Different results with same input in parallel yambo 4.0.1.
Moderators: Davide Sangalli, andrea.ferretti, myrta gruning, andrea marini, Daniele Varsano, Conor Hogan, Nicola Spallanzani
-
Michael.Friedrich
- Posts: 15
- Joined: Tue Aug 11, 2015 11:51 am
Re: Different results with same input in parallel yambo 4.0.
I forgot to mention that the ground state supercell calculations were done with a charged cell and spin in contrast to the ones for the unit cell which of coarse is a major difference...
You do not have the required permissions to view the files attached to this post.
Michael Friedrich, Universität Paderborn, Germany
-
Daniele Varsano
- Posts: 3816
- Joined: Tue Mar 17, 2009 2:23 pm
- Contact:
Re: Different results with same input in parallel yambo 4.0.
Dear Michael,
the two cases are not equivalent as in one of the two cases you run a spin-polarized calculation.
In this case, if you build the input adding the "-V qp" option you can recognize that the name of the variable(s) will be:
this is reason why the XfnQP_E variable was not recognized.
Best,
Daniele
the two cases are not equivalent as in one of the two cases you run a spin-polarized calculation.
In this case, if you build the input adding the "-V qp" option you can recognize that the name of the variable(s) will be:
Code: Select all
XfnQP_up_E=
XfnQP_dn_E=Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
-
Michael.Friedrich
- Posts: 15
- Joined: Tue Aug 11, 2015 11:51 am
Re: Different results with same input in parallel yambo 4.0.
Sorry for that question. I just used the input from non spin polarized calculations without checking on further changes to be made 
I was able to produce nice results with YAMBO
I still can use some piece of advice to perform a more efficient parallelization (on 64 cpus) because the run on 4 of 64 cpus is always taking much more time and seems to start later than on the others which delays the whole calculation very much...
I'm using
which ends up like
Do you have some ideas to overcome this problem?
Thank you,
Michael
I was able to produce nice results with YAMBO
I still can use some piece of advice to perform a more efficient parallelization (on 64 cpus) because the run on 4 of 64 cpus is always taking much more time and seems to start later than on the others which delays the whole calculation very much...
I'm using
Code: Select all
X_q_0_CPU= "1 8 8" # [PARALLEL] CPUs for each role
X_q_0_ROLEs= "k v c" # [PARALLEL] CPUs roles (k,c,v)
X_q_0_nCPU_invert=0 # [PARALLEL] CPUs for matrix inversion
X_finite_q_CPU= "1 1 8 8" # [PARALLEL] CPUs for each role
X_finite_q_ROLEs= "q k v c" # [PARALLEL] CPUs roles (q,k,c,v)
X_finite_q_nCPU_invert=0 # [PARALLEL] CPUs for matrix inversion
Code: Select all
<04d-05h-11m-28s> P0001: Xo@q[1] |################################## | [085%] 04d-05h-06m-39s(E) 04d-22h-56m-40s(X)
<02d-12h-28m-45s> P0002: Xo@q[1] |####################################### | [097%] 02d-12h-23m-53s(E) 02d-13h-56m-44s(X)
<04d-05h-25m-04s> P0003: Xo@q[1] |#################################### | [090%] 04d-05h-20m-07s(E) 04d-16h-35m-09s(X)
<04d-05h-31m-38s> P0004: Xo@q[1] |##################################### | [092%] 04d-05h-26m-47s(E) 04d-13h-40m-02s(X)
<03d-11h-09m-23s> P0005: Xo@q[1] |####################################### | [097%] 03d-11h-04m-34s(E) 03d-13h-12m-16s(X)
<03d-10h-58m-54s> P0006: Xo@q[1] |####################################### | [097%] 03d-10h-54m-05s(E) 03d-13h-01m-15s(X)
<03d-04h-25m-32s> P0007: Xo@q[1] |####################################### | [097%] 03d-04h-20m-51s(E) 03d-06h-17m-54s(X)
<03d-04h-29m-49s> P0008: Xo@q[1] |####################################### | [097%] 03d-04h-25m-09s(E) 03d-06h-22m-02s(X)
<04d-04h-52m-39s> P0009: Xo@q[1] |##################################### | [092%] 04d-04h-47m-56s(E) 04d-12h-57m-56s(X)
<02d-07h-31m-28s> P0010: Xo@q[1] |####################################### | [097%] 02d-07h-26m-45s(E) 02d-08h-51m-48s(X)
<04d-06h-50m-11s> P0011: Xo@q[1] |###################################### | [095%] 04d-06h-45m-22s(E) 04d-12h-09m-41s(X)
<04d-04h-55m-46s> P0012: Xo@q[1] |###################################### | [095%] 04d-04h-51m-02s(E) 04d-10h-09m-28s(X)
<03d-10h-37m-17s> P0013: Xo@q[1] |####################################### | [097%] 03d-10h-32m-41s(E) 03d-12h-39m-23s(X)
<03d-11h-12m-09s> P0014: Xo@q[1] |####################################### | [097%] 03d-11h-07m-26s(E) 03d-13h-14m-32s(X)
<03d-02h-06m-31s> P0015: Xo@q[1] |####################################### | [097%] 03d-02h-01m-56s(E) 03d-03h-55m-40s(X)
<03d-00h-19m-54s> P0016: Xo@q[1] |####################################### | [097%] 03d-00h-15m-20s(E) 03d-02h-06m-27s(X)
<04d-05h-24m-20s> P0017: Xo@q[1] |############################ | [070%] 04d-05h-19m-44s(E) 06d-00h-45m-00s(X)
<03d-09h-08m-52s> P0018: Xo@q[1] |####################################### | [097%] 03d-09h-04m-26s(E) 03d-11h-08m-36s(X)
<04d-05h-18m-49s> P0019: Xo@q[1] |############################## | [075%] 04d-05h-14m-21s(E) 05d-14h-58m-52s(X)
<04d-05h-10m-11s> P0020: Xo@q[1] |############################### | [077%] 04d-05h-05m-49s(E) 05d-10h-26m-39s(X)
<04d-00h-41m-38s> P0021: Xo@q[1] |####################################### | [097%] 04d-00h-37m-15s(E) 04d-03h-05m-12s(X)
<03d-21h-20m-35s> P0022: Xo@q[1] |####################################### | [097%] 03d-21h-15m-52s(E) 03d-23h-39m-14s(X)
<03d-16h-05m-25s> P0023: Xo@q[1] |####################################### | [097%] 03d-16h-00m-49s(E) 03d-18h-15m-48s(X)
<03d-19h-18m-20s> P0024: Xo@q[1] |####################################### | [097%] 03d-19h-13m-44s(E) 03d-21h-34m-01s(X)
<04d-04h-54m-20s> P0025: Xo@q[1] |########################## | [065%] 04d-04h-49m-44s(E) 06d-11h-06m-09s(X)
<03d-14h-04m-13s> P0026: Xo@q[1] |####################################### | [097%] 03d-13h-59m-26s(E) 03d-16h-11m-29s(X)
<04d-05h-18m-20s> P0027: Xo@q[1] |############################# | [072%] 04d-05h-13m-29s(E) 05d-19h-37m-13s(X)
<04d-05h-35m-30s> P0028: Xo@q[1] |############################## | [075%] 04d-05h-30m-46s(E) 05d-15h-20m-45s(X)
<04d-03h-47m-30s> P0029: Xo@q[1] |####################################### | [097%] 04d-03h-42m-45s(E) 04d-06h-15m-31s(X)
<04d-03h-30m-42s> P0030: Xo@q[1] |####################################### | [097%] 04d-03h-25m-50s(E) 04d-05h-58m-13s(X)
<03d-22h-01m-45s> P0031: Xo@q[1] |####################################### | [097%] 03d-21h-57m-01s(E) 04d-00h-21m-28s(X)
<01d-22h-56m-39s> P0032: Xo@q[1] |########### | [027%] 01d-22h-53m-26s(E) 07d-02h-29m-00s(X)
<01d-20h-11m-16s> P0033: Xo@q[1] |####### | [017%] 01d-20h-07m-56s(E) 10d-12h-08m-24s(X)
<02d-00h-56m-16s> P0034: Xo@q[1] |############## | [035%] 02d-00h-52m-53s(E) 05d-19h-37m-04s(X)
<01d-22h-22m-34s> P0035: Xo@q[1] |######## | [020%] 01d-22h-19m-03s(E) 09d-15h-31m-17s(X)
<03d-08h-11m-45s> P0036: Xo@q[1] |######################## | [060%] 03d-08h-09m-39s(E) 05d-13h-35m-50s(X)
<03d-09h-13m-24s> P0037: Xo@q[1] |################################ | [080%] 03d-09h-11m-19s(E) 04d-05h-29m-09s(X)
<04d-04h-20m-53s> P0038: Xo@q[1] |####################################### | [097%] 04d-04h-15m-54s(E) 04d-06h-49m-58s(X)
<03d-22h-56m-07s> P0039: Xo@q[1] |####################################### | [097%] 03d-22h-51m-23s(E) 04d-01h-17m-11s(X)
<04d-00h-25m-45s> P0040: Xo@q[1] |####################################### | [097%] 04d-00h-20m-59s(E) 04d-02h-48m-31s(X)
<04d-04h-00m-16s> P0041: Xo@q[1] |############################# | [072%] 04d-03h-55m-51s(E) 05d-17h-49m-44s(X)
<03d-01h-56m-37s> P0042: Xo@q[1] |####################################### | [097%] 03d-01h-51m-55s(E) 03d-03h-44m-52s(X)
<04d-03h-50m-08s> P0043: Xo@q[1] |############################### | [077%] 04d-03h-45m-19s(E) 05d-08h-42m-16s(X)
<04d-03h-59m-10s> P0044: Xo@q[1] |################################ | [080%] 04d-03h-54m-28s(E) 05d-04h-52m-30s(X)
<03d-22h-06m-49s> P0045: Xo@q[1] |####################################### | [097%] 03d-22h-02m-07s(E) 04d-00h-26m-47s(X)
<03d-19h-27m-13s> P0046: Xo@q[1] |####################################### | [097%] 03d-19h-22m-30s(E) 03d-21h-43m-05s(X)
<03d-15h-37m-28s> P0047: Xo@q[1] |####################################### | [097%] 03d-15h-32m-45s(E) 03d-17h-47m-12s(X)
<03d-16h-26m-39s> P0048: Xo@q[1] |####################################### | [097%] 03d-16h-21m-57s(E) 03d-18h-37m-13s(X)
<04d-03h-42m-38s> P0049: Xo@q[1] |############################### | [077%] 04d-03h-37m-54s(E) 05d-08h-32m-35s(X)
<02d-20h-19m-07s> P0050: Xo@q[1] |####################################### | [097%] 02d-20h-14m-24s(E) 02d-21h-58m-59s(X)
<04d-06h-25m-10s> P0051: Xo@q[1] |################################### | [087%] 04d-06h-20m-27s(E) 04d-20h-57m-40s(X)
<04d-06h-14m-42s> P0052: Xo@q[1] |#################################### | [090%] 04d-06h-10m-07s(E) 04d-17h-30m-59s(X)
<03d-15h-39m-20s> P0053: Xo@q[1] |####################################### | [097%] 03d-15h-34m-37s(E) 03d-17h-48m-33s(X)
<03d-15h-39m-12s> P0054: Xo@q[1] |####################################### | [097%] 03d-15h-34m-24s(E) 03d-17h-48m-32s(X)
<03d-09h-33m-51s> P0055: Xo@q[1] |####################################### | [097%] 03d-09h-29m-09s(E) 03d-11h-34m-30s(X)
<03d-10h-59m-56s> P0056: Xo@q[1] |####################################### | [097%] 03d-10h-55m-14s(E) 03d-13h-02m-26s(X)
<04d-04h-01m-26s> P0057: Xo@q[1] |################################# | [082%] 04d-03h-56m-49s(E) 05d-01h-08m-47s(X)
<02d-14h-28m-34s> P0058: Xo@q[1] |####################################### | [097%] 02d-14h-23m-57s(E) 02d-15h-59m-27s(X)
<04d-03h-21m-13s> P0059: Xo@q[1] |################################### | [087%] 04d-03h-16m-28s(E) 04d-17h-27m-04s(X)
<04d-04h-16m-16s> P0060: Xo@q[1] |#################################### | [090%] 04d-04h-11m-41s(E) 04d-15h-19m-18s(X)
<03d-14h-16m-36s> P0061: Xo@q[1] |####################################### | [097%] 03d-14h-11m-54s(E) 03d-16h-24m-07s(X)
<03d-13h-42m-49s> P0062: Xo@q[1] |####################################### | [097%] 03d-13h-38m-00s(E) 03d-15h-49m-06s(X)
<03d-07h-14m-42s> P0063: Xo@q[1] |####################################### | [097%] 03d-07h-10m-00s(E) 03d-09h-11m-46s(X)
<03d-08h-36m-43s> P0064: Xo@q[1] |####################################### | [097%] 03d-08h-32m-01s(E) 03d-10h-35m-06s(X)
[michaelf@fe1 work]$ ./statuscheck 64 4d24000/LOG/
cpus: 64
path: 4d24000/LOG/
<01d-21h-22m-20s> P0001: Xo@q[1] |##################### | [052%] 01d-21h-15m-34s(E) 03d-14h-10m-25s(X)
<01d-21h-24m-04s> P0002: Xo@q[1] |############################ | [070%] 01d-21h-17m-20s(E) 02d-16h-41m-54s(X)
<01d-21h-26m-20s> P0003: Xo@q[1] |##################### | [052%] 01d-21h-19m-28s(E) 03d-14h-19m-49s(X)
<01d-21h-00m-32s> P0004: Xo@q[1] |##################### | [052%] 01d-20h-53m-49s(E) 03d-13h-30m-37s(X)
<02d-18h-37m-17s> P0005: Xo@q[1] |####################################### | [097%] 02d-18h-27m-59s(E) 02d-20h-09m-28s(X)
<02d-17h-59m-12s> P0006: Xo@q[1] |####################################### | [097%] 02d-17h-49m-56s(E) 02d-19h-30m-18s(X)
<02d-12h-03m-32s> P0007: Xo@q[1] |####################################### | [097%] 02d-11h-54m-25s(E) 02d-13h-26m-15s(X)
<01d-14h-22m-34s> P0008: Xo@q[1] |####################################### | [097%] 01d-14h-06m-30s(E) 01d-15h-04m-51s(X)
<01d-18h-22m-03s> P0009: Xo@q[1] |####################################### | [097%] 01d-18h-06m-02s(E) 01d-19h-10m-29s(X)
<01d-07h-47m-47s> P0010: Xo@q[1] |####################################### | [097%] 01d-07h-31m-42s(E) 01d-08h-20m-05s(X)
<01d-22h-20m-48s> P0011: Xo@q[1] |####################################### | [097%] 01d-22h-04m-33s(E) 01d-23h-14m-59s(X)
<01d-21h-32m-27s> P0012: Xo@q[1] |####################################### | [097%] 01d-21h-16m-20s(E) 01d-22h-25m-58s(X)
<01d-17h-30m-20s> P0013: Xo@q[1] |####################################### | [097%] 01d-17h-14m-13s(E) 01d-18h-17m-27s(X)
<01d-17h-24m-08s> P0014: Xo@q[1] |####################################### | [097%] 01d-17h-07m-53s(E) 01d-18h-11m-06s(X)
<01d-14h-28m-05s> P0015: Xo@q[1] |####################################### | [097%] 01d-14h-12m-01s(E) 01d-15h-10m-38s(X)
<01d-13h-04m-03s> P0016: Xo@q[1] |####################################### | [097%] 01d-12h-48m-00s(E) 01d-13h-44m-30s(X)
<01d-21h-59m-30s> P0017: Xo@q[1] |####################################### | [097%] 01d-21h-43m-26s(E) 01d-22h-53m-25s(X)
<01d-14h-39m-43s> P0018: Xo@q[1] |####################################### | [097%] 01d-14h-23m-39s(E) 01d-15h-21m-58s(X)
<01d-22h-46m-10s> P0019: Xo@q[1] |####################################### | [097%] 01d-22h-29m-58s(E) 01d-23h-40m-52s(X)
<01d-23h-21m-57s> P0020: Xo@q[1] |####################################### | [097%] 01d-23h-05m-53s(E) 02d-00h-17m-47s(X)
<01d-17h-23m-21s> P0021: Xo@q[1] |####################################### | [097%] 01d-17h-07m-18s(E) 01d-18h-10m-18s(X)
<01d-17h-28m-47s> P0022: Xo@q[1] |####################################### | [097%] 01d-17h-12m-36s(E) 01d-18h-15m-28s(X)
<01d-17h-01m-12s> P0023: Xo@q[1] |####################################### | [097%] 01d-16h-45m-09s(E) 01d-17h-47m-47s(X)
<01d-17h-33m-09s> P0024: Xo@q[1] |####################################### | [097%] 01d-17h-17m-05s(E) 01d-18h-20m-29s(X)
<01d-15h-59m-51s> P0025: Xo@q[1] |####################################### | [097%] 01d-15h-43m-48s(E) 01d-16h-44m-45s(X)
<01d-16h-49m-07s> P0026: Xo@q[1] |####################################### | [097%] 01d-16h-33m-03s(E) 01d-17h-35m-23s(X)
<02d-00h-43m-44s> P0027: Xo@q[1] |####################################### | [097%] 02d-00h-27m-32s(E) 02d-01h-41m-45s(X)
<02d-00h-42m-28s> P0028: Xo@q[1] |####################################### | [097%] 02d-00h-26m-24s(E) 02d-01h-40m-35s(X)
<03d-10h-14m-45s> P0029: Xo@q[1] |####################################### | [097%] 03d-09h-55m-40s(E) 03d-12h-01m-33s(X)
<03d-09h-45m-15s> P0030: Xo@q[1] |####################################### | [097%] 03d-09h-25m-54s(E) 03d-11h-30m-55s(X)
<03d-09h-13m-03s> P0031: Xo@q[1] |####################################### | [097%] 03d-08h-54m-08s(E) 03d-10h-58m-31s(X)
<03d-09h-36m-46s> P0032: Xo@q[1] |####################################### | [097%] 03d-09h-17m-50s(E) 03d-11h-21m-49s(X)
<03d-21h-20m-33s> P0033: Xo@q[1] |####################################### | [097%] 03d-21h-01m-14s(E) 03d-23h-23m-51s(X)
<03d-07h-46m-19s> P0034: Xo@q[1] |####################################### | [097%] 03d-07h-26m-59s(E) 03d-09h-28m-34s(X)
<03d-22h-10m-57s> P0035: Xo@q[1] |####################################### | [097%] 03d-21h-51m-22s(E) 04d-00h-14m-40s(X)
<03d-21h-47m-18s> P0036: Xo@q[1] |####################################### | [097%] 03d-21h-27m-58s(E) 03d-23h-51m-31s(X)
<01d-18h-29m-05s> P0037: Xo@q[1] |####################################### | [097%] 01d-18h-12m-45s(E) 01d-19h-17m-13s(X)
<01d-18h-20m-23s> P0038: Xo@q[1] |####################################### | [097%] 01d-18h-03m-55s(E) 01d-19h-08m-18s(X)
<01d-17h-54m-44s> P0039: Xo@q[1] |####################################### | [097%] 01d-17h-38m-22s(E) 01d-18h-41m-53s(X)
<01d-17h-26m-48s> P0040: Xo@q[1] |####################################### | [097%] 01d-17h-10m-26s(E) 01d-18h-13m-16s(X)
<04d-08h-14m-41s> P0041: Xo@q[1] |####################################### | [097%] 04d-08h-04m-52s(E) 04d-10h-44m-57s(X)
<03d-17h-25m-07s> P0042: Xo@q[1] |####################################### | [097%] 03d-17h-15m-23s(E) 03d-19h-31m-30s(X)
<04d-09h-57m-40s> P0043: Xo@q[1] |####################################### | [097%] 04d-09h-47m-44s(E) 04d-12h-30m-04s(X)
<04d-09h-39m-08s> P0044: Xo@q[1] |####################################### | [097%] 04d-09h-29m-25s(E) 04d-12h-10m-46s(X)
<04d-02h-30m-48s> P0045: Xo@q[1] |####################################### | [097%] 04d-02h-21m-05s(E) 04d-04h-50m-49s(X)
<04d-02h-24m-28s> P0046: Xo@q[1] |####################################### | [097%] 04d-02h-14m-32s(E) 04d-04h-44m-35s(X)
<04d-01h-54m-59s> P0047: Xo@q[1] |####################################### | [097%] 04d-01h-45m-16s(E) 04d-04h-14m-11s(X)
<04d-00h-18m-09s> P0048: Xo@q[1] |####################################### | [097%] 04d-00h-08m-25s(E) 04d-02h-35m-23s(X)
<03d-18h-02m-00s> P0049: Xo@q[1] |####################################### | [097%] 03d-17h-45m-03s(E) 03d-20h-01m-39s(X)
<03d-03h-21m-42s> P0050: Xo@q[1] |####################################### | [097%] 03d-03h-04m-44s(E) 03d-04h-58m-52s(X)
<03d-18h-23m-14s> P0051: Xo@q[1] |####################################### | [097%] 03d-18h-06m-02s(E) 03d-20h-24m-29s(X)
<03d-18h-09m-23s> P0052: Xo@q[1] |####################################### | [097%] 03d-17h-52m-24s(E) 03d-20h-10m-28s(X)
<03d-10h-21m-44s> P0053: Xo@q[1] |####################################### | [097%] 03d-10h-04m-23s(E) 03d-12h-09m-25s(X)
<03d-10h-03m-46s> P0054: Xo@q[1] |####################################### | [097%] 03d-09h-46m-34s(E) 03d-11h-51m-16s(X)
<03d-10h-39m-02s> P0055: Xo@q[1] |####################################### | [097%] 03d-10h-22m-06s(E) 03d-12h-28m-46s(X)
<01d-14h-01m-21s> P0056: Xo@q[1] |####################################### | [097%] 01d-13h-45m-18s(E) 01d-14h-43m-16s(X)
<01d-12h-11m-08s> P0057: Xo@q[1] |####################################### | [097%] 01d-11h-55m-05s(E) 01d-12h-49m-42s(X)
<01d-11h-30m-52s> P0058: Xo@q[1] |####################################### | [097%] 01d-11h-14m-49s(E) 01d-12h-08m-50s(X)
<01d-20h-59m-03s> P0059: Xo@q[1] |####################################### | [097%] 01d-20h-42m-52s(E) 01d-21h-50m-55s(X)
<03d-20h-48m-10s> P0060: Xo@q[1] |####################################### | [097%] 03d-20h-31m-33s(E) 03d-22h-53m-04s(X)
<01d-15h-08m-07s> P0061: Xo@q[1] |####################################### | [097%] 01d-14h-52m-04s(E) 01d-15h-51m-08s(X)
<01d-07h-53m-02s> P0062: Xo@q[1] |####################################### | [097%] 01d-07h-36m-55s(E) 01d-08h-25m-33s(X)
<01d-14h-20m-57s> P0063: Xo@q[1] |####################################### | [097%] 01d-14h-04m-54s(E) 01d-15h-02m-52s(X)
<01d-06h-52m-01s> P0064: Xo@q[1] |####################################### | [097%] 01d-06h-35m-58s(E) 01d-07h-22m-32s(X)
Thank you,
Michael
Michael Friedrich, Universität Paderborn, Germany
-
Daniele Varsano
- Posts: 3816
- Joined: Tue Mar 17, 2009 2:23 pm
- Contact:
Re: Different results with same input in parallel yambo 4.0.
Dear Michael,
Indeed your calculation seems totally unbalanced.
can you please post the complete input/report/and output files?
It can also be a problem of some nodes of your machines, or unbalanaced parallelization.
Please upload a tgz files with all the info including the LOG directory content (copy paste in the post is not easy to read!).
Best,
Daniele
Indeed your calculation seems totally unbalanced.
can you please post the complete input/report/and output files?
It can also be a problem of some nodes of your machines, or unbalanaced parallelization.
Please upload a tgz files with all the info including the LOG directory content (copy paste in the post is not easy to read!).
Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
-
Michael.Friedrich
- Posts: 15
- Joined: Tue Aug 11, 2015 11:51 am
Re: Different results with same input in parallel yambo 4.0.
Dear Daniele,
After all it was a problem at some nodes on the cluster. yambo 4.0.2. works quick and reliable now concerning tddft. The problems I encountered so far made me ask quick for help instead of checking everything thoroughly; I'm sorry for that.
One last question - can I use the output gained from a quantum-espresso calculation with hybrid functionals (like HSE06) in yambo to do tddft calculations?
Best,
Michael
After all it was a problem at some nodes on the cluster. yambo 4.0.2. works quick and reliable now concerning tddft. The problems I encountered so far made me ask quick for help instead of checking everything thoroughly; I'm sorry for that.
One last question - can I use the output gained from a quantum-espresso calculation with hybrid functionals (like HSE06) in yambo to do tddft calculations?
Best,
Michael
Michael Friedrich, Universität Paderborn, Germany
-
Daniele Varsano
- Posts: 3816
- Joined: Tue Mar 17, 2009 2:23 pm
- Contact:
Re: Different results with same input in parallel yambo 4.0.
Dera Michael,
good to hear things are working.
I've never done it, anyway I do not see any problem in using energies and wfs coming from an HSE calculation for TDDFT. Actually GW is another story instead.
Best,
Daniele
good to hear things are working.
I've never done it, anyway I do not see any problem in using energies and wfs coming from an HSE calculation for TDDFT. Actually GW is another story instead.
Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
-
Davide Sangalli
- Posts: 614
- Joined: Tue May 29, 2012 4:49 pm
- Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
- Contact:
Re: Different results with same input in parallel yambo 4.0.
Dera Michael,
if you are thinking about computing absorption spectra within TD-DFT on top of HSE it might work.
However, yambo should compute <x>=< (p - [x,V_nl(pseudo)]-[x,V_xc(HSE)]) > / delta_E (The square brackets are for the commutator.)
I.e. HSE, if I'm not wrong, is an hybrid functional with some fraction of exchange which is non local.
Instead yambo will neglect the term [x,V_xc(HSE)] assuming V_xc is still local.
On the other hand for QP correction the new libxc support should work also for HSE, thus you should not have many problems in that case.
The wrong <x> could induce a small error in the screening at q=0 but I expect the effect to be really negligile.
(... well, indeed, I'm not sure it has been already released inside the GPL version.)
Best,
D.
if you are thinking about computing absorption spectra within TD-DFT on top of HSE it might work.
However, yambo should compute <x>=< (p - [x,V_nl(pseudo)]-[x,V_xc(HSE)]) > / delta_E (The square brackets are for the commutator.)
I.e. HSE, if I'm not wrong, is an hybrid functional with some fraction of exchange which is non local.
Instead yambo will neglect the term [x,V_xc(HSE)] assuming V_xc is still local.
On the other hand for QP correction the new libxc support should work also for HSE, thus you should not have many problems in that case.
The wrong <x> could induce a small error in the screening at q=0 but I expect the effect to be really negligile.
(... well, indeed, I'm not sure it has been already released inside the GPL version.)
Best,
D.
Davide Sangalli, PhD
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/
-
Michael.Friedrich
- Posts: 15
- Joined: Tue Aug 11, 2015 11:51 am
Re: Different results with same input in parallel yambo 4.0.
Dear Daniele and Davide,
In general, I didn't think tddft on top of a hybrid functional calculation would be much of a problem, too.
I'm currently testing but unfortunately, there always occurs this problem in yambo, probably related to the format of the wavefunction files?!
(there was no problem within p2y)
in r_optics_* there are no WF G-vectors displayed
Best,
Michael
In general, I didn't think tddft on top of a hybrid functional calculation would be much of a problem, too.
I'm currently testing but unfortunately, there always occurs this problem in yambo, probably related to the format of the wavefunction files?!
(there was no problem within p2y)
Code: Select all
<01s> [01] CPU structure, Files & I/O Directories
<01s> CPU-Threads:1(CPU)-24(threads)-24(threads@X)-24(threads@DIP)
<01s> [02] CORE Variables Setup
<01s> [02.01] Unit cells
<01s> [02.02] Symmetries
<01s> [02.03] RL shells
<01s> [02.04] K-grid lattice
<01s> [02.05] Energies [ev] & Occupations
<01s> [03] Transferred momenta grid
<01s> [04] External corrections
<02s> [05] Optics
<02s> [DIP] Checking dipoles header
<02s> [x,Vnl] computed using 130 projectors
<02s> [WARNING] [x,Vnl] slows the Dipoles computation. To neglect it rename the ns.kb_pp file
<02s> [WF] Performing Wave-Functions I/O from ./SAVE[compute-4-4:17616] *** Process received signal ***
[compute-4-4:17616] Signal: Bus error (7)
[compute-4-4:17616] Signal code: (128)
[compute-4-4:17616] Failing at address: (nil)
[compute-4-4:17616] [ 0] /lib64/libpthread.so.0 [0x3493e0e4c0]
[compute-4-4:17616] [ 1] /opt/pc2_amd64/openmpi/1.3.3-intel/lib/libopen-pal.so.0(opal_memory_ptmalloc2_int_free+0x75) [0x2b510c586615]
[compute-4-4:17616] [ 2] /opt/pc2_amd64/openmpi/1.3.3-intel/lib/libopen-pal.so.0 [0x2b510c587040]
[compute-4-4:17616] [ 3] /lib64/libc.so.6(cfree+0x27) [0x34936758a7]
[compute-4-4:17616] [ 4] ../../Tools/yambo-4.0.2-rev.90/bin/yambo(for_dealloc_allocatable+0xf8) [0x80a988]
[compute-4-4:17616] [ 5] ../../Tools/yambo-4.0.2-rev.90/bin/yambo(wf_load_+0x1e0a) [0x4e6ffa]
[compute-4-4:17616] [ 6] ../../Tools/yambo-4.0.2-rev.90/bin/yambo(dipole_transverse_+0x513) [0x48fff3]
[compute-4-4:17616] [ 7] ../../Tools/yambo-4.0.2-rev.90/bin/yambo(dipole_driver_+0x4f0) [0x48deb0]
[compute-4-4:17616] [ 8] ../../Tools/yambo-4.0.2-rev.90/bin/yambo(x_os_+0x2465) [0x4955c5]
[compute-4-4:17616] [ 9] ../../Tools/yambo-4.0.2-rev.90/bin/yambo(o_driver_+0xa2e) [0x48aa9e]
[compute-4-4:17616] [10] ../../Tools/yambo-4.0.2-rev.90/bin/yambo(yambo_driver_+0x56e) [0x439e5e]
[compute-4-4:17616] [11] ../../Tools/yambo-4.0.2-rev.90/bin/yambo(main+0xd1b) [0x43833b]
[compute-4-4:17616] [12] /lib64/libc.so.6(__libc_start_main+0xf4) [0x349361d974]
[compute-4-4:17616] [13] ../../Tools/yambo-4.0.2-rev.90/bin/yambo [0x437569]
[compute-4-4:17616] *** End of error message ***
Bus error
Code: Select all
[RD./SAVE//ns.db1]------------------------------------------
Bands : 60
K-points : 12
G-vectors [RL space]: 67681
Components [wavefunctions]: 8449
Symmetries [spatial+T-rev]: 6
Spinor components : 1
Spin polarizations : 1
Temperature [ev]: 0.000000
Electrons : 68.00000
WF G-vectors :0
Max atoms/species : 6
No. of atom species : 3
Magnetic symmetries : no
Michael
You do not have the required permissions to view the files attached to this post.
Michael Friedrich, Universität Paderborn, Germany
-
Michael.Friedrich
- Posts: 15
- Joined: Tue Aug 11, 2015 11:51 am
Re: Different results with same input in parallel yambo 4.0.
Dear all,
first of all a happy new year!
I did some tests with silicon. There p2y worked and I could run some optial calculations but the result is three times lower for hse compared to calculations with pbesol.
A spectrum is attached hse pbesol si.txt. It's an eps file.
For lithium niobate those calculations still aren't possible at all due to a problem possibly within p2y. As expected, a pbesol calculation with pw.x produces files of the same size and with the same information (excluding XC and different eigenenergies; as far as human readable) as a calculation with hse and everything works fine. Just for a test I exchanged the information in data-file.xml from hse to pbesol but the G vectors are still not found.
So I have no idea what to do anymore...
Best,
Michael
first of all a happy new year!
I did some tests with silicon. There p2y worked and I could run some optial calculations but the result is three times lower for hse compared to calculations with pbesol.
Code: Select all
<---> DBs path set to .
<---> Index file set to data-file.xml
<---> Header/K-points/Energies... done
<---> Cell data... done
<---> Atomic data... done
<---> Symmetries...[SI no]...[I yes]...[-I no]...[TR yes]
<---> XC functional...HSE06(XC)
<---> EXX fraction... 0.250000
<---> EXX screening... 0.106000
<---> K-points mesh... done
<---> RL vectors... done
<---> IGK arrays... done
<---> Energies... done
<---> :: Electrons : 8.000000
<---> :: Temperature [ev]: 0.000000
<---> :: Lattice factors [a.u.]: 5.164295 5.164295 5.164295
<---> :: K-points : 16
<---> :: Bands : 16
<---> :: Spinor components : 1
<---> :: Spin polarizations : 1
<---> :: Spin orbit coupling : no
<---> :: Symmetries [spatial]: 24
<---> :: [T-rev]: yes
<---> :: Max WF components : 1676
<---> :: RL vectors (WF): 2181
<---> :: RL vectors (CHARGE): 13323
<---> :: XC potential : HSE06(XC)
<---> :: Atomic species : 1
<---> :: Max atoms/species : 2
<---> == DB1 (Gvecs and more) ...
<---> ... Database done
<---> == DB2 (wavefunctions) ... done ==
<---> == DB3 (PseudoPotential) ... done ==
<01s> == P2Y completed ==
For lithium niobate those calculations still aren't possible at all due to a problem possibly within p2y. As expected, a pbesol calculation with pw.x produces files of the same size and with the same information (excluding XC and different eigenenergies; as far as human readable) as a calculation with hse and everything works fine. Just for a test I exchanged the information in data-file.xml from hse to pbesol but the G vectors are still not found.
So I have no idea what to do anymore...
Best,
Michael
You do not have the required permissions to view the files attached to this post.
Michael Friedrich, Universität Paderborn, Germany
-
myrta gruning
- Posts: 240
- Joined: Tue Mar 17, 2009 11:38 am
- Contact:
Re: Different results with same input in parallel yambo 4.0.
Dear Michael
Thank you for sharing the test on Si. The onset for the HSE spectrum seems quite large, I will check this. Do you know if there are known values from the literature?
Anyway as Davide pointed out, the <x> is not calculated properly when nonlocal xc potential are used as in the case of HSE.
<x>=< (p - [x,V_nl(pseudo)]-[x,V_xc(HSE)]) > / delta_E
This is easy to understand: in fact <p> is +/- the same for PBE and for HSE. On the other hand delta_E is much larger for HSE. Then when [x,V_xc(HSE)] is not accounted for (as it is the case in the present implementation), the oscillator strengths of HSE are much smaller than for the PSESOL and as a consequence the optical spectrum too low as in your results.
So I think that unfortunately HSE cannot be used for TDDFT calculations for the moment.
Thank you for sharing the test on Si. The onset for the HSE spectrum seems quite large, I will check this. Do you know if there are known values from the literature?
Anyway as Davide pointed out, the <x> is not calculated properly when nonlocal xc potential are used as in the case of HSE.
<x>=< (p - [x,V_nl(pseudo)]-[x,V_xc(HSE)]) > / delta_E
This is easy to understand: in fact <p> is +/- the same for PBE and for HSE. On the other hand delta_E is much larger for HSE. Then when [x,V_xc(HSE)] is not accounted for (as it is the case in the present implementation), the oscillator strengths of HSE are much smaller than for the PSESOL and as a consequence the optical spectrum too low as in your results.
So I think that unfortunately HSE cannot be used for TDDFT calculations for the moment.
Dr Myrta Grüning
School of Mathematics and Physics
Queen's University Belfast - Northern Ireland
http://www.researcherid.com/rid/B-1515-2009
School of Mathematics and Physics
Queen's University Belfast - Northern Ireland
http://www.researcherid.com/rid/B-1515-2009