)*v_xc2
v_x_tmp=v_x_tmp+w(m)*v_x
! release v_xc, v_xc2 and v_x
deallocate(v_xc, v_xc2, v_x)
!calculate the ex.-cor energy density
IF (ALLOCATED(exc%pw_w)) THEN
ALLOCATE ( e_xc(SIZE(rho,1),1) ); e_xc=0.0
#ifdef CPP_LIBXC
IF(kinED%set) THEN
CALL xcpot%get_exc(input%jspins,rho,e_xc(:,1),grad, kinED%is, mt_call=.False.)
ELSE
CALL xcpot%get_exc(input%jspins,rho,e_xc(:,1),grad, mt_call=.False.)
ENDIF

#else
call timestart("get_exc")
CALL xcpot%get_exc(input%jspins,rho,e_xc(:,1),grad, mt_call=.False.)
call timestop("get_exc")
#endif
! do sum of e_xc
e_xc_tmp=e_xc_tmp+w(m)*e_xc
! release e_xc
deallocate(e_xc)
ENDIF
! End of Lebedev grids sampling loop
END DO
! reallocation
ALLOCATE(v_xc,mold=rho)
ALLOCATE(v_xc2,mold=rho)
ALLOCATE(v_x,mold=rho)
IF (ALLOCATED(exc%pw_w)) THEN
ALLOCATE ( e_xc(SIZE(rho,1),1) ); e_xc=0.0
ENDIF
!recalculate the potentials
v_xc=v_xc_tmp
v_xc2=v_xc2_tmp
v_x=v_x_tmp
!recalculate the ex.-cor energy density
e_xc=e_xc_tmp
!Put the potentials in rez. space.
call timestart("pw_from_grid")
CALL pw_from_grid(stars,v_xc,vTot%pw,vTot%pw_w)
CALL pw_from_grid(stars,v_xc2,vxc%pw)
CALL pw_from_grid(stars,v_x,vx%pw,vx%pw_w)
call timestop("pw_from_grid")
!Put the ex.-cor energy density in rez. space.
call timestart("pw_from_grid")
CALL pw_from_grid(stars,e_xc,exc%pw,exc%pw_w)
call timestop("pw_from_grid")
call timestart("finish_pw_grid")
CALL finish_pw_grid()
call timestop("finish_pw_grid")
deallocate(x, y, z, w)
deallocate(v_xc_tmp, v_xc2_tmp, v_x_tmp, e_xc_tmp)

the bug:
Signal 11 detected on PE: 0
FLEUR tried an illegal memory access.
This might indicate a bug, or your calculation
failed because you were running out of memory
Please check and report if you believe you found a bug
Rank:0 used 2.546 0.108 GB/ 2670088 kB
Last known location:
Last timer:pw_to_grid
Timerstack:
Timer:Vxc in interstitial
Timer:generation of potential
Timer:Iteration
Timer:Total Run
*****************************************
Abort(0) on node 0 (rank 0 in comm 0): application called MPI_Abort(MPI_COMM_WORLD, 0) - process 0


some explanation: pw_to_grid was also modified and added three optional input.

this problem is fixed