-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathCompiling_OpenMolcas.txt
98 lines (75 loc) · 10.1 KB
/
Compiling_OpenMolcas.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
On Cedar, 2022 November 23 version of HPQC/Add_Basis_Sets: make -j6 started at 6:17am finished at 6:29am
On Cedar, 2022 November 23 version of HPQC/Add_Basis_Sets: cmake from a folder where O'Molcas was already built, started at 6:34am finished at 6:37am (just chnaged the names of two basis_library files)
On Cedar, 2022 November 23 version of HPQC/Add_Basis_Sets: make -j6 from a folder where O'Molcas was already built, started at 6:37 finished at 6:39am (just chnaged the names of two basis_library files)
On Cedar, 2022 December 5 version of HPQC/master: cmake -j ../OpenMolcas_HPQC took only from 12:11-12:13pm, when running cmake for the *first* time (which I'd think would be slower than when I ran it on 23 November). I had forgotten to load HDF5 though.
On Cedar, 2022 December 5 version of HPQC/master: cmake -j ../OpenMolcas_HPQC after loading HDF5: real 1m29.423s, user 0m5.035s, sys 0m7.625s
On Cedar, 2022 December 5 version of HPQC/master: cmake ../OpenMolcas_HPQC after loading HDF5: real 1m3.883s, user 0m4.754s, sys 0m8.022s
On Cedar, 2022 December 5 version of HPQC/master: cmake ../OpenMolcas_HPQC after loading HDF5: real 1m0.627s, user 0m4.774s, sys 0m9.020s
On Cedar, 2022 December 5 version of HPQC/master: make -j6 from scratch and after loading HDF5 (make -j crashes): real 9m44.577s, user 15m15.985s, sys 12m20.209s
------------------------------------------------------------------
CONCLUSION: "cmake -j" is not noticably better than "cmake".
On Cedar1, 2022 December 8 Version of HPQC/LEVEL: make -j8 from scratch after loading HDF5: real 12m22.927s, user 15m11.978s, sys 15m4.729s
On Cedar1, 2022 December 22 Version of HPQC/LEVEL: make -j in already compiled folder after loading HDF5 and python, but after signing back in after a power shut-down: real 4m2.268s user 0m58.857s sys 0m50.165s
On Cedar1, 2022 December 28 Version of HPQC: make -j12 in already compiled folder but after a lot of changes were made (HPQC/LEVEL branch was merged into HPQC fork), after loading HDF5 and python: real 2m41.049s, user 0m56.186s, sys 0m43.772s
On Graham1, 2022 December 28 Version of HPQC: make -j12 from scratch wtih hdf5 and python loaded, and using GCC instead of Intel: took only about 4 minutes !!! (don't know exact amount because it ended with an error at real 3m46.388s, user 9m21.013s, sys 7m29.270s, then I restarted it, but the restart didn't take long at all)
On Graham2, 2022 December 30 Version of HPQC: cmake from scratch with python and hdf5 laoded: real 3m15.161s, user 0m3.118s, sys 0m5.017s
On Graham2, 2022 December 30 Version of HPQC: make -j16 from scratch with python and hdf5 laoded: (errors came at the very end due to running out of disk space in /home directory which only had 53G available, but we'd already finished 4 lines starting with [100%].. not much more time is required after that) real 6m9.004s, user 13m37.276s, sys 10m14.497s
On Cedar1 2022 December 30 Version of HPQC: cmake from scratch with python and hdf5 loaded: real 0m28.378s, user 0m5.178s, sys 0m8.920s
On Cedar1 2022 December 30 Version of HPQC: make -j12 from scratch with python and hdf5 loaded: forgot to time it but it was FAST and successful!
On Cedar1 0f529f99 of HPQC (but done on 2/1/23): cmake from scratch with python loaded but hdf5 UNLOADED: real 0m21.768s, user 0m4.894s, sys 0m7.792s
On Cedar1 0f529f99 of HPQC (but done on 2/1/23): make -j16 from scratch with python loaded but hdf5 UNLOADED: real 2m11.886s, user 14m9.632s, sys 3m43.564s (WOW THAT WAS FAST! THANKS TO NO HDF5!)
On Alavi, 2022 December 28 Version of HPQC: make -j2 from scratch without hdf5 or python loaded (since it's a laptop without lmod) and using GCC (GNU 5.4.0) took this long: real 19m50.350s, user 11m6.961s, sys 1m24.856s but took FOREVER just before the HEAD command here:
============
[ 3%] Building Fortran object unit_tests/linalg_mod/bin/CMakeFiles/fruit_molcas.dir/fruit.f90.o
[ 3%] Linking Fortran static library ../../../lib/libfruit_molcas.a
[ 3%] Built target fruit_molcas
HEAD is now at 9bf3b64... Remaple hyb_gga_x_cam_s12
[ 3%] Performing update step for 'Libxc'
============
and during that time "top" suggested that nothing was compiling, so it was very furstrating but fortunately I waited it out and didn't terminate the compilation with CTRL+C.
On Alavi, 2022 December 31 Version of Deyan: time cmake ../OpenMolcas_Deyan -D BIGOT=ON (from scratch, no HDF5): real 0m9.270s, user 0m3.623s, sys 0m0.819s
!WARNING! On Alavi, 2022 December 31 Version of Deyan: time make -j (from scratch, no hdf5): [ 64%] Built target fmm_util_obj Makefile:145: recipe for target 'all' failed make: *** [all] Error 2 real 119m37.677s, user 9m24.948s, sys 319m1.586s. [computer froze fan went loud (luckily only for a few minutes), and lots of data from RAM went to disk.. took about 2 hours for the computer to become usable again]
On Alavi (NAG image: registry.gitlab.com/molcas/dockerfiles/nag, however by default it still used GCC), 2023 January 10 Version of HPQC: make -j2 from scratch without hdf5 or python laoded: real 19m25.374s, user 15m52.934s, sys 1m51.996s
On Cedar1 2023 January 10 Version of HPQC: after using cmake ../OpenMolcas_HPQC_2022_12_30 -D CMAKE_Fortran_COMPILER=/cvmfs/restricted.computecanada.ca/easybuild/software/2020/Core/intel/2020.1.217/compilers_and_libraries_2020.1.217/linux/bin/intel64/ifort without HDF5 and with python, from scratch: make -j18 took real 11m12.933s, user 27m45.794s, sys 7m11.806s
On lamodel.unifal-mg.edu.br/orange i7-4770 CPU @ 3.40GHz (Haswell 2013), 2023 January 11 Version of HPQC: after using cmake ../OpenMolcas -D BIGOT=ON -D CMAKE_Fortran_COMPILER=/opt/intel/oneapi/compiler/2023.0.0/linux/bin/intel64/ifort without hdf5 I hope: time make -j4 leads to: [ 83%] Built target gugaci_obj make: *** [Makefile:158: all] Error 2, real 39m42,091s, user 13m49,078s, sys 1m48,013s. After fixing the error, it took real 4m59,598s, user 3m16,962s, sys 0m26,274s to finish 100.0%
On Alavi (Intel-phusion image: registry.gitlab.com/molcas/dockerfiles/intel-phusion, using /opt/intel/oneapi/compiler/2021.4.0/linux/bin/intel64/ifort), 2023 January 11 Version of HPQC: make -j2 from scratch without hdf5 or python loaded: real 26m58.355s, user 26m4.155s, sys 3m7.291s)
On Alavi (Intel-phusion image: registry.gitlab.com/molcas/dockerfiles/intel-phusion, using /opt/intel/oneapi/compiler/2021.4.0/linux/bin/intel64/ifort), 2023 January 13 Version of HPQC: make -j from scratch without hdf5 or python loaded: real 49m0.798s user 23m55.496s sys 3m1.829s
On Alavi, 2022 January 14 version of HPQC: Ran "make -j2" after minor change, and it was fast. Ran "cmake" with BIGOT=ON and "make -j2" was slower, but seemed like it would take as long as what the next compilation took (see later in this line), but it failed due to the rdvec problem that's unrelated to LEVEL, so I re-ran "cmake" with BIGOT=OFF then "make -j2" took: real 4m19.909s, user 7m36.317s, sys 0m54.000s
On Alavi, 2022 January 17 version of HPQC (Nagfor, garble, bigot=on, but with: set (FFLAGS_NAG_GARBLE "-O0 -g -gline -nan") instead of -O0: cmake ../OpenMolcas -D CMAKE_Fortran_COMPILER=/usr/local/bin/nagfor -D CMAKE_BUILD_TYPE=Garble -D BIGOT=ON, followed by real 17m5.837s, user 8m38.185s, sys 1m58.930s. With -02 we get (when building from scratch): real 21m31.907s, user 16m52.390s, sys 2m5.420s, then switch to "Release" and build not from scratch: real 7m41.194s, user 13m13.827s, sys 1m55.055s
GitLab runners:
============
test:garble: ~103 minutes https://gitlab.com/Molcas/OpenMolcas/-/jobs/3605102717
build:bound: ~82 minutes https://gitlab.com/hpqc-labs/OpenMolcas/-/jobs/3550933293 to ~93 minutes: https://gitlab.com/Molcas/OpenMolcas/-/jobs/3612638761
build:qcmaquis: ~42 minutes https://gitlab.com/hpqc-labs/OpenMolcas/-/jobs/3604051174
buid:intel: ~31 minutes https://gitlab.com/Molcas/OpenMolcas/-/jobs/3612638757
build: nag: ~28 minutes https://gitlab.com/Molcas/OpenMolcas/-/jobs/3612638759
build:garble: ~21 minutes https://gitlab.com/hpqc-labs/OpenMolcas/-/jobs/3604051171
build:linalg: ~20 minutes https://gitlab.com/hpqc-labs/OpenMolcas/-/jobs/3604051166
build:mpi: ~18 minutes https://gitlab.com/hpqc-labs/OpenMolcas/-/jobs/3604051175
build:plain: ~14 minutes https://gitlab.com/hpqc-labs/OpenMolcas/-/jobs/3604051164
doc: ~11 minutes https://gitlab.com/hpqc-labs/OpenMolcas/-/jobs/3604051161
On Cedar1, 2022 January 14 version of HPQC: time pymolcas verify. real 679m3.057s, user 357m46.428s, sys 94m1.227s, additional:896 Failed! (rasscf), grayzone:832 Failed! (scf), grayzone:834 Skipped!, grayzone:898 Failed! (caspt2). compiled with: cmake -D BIGOT=ON ../OpenMolcas_HPQC_2022_12_30 -D MAKE_BUILD_TYPE=DEBUG .... ooohh maybe that's why it was so slow !!!
Size:
============
[nike@cedar5 ~]$ time XZ_OPT=-5 tar cJf build_OpenMolcas_Add_Basis_Sets_2022_11_23.tar.xz build_OpenMolcas_Add_Basis_Sets_2022_11_23
real 15m51.207s, user 9m43.889s, sys 0m34.246s
-rw-r----- 1 nike nike 752M Jan 26 16:13 build_OpenMolcas_Add_Basis_Sets_2022_11_23.tar.xz
[nike@cedar1 ~]$ XZ_OPT=-e9 tar cJf build_OpenMolcas_Add_Basis_Sets_2022_11_21.tar.xz build_OpenMolcas_Add_Basis_Sets_2022_11_21
[nike@cedar1 ~]$ timed out waiting for input: auto-logout
-rw-r----- 1 nike nike 753810876 Jan 26 16:09 build_OpenMolcas_Add_Basis_Sets_2022_11_21.tar.xz
On Graham:
2.1G build_OpenMolcas_HPQC_2022_12_28_gfortran_garble
1.4G build_OpenMolcas_HPQC_Add_Basis_Sets
1.4G build_OpenMolcas_HPQC_2022_12_30
1.4G build_OpenMolcas_HPQC_2022_12_28_gfortran
1.4G build_OpenMolcas_HPQC_2022_11_02
1.4G build_OpenMolcas_HPQC_2022_10_31
1.4G build_OpenMolcas_2022_10_29
[nike@gra-login3 ~]$ time XZ_OPT=-3 tar cJf build_OpenMolcas_HPQC_2022_12_28_gfortran_garble.tar.xz build_OpenMolcas_HPQC_2022_12_28_gfortran_garble
real 6m12.603s, user 6m1.463s, sys 0m5.121s
-rw-r----- 1 nike nike 775M Feb 3 00:20 build_OpenMolcas_HPQC_2022_10_31.tar.xz
-rw-r----- 1 nike nike 774M Feb 3 00:20 build_OpenMolcas_HPQC_2022_11_02.tar.xz
-rw-r----- 1 nike nike 774M Feb 3 00:20 build_OpenMolcas_HPQC_2022_12_28_gfortran.tar.xz
-rw-r----- 1 nike nike 774M Feb 3 00:20 build_OpenMolcas_HPQC_2022_12_30.tar.xz
-rw-r----- 1 nike nike 774M Feb 2 23:50 build_OpenMolcas_HPQC_Add_Basis_Sets.tar.xz
-rw-r----- 1 nike nike 910M Feb 2 23:47 build_OpenMolcas_HPQC_2022_12_28_gfortran_garble.tar.xz