- Comp-phys-alps-users - lists.phys.ethz.ch

Re: [ALPS-users] BLAS/LAPACK troubles configuring ALPS 1.3.5
by Kirill Shtengel 30 Apr '10

30 Apr '10

Emanuel, Thanks a lot, but this does not seem to help. As I have mentioned in my first email, gfortran is not installed as a part of acml libraries on the cluster in question (unlike some other installations where there is a /usr/local/acml3.6.0/gfortran64 directory). On the machine I am trying to use, gfortran seems to be a part of the standard gcc installation: /usr/lib/gcc/x86_64-redhat-linux/4.1.2/libgfortran.so -> ../../../../lib64/libgfortran.so.1.0.0 Yet trying g++ -L /usr/local/acml3.6.0/gnu64/lib -L /usr/lib/gcc/x86_64-redhat- linux/4.1.2/ -lacml -lacml_mv -lgfortran test.cpp produces exactly the same output as before :-( What am I missing? Is the acml installation I am trying to use actually unusable for compiling ALPS with? Should I try installing acml libraries locally? Sorry is these questions seem dumb -- I am a novice in these matters. Best, Kirill On Thursday 29 April 2010 13:11:51 Emanuel Gull wrote: > Hi, > > this looks like you're forgetting to link gfortran. Please try > > g++ -L /usr/local/acml3.6.0/gnu64/lib -lacml -lacml_mv -lgfortran > > Emanuel > > On Apr 29, 2010, at 4:07 PM, Kirill Shtengel wrote: > > Matthias, > > > > Here is the best I seem to be able to do with your test program: > > > > [shtengel@gigio ~]$ g++ -L /usr/local/acml3.6.0/gnu64/lib -lacml > > -lacml_mv test.cpp > > /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference to > > `e_wsfi' /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference > > to `s_stop' /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined > > reference to `s_wsfe' /usr/local/acml3.6.0/gnu64/lib/libacml.so: > > undefined reference to `e_rsfi' > > /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference to `s_cmp' > > /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference to > > `e_wsfe' /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference > > to `z_abs' /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference > > to `do_lio' /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined > > reference to `s_wsle' /usr/local/acml3.6.0/gnu64/lib/libacml.so: > > undefined reference to `z_sqrt' > > /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference to `s_cat' > > /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference to > > `c_sqrt' /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference > > to `s_copy' /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined > > reference to `s_rsfi' /usr/local/acml3.6.0/gnu64/lib/libacml.so: > > undefined reference to `do_fio' > > /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference to `z_exp' > > /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference to `c_abs' > > /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference to > > `e_wsle' /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference > > to `c_exp' /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined reference > > to `s_wsfi' /usr/local/acml3.6.0/gnu64/lib/libacml.so: undefined > > reference to `i_indx' collect2: ld returned 1 exit status > > > > Could someone please help me with these errors? What are the libraries I > > am still missing? > > The other installation of acml/blas/lapack libraries I have mentioned in > > my previous email does not work at all: the linker reports skipping these > > libraries as incompatible (?). > > Once again, any help will be greatly appreciated! > > > > Kirill > > > > On Monday 26 April 2010 21:45:23 Matthias Troyer wrote: > >> On 27 Apr 2010, at 04:19, Kirill Shtengel wrote: > >>> Hello, > >>> > >>> I wonder if anyone can help me with configuring/compiling ALPS-1.3.5 on > >>> a Linux machine with AMD Opteron Processor. My problem seems to be > >>> making the configuration routine see either acml or blas/lapack > >>> libraries. From what I can tell, there are two separate installations > >>> of > >>> acml/blas/lapack libraries, one is the set of ACML libraries: > >> > >> ... > >> > >>> Removing references to lapack (i.e. keeping only references to blas in > >>> the configuration options makes no difference whatsoever). In either > >>> case, I am getting " LAPACK support no" at the end of the > >>> configuration process, and subsequent use of make does not compile dmrg > >>> (which is what I intend to run). > >>> Any help or suggestions will be greatly appreciated! > >> > >> Hi Kirill > >> > >> The best way to figure out how to link LAPACL/BLAS is to write a test > >> program: > >> > >> extern "C" {void dgeev_();} > >> int main() > >> { > >> dgeev_(); > >> } > >> > >> and try to compile this with the -L and -l options you mention above. > >> You'll immediately see the linker errors, which might give you hints > >> regarding libraries you need to link in addition. We can help you more > >> if you tell us what happens when you try this. > >> > >> Matthias > -- Kirill Shtengel Associate Professor Department of Physics & Astronomy UC Riverside, CA 92521 Tel: (951) 827-1058

4 4

BLAS/LAPACK troubles configuring ALPS 1.3.5
by Kirill Shtengel 29 Apr '10

29 Apr '10

Hello, I wonder if anyone can help me with configuring/compiling ALPS-1.3.5 on a Linux machine with AMD Opteron Processor. My problem seems to be making the configuration routine see either acml or blas/lapack libraries. From what I can tell, there are two separate installations of acml/blas/lapack libraries, one is the set of ACML libraries: [shtengel@gigio ~]$ ls /usr/local/acml3.6.0/ Doc gnu64 ReleaseNotes ReleaseNotes.acml_mv util [shtengel@gigio ~]$ ls /usr/local/acml3.6.0/gnu64/lib libacml.a libacml_mv.a libacml_mv.so libacml.so [shtengel@gigio ~]$ ls /usr/local/acml3.6.0/gnu64/include/ acml.h acml_mv.h acml_mv_m128.h Notice is that there is no gfortran64 subdirectory in /usr/local/acml3.6.0/ (I can't tell if it's really needed, but someone suggested it may be.) The other set of libraries is in a PGI installation and it includes ACML, BLAS and LAPACK libraries separately: ls /usr/local/pgi-7.2-3/linux86-64/7.2-3/lib 90alt.o libjpgdbg.so libpgf90_rpm1.a libpgnod_prof.a f90main.o libjpgprof.so libpgf90_rpm1.ipl libpgnod_prof_g.a hpfalt.o liblapack.a libpgf90_rpm1_p.a libpgnod_prof_hpmpi.a hpfmain.o liblapack.ipl libpgf90rtl.a libpgnod_prof_mpi2.a hpmpif.o libmem.il libpgf90rtl.ipl libpgnod_prof_mpi.a hugebss.ld libm.ipl libpgftnrtl.a libpgnod_prof_mvapich.a initmp.o libmpich.ipl libpgftnrtl.ipl libpgnod_prof_papi.a init_pgpf.o libnspgc.a libpghpf2.a libpgnod_prof_pfo.a libacml.a libnuma.so libpghpf.a libpgnod_prof_time.a libacml.ipl libnuma.so.1 libpghpf_mpi.a libpthread.ipl libacml_mp.a libpgbind.a libpghpf_mpi_p.a libstd.a libacml_mv.a libpgbind.so libpghpf_prof.a libstdz.a libblas.a libpgc.a libpghpf_rpm1.a libzceh.a libblas.ipl libpgc.ipl libpghpf_rpm1_p.a nonuma.o libC.a libpgf902.a libpghpf_rpm.a pgfalt.o libc.ipl libpgf902.ipl libpghpf_rpm_p.a pgfi8st.o libCz.a libpgf90.a libpghpf_smp.a pgfmain.o libfmpich.ipl libpgf90.ipl libpghpf_smp_p.a pgi.ld libhugetlbfs_pgi.a libpgf90_prof.a libpgmp.a subchk.o libintrinsics.il libpgf90_prof.ipl libpgmp.ipl trace_init.o [shtengel@gigio ~]$ ls /usr/local/pgi-7.2-3/linux86-64/7.2-3/include/ acml.h byteswap.h iso_fortran_env.mod omp_lib_kinds.mod sys acml_mv.h CC lib3f.h omp_lib.mod tmmintrin.h acml_mv_m128.h curses.h limits.h pglocal.f va_list.h alloca.h emmintrin.h math.h pglocal.h varargs.h ammintrin.h error.h mmintrin.h pmmintrin.h xmmintrin.h ansidecl.h float.h ncurses.h stdarg.h argp.h intrin.h omp.h stddef.h bits iso_c_binding.mod omp_lib.h stdlib.h (Notice the absence of blas.h and lapac.h header files, acml header files are in place). Anyway, any attempts of making the ./configure routine see either those libraries fails, as far as I can tell: [shtengel@gigio ~/alps-1.3.5]$ ./configure --with-blas-dir=/usr/local/acml3.6.0/gnu64/lib --with-blas=-lacml --with-lapack-dir=/usr/local/acml3.6.0/gnu64/lib --with-lapack=-lacml --with-boost=/home/shtengel/boost_1_42_0 ...... configure: checking for BLAS Library checking for BLAS library directory... /usr/local/acml3.6.0/gnu64/lib checking for dgemm_ in -L/usr/local/acml3.6.0/gnu64/lib -lacml... no checking for dgemm_ in -L/usr/local/acml3.6.0/gnu64/lib -lsci... no checking for dgemm_ in -L/usr/local/acml3.6.0/gnu64/lib -lacml -lacml_mv - lgfortran... no checking for dgemm_ in -L/usr/local/acml3.6.0/gnu64/lib -lblas... no checking for dgemm_ in -L/usr/local/acml3.6.0/gnu64/lib -lblas -lf2c... no checking for dgemm_ in -L/usr/local/acml3.6.0/gnu64/lib -lblas -lg2c... no checking for dgemm_ in -L/usr/local/acml3.6.0/gnu64/lib -lblas -lpgftnrtl - lpgc... no checking for dgemm_ in -L/usr/local/acml3.6.0/gnu64/lib -lblas /usr/lib/libg2c.so.0.0.0... no configure: disabling BLAS support configure: disabling LAPACK support ...... [shtengel@gigio ~/alps-1.3.5]$ ./configure --with-blas-dir=/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib --with-blas=-lacml --with-lapack-dir=/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib --with-lapack=- lacml --with-boost=/home/shtengel/boost_1_42_0 ..... configure: checking for BLAS Library checking for BLAS library directory... /usr/local/pgi-7.2-3/linux86-64/7.2-3/lib checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lacml... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lsci... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lacml - lacml_mv -lgfortran... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lblas... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lblas - lf2c... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lblas - lg2c... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lblas - lpgftnrtl -lpgc... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lblas /usr/lib/libg2c.so.0.0.0... no configure: disabling BLAS support configure: disabling LAPACK support ..... And finally: [shtengel@gigio ~/alps-1.3.5]$ ./configure --with-blas-dir=/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib --with-blas=-lblas --with-lapack-dir=/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib --with-lapack=-llapack --with-boost=/home/shtengel/boost_1_42_0 ..... configure: checking for BLAS Library checking for BLAS library directory... /usr/local/pgi-7.2-3/linux86-64/7.2-3/lib checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lblas... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lsci... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lacml - lacml_mv -lgfortran... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lblas... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lblas - lf2c... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lblas - lg2c... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lblas - lpgftnrtl -lpgc... no checking for dgemm_ in -L/usr/local/pgi-7.2-3/linux86-64/7.2-3/lib -lblas /usr/lib/libg2c.so.0.0.0... no configure: disabling BLAS support configure: disabling LAPACK support ..... Removing references to lapack (i.e. keeping only references to blas in the configuration options makes no difference whatsoever). In either case, I am getting " LAPACK support no" at the end of the configuration process, and subsequent use of make does not compile dmrg (which is what I intend to run). Any help or suggestions will be greatly appreciated! Kirill -- Kirill Shtengel Associate Professor Department of Physics & Astronomy UC Riverside, CA 92521 Tel: (951) 827-1058

3 3

Announcement: CECAM/ESF/MaNEP summer school on simulations of strongly correlated systems
by Matthias Troyer 25 Apr '10

25 Apr '10

Dear colleagues, I want to draw your attention to a summer school / tutorial on the simulation of strongly correlated quantum systems to be held September 13-17, 2010 at ETH Zurich. The school will consist of lectures on methods in the morning, followed by an experimental talk in the afternoon and hands-on tutorials with the forthcoming release 2.0 of the ALPS simulation package. Confirmed speakers include A. Laeuchli (MPI Dresden) U. Schollwoeck (LMU Munich) S. Trebst (Microsoft) M. Troyer (ETH) P. Werner (ETH) We have secured funding for local expenses (housing and meals) of all admitted participants from CECAM, Intelbiomat and the Swiss NCCR MaNEP. For details and to register please see the web page http://www.cecam.org/workshop-473.html Please forward this announcement to any interested PhD student and postdoc. Best regards Matthias Troyer

1 0

Lanczos
by Sebastian Huber 22 Apr '10

22 Apr '10

I made use of the sparsediag 1.1 application part of alps-applications 1.3.4. The maximal interations for the eigenvalues MAX_ITERATIONS can be set in the parameter file. For the calculation of the eigenvectors (correlations) this parameter is not adjustable outside the source code (set to 10). I had to fix this for my purposes. I am sure the way I implemented it is not up to the standards, but it maybe interesting to note neverthless. Sebastian -- Sebastian Huber Department of Condensed Matter Physics The Weizmann Institute of Science 76100 Rehovot Web: http://www.weizmann.ac.il/~huber Tel: +972 8 934 4258

2 1

ALPS user OS survey
by Matthias Troyer 15 Apr '10

15 Apr '10

Dear ALPS users, To prepare for version 2.0 of ALPS we want to conduct a survey of operating systems used by ALPS users. ALPS 2.0 will support all OSs supported by ALPS 1.3 and in addition also Windows and the Cray XT 5. Please follow this link: http://doodle.ethz.ch/hhbvg8ixvpi8kuaa and fill in the short survey. Thanks in advance Matthias Troyer

1 1

Winding numbers
by "Michał Maik" 13 Apr '10

13 Apr '10

Hi, I am having a problem calculating the "Winding Number vs Energy" graph for certain models. When I do it for the "4-site mixed" I am not able to get it to work but if I use a "square lattice" with L=2 it seems to work fine. I am wondering how I have to define a model in order to calculate the winding numbers. Regards Michal Maik

2 1

Call for Papers: International Workshop on "Statistical Physics of Quantum Systems" (Tokyo, August 2-4 2010)
by Synge Todo 04 Mar '10

04 Mar '10

International Workshop on "Statistical Physics of Quantum Systems" First Announcement and Call for Papers We cordially invite you to attend and/or present a paper at the International Workshop on Statistical Physics of Quantum Systems held during 2-4 August 2010 at Institute of Industrial Science, University of Tokyo. http://www.iis.u-tokyo.ac.jp/access_e/access_e.html (In the convention hall of the An block of the campus.) http://www.iis.u-tokyo.ac.jp/access_e/campusmap_e.html This is a satellite meeting of the international conference STATPHYS 24 http://www.statphys.org.au/ In this workshop, we will discuss on the recent developments of studies on various novel properties of strongly interacting quantum systems, non-equilibrium quantum dynamics under time-depending fields, and quantum response as well as other related topics. Specific topics include, but not restricted to a.. Phase transitions in quantum systems b.. Novel quantum properties in the ground state c.. Quantum dynamics d.. Quantum transport phenomena e.. Quantum information processing We ask you to register and/or submit a paper at the following web site: http://looper.t.u-tokyo.ac.jp/spqs2010/ The registration will be open by 8 March, 2010. The deadline of registration and/or submission is 30 April, 2010. We will choose poster presentations as well as a small number of oral presentations from the submitted papers and tell the authors the decision by 1 June, 2010. The registration fee is 20,000 yen (JPY20,000) for regular participants and 10,000 yen (JPY10,000) for students. It includes the attendance as well as the abstract book/CD. Please pay the fee in cash at the on-site registration desk when you arrive at the workshop. (Sorry, we do not accept credit cards.) We will have a party on the night of 3 August. The fee for the party is 5,000 yen for regular participants and 2,000 yen for students. We will collect the fee at the workshop site. Confirmed invited speakers include M. Batchelor (Australian National University, http://wwwmaths.anu.edu.au/~murrayb/Murray_Batchelor2.html) O. Cepas (Grenoble, CNRS, http://perso.neel.cnrs.fr/olivier.cepas/) G. Ordonez (Butler University, http://www.butler.edu/physics/faculty-staff) T. Petrosky (University of Texas, http://order.ph.utexas.edu/people/Petrosky.htm) H.A. De Raedt (University of Groningen, http://www.rug.nl/staff/h.a.de.raedt/index) M. Troyer (ETH, http://www.itp.phys.ethz.ch/people/troyer) G. Vidal (University of Queensland, http://www.physics.uq.edu.au/people/vidal/) T. Xiang (Chinese Academy of Sciences, http://power.itp.ac.cn/~txiang/) Local Organizers: S. Miyashita, N. Hatano, M. Imada, N. Kawakami, N. Kawashima, A. Kuniba, H. Nishimori, M. Oshikawa, T. Sakai, and S. Todo

1 0

sigma_y interaction
by Fernando Cucchietti 28 Feb '10

28 Feb '10

Hello, I probably know the answer to this already, but just wanted to check: I would like to use the Montecarlo code to simulate a system with a Hamiltonian that has the Sy operator, that is Sy=-I/2(Splus-Sminus). The worm and sse codes return a runtime error "can not convert complex number into real one". If I don't include de I constant (or (0.,1.) ) in the model.xml file the code starts, but then complains that it cannot simulate this term (which without the complex unit is non-Hermitian). I cannot rotate my basis because I also have the Sx operator, so I am stuck with Sy. Is there a way to do this? It looks like a no-no... Thanks Fernando -- ============================== Fernando Cucchietti Research Fellow ICFO-Institut de Ciencies Fotoniques Mediterranean Technology Park Av. Canal Olimpic s/n 08860 Castelldefels (Barcelona), Spain ============================== 1+1=3, for large values of 1 ------------------------------

2 1

Exact Diagonalization for Bose-Hubbard
by Vipin Varma 19 Feb '10

19 Feb '10

Hi I am trying the Lanczos method for solving the Bose-Hubbard model on a square lattice, to get the t/U dependence of the ground state energy per site. I tried a simple 2x1 cluster with open and periodic boundary conditions and I get a linear dependence for the energy on t/U; analytically (which I am reasonably sure is correct), I get a parabolic dependence for the energy on t/U, if I consider 2 bosons over 2 sites. Here is my parameter file: /////////////////////////// LATTICE_LIBRARY="../lattices.xml"; LATTICE="square lattice"; MODEL_LIBRARY="../models.xml"; MODEL="boson Hubbard"; U = 1; Nmax = 2; L=2; W =1; {t=0.1;} {t=0.01;} {t=0.02;} {t=0.03;} {t=0.04;} {t=0.05;} {t=0.06;} {t=0.07;} {t=0.08;} {t=0.09;} /////////////////// I have tried Nmax = 2,4,6, but all give me a linear dependence; also strangely, the energy range does not change much when I vary the value of U by orders of magnitude. I guess I am missing something simple. Any ideas will be appreciated. Vipin

2 5

Python distributions
by Matthias Troyer 05 Feb '10

05 Feb '10

Dear ALPS developers and users, Since ALPS 2.0, to be released later this year in a beta version, will rely heavily on Python for data analysis I want to start a survey about the Python distributions you are using. We might be able to prepare compiled binary installation packages for the most common combinations of OS/compiler/Python distribution. Please reply to the list with the Python distribution you are using if you feel strongly about that distribution you are using now and don't want to change it when using ALPS 2.0 in the future. Matthias PS: you will always be able to compile ALPS from source against any distribution, but many people prefer the new binary installations that can be installed in less than a minute.

6 15