About Installing ATLAS with Athlon 64 X2 for Windows users

1.1 Requirement: cygwin (gcc, g++, g77, make, gdb )
    (#Use cygwin.bat and don't use RXVT,VT102 terminal
    emulator or you can't go after 1.5.6.)
1.2 Access to http://www.netlib.org/atlas/
    and obtain atlas3.6.0.gz.
1.3 After downloading atlas3.6.0.gz in /cygwin/usr/local/,
    unzip atlas3.6.0.gz, when the command line is
    gunzip -c atlas3.6.0.gz | tar xv or
    tar xvfz atlas3.6.0.gz.
1.4 Since current directory is /cygwin/usr/local/,
    move it to /cygwin/usr/local/ATLAS/, when
    the command line is cd ATLAS.
1.5 Build xconfig.exe, when the command line is make
    and answer the question below appropriately.
1.5.1 160
        159
        158
         ...
         3
         2
         1
        Enter number at top left of screen [0]: 160
1.5.2 Have you scoped the errata file? [y]: y
1.5.3 Are you ready to continue? [y]: y
1.5.4 Enter your machine type:
         1. Other/Unknown
         2. AMD Athlon
         3. 32 bit AMD Hammer
         4. 64 bit AMD Hammer
         5. Pentium PRO
         6. Pentium Ⅱ
         7. Pentium Ⅲ
         8. Pentium 4
         Enter machine number [1]: 2
1.5.5 enable Posix threads support? [n]: y
1.5.6 Enter the number processors in system [0]: 2
1.5.7 use express setup? [y]: y
1.5.8 Enter Architecture name (ARCH) [WinNT_ATHLONSSE2_2]:
      WinNT_ATHLONSSE2_2
1.5.9 Enter Maximum cache size (KB) [4096]: 4096
1.5.10 Enter File creation delay in seconds [0]: 0
1.5.11 Tune the Level 1 BLAS? [y]: y
1.6 Build ATLAS, when the command line is
    make install arch=WinNT_ATHLONSSE2_2.
    This work will take more than 2.5 hours to complete.
1.7 Copy the library files, when the command lines are
    given below.
    cd ./lib/WinNT_ATHLONSSE2_2,
    cp *.a /lib,
    ranlib /lib/liblapack.a, ranlib /lib/libatlas.a,
    ranlib /lib/libcblas.a, ranlib /lib/libf77blas.a,
    ranlib /lib/libptf77blas.a, ranlib /lib/libptcblas.a,
    ranlib /lib/libtstatlas.a
1.8 Now you can use ATLAS, when the command lines are
    given below.
    g77 -o file01 file01.f -llapack -lf77blas -lcblas
     -latlas -lg2c -lm or
    g95 -o file01 file01.f90 -llapack -lf77blas -lcblas
     -latlas -lg2c -lm
    file01 is an example and you can use other
    fortran 77 or 95 programs.

2.1 About the LINPACK benchmark

    Access to http://www.netlib.org/benchmark/
    and you can obtain 1000s, 1000d, linpacks,
    and linpackd, which are benchmark programs.
    The data which has the largest mflops in the
    several trials is adopted.

2.1.1 The results from LINPACK 1000s benchmark

    $ g77 -o 1000s 1000s.f

norm resid	resid	machep
9.56832123E+00	5.70633099E-04	1.19209290E-07
X(1)	X(n)
1.00003088E+00	9.99999046E-01
factor	solve	total
2.418E+00	0.000E+00	2.418E+00
mflops	unit	ratio
2.765E+02	7.232E-03	4.318E+01

2.1.2 The results from LINPACK 1000d benchmark

    $ g77 -o 1000d 1000d.f

norm resid	resid	machep
1.05174252E+01	1.16766853E-12	2.22044605E-16
X(1)	X(n)
1.00000000E+00	1.00000000E+00
factor	solve	total
2.995E+00	0.000E+00	2.995E+00
mflops	unit	ratio
2.233E+02	8.958E-03	5.348E+01

2.2 About the DGEMM benchmark

    Access to http://www.mcs.anl.gov/index.php,
    go to Software>MPICH>Win IA32 Binary (1.2.1p1),
    and obtain mpich2-1.2.1p1-win-ia32.msi. After
    downloading mpich2-1.2.1p1-win-ia32.msi, we
    install it on Windows system. Then we add C:\
    Program Files\MPICH2\bin to path in environment
    variables, copy libfmpich2g.a, libmpi.a, and
    libmpicxx.a into /lib, and enter the command lines
    given below.

    $ ranlib /lib/libfmpich2g.a
    $ ranlib /lib/libmpi.a
    $ ranlib /lib/libmpicxx.a

    Then access to https://computecanada.org/
    and go to Committees>TECC>Working groups>
    Benchmarking>Benchmark Collection>
    Microbenchmarks>DGEMM, and you can obtain
    dgemm-1.0.0.tar.gz.

    $ tar xvfz dgemm-1.0.0.tar.gz
    $ cd dgemm-1.0.0
    $ cp ./setup/Make.Linux_AtlasFBLAS_Lam ./

    We edit Make.Linux_AtlasFBLAS_Lam to solve
    the problem about Message Passing library (MPI).

    $ vi Make.Linux_AtlasFBLAS_Lam
      MPlib = -lfmpich2g -lmpi -lmpicxx
      # When the value of MPlib was null, it also
      # went well.

    We build dgemm-1.0.0, when the command line is
    given below.

    $ make arch=Linux_AtlasFBLAS_Lam

    Now we can use mpiexec.exe and hpcc-dgemm.exe.
    Then we will confirm the performance differences
    between ATLAS+BLAS and BLAS which is included
    in LAPACK using the benchmark software.

    $ mpiexec -np 1 hpcc-dgemm 100

    Figure1 shows the processing speed of ATLAS+BLAS
    is approximately 5.6 times faster than the speed of
    BLAS mentioned above regarding the performance
    measured by Single DGEMM Gflop/s and the performance
    differeneces related to the calculation accuracy is
    not mentioned.

Figure 1  The DGEMM Performance

Finally, I am happy to assist you in installing ATLAS and
the DGEMM benchmark and evaluating the performance
differences of some software libraries.