installing lapack 3.3.1 - a BLAS test fails.

[plwm2_cam]

Hi,

I'm quite a newbie to the whole lapack/blas thing, but i've just tried to install lapack 3.3.1 using gcc 4.4.3 on a linux machine (with Fedora 12 installed). I set it so that I link to a previous GotoBLAS2 installation. However, absolutely everything is fine until one of the BLAS test executables ( xblat1d ) is performed - this produces the following output:

Real BLAS Test Program Results


Test of subprogram number 1 DDOT
----- PASS -----

Test of subprogram number 2 DAXPY
----- PASS -----

Test of subprogram number 3 DROTG
----- PASS -----

Test of subprogram number 4 DROT
----- PASS -----

Test of subprogram number 5 DCOPY
----- PASS -----

Test of subprogram number 6 DSWAP
----- PASS -----

Test of subprogram number 7 DNRM2
----- PASS -----

Test of subprogram number 8 DASUM
----- PASS -----

Test of subprogram number 9 DSCAL
----- PASS -----

Test of subprogram number 10 IDAMAX
----- PASS -----

Test of subprogram number 11 DROTMG
FAIL

CASE N INCX INCY I COMP(I) TRUE(I) DIFFERENCE SIZE(I)

11 1 9999 9999 1 0.23076923D-01 0.92307692D-01 -0.6923D-01 0.9231D-01
11 1 9999 9999 2 0.12905551D+06 0.27692308D+00 0.1291D+06 0.2769D+00
11 1 9999 9999 3 0.26000000D+01 0.13000000D+01 0.1300D+01 0.1300D+01
11 1 9999 9999 5 -0.10000000D+01 0.00000000D+00 -0.1000D+01 0.0000D+00
11 1 9999 9999 6 0.20000000D+01 0.00000000D+00 0.2000D+01 0.0000D+00
11 1 9999 9999 7 -0.24414062D-03 -0.16666667D+00 0.1664D+00 -0.1667D+00
11 1 9999 9999 8 0.10000000D+01 0.50000000D+00 0.5000D+00 0.5000D+00
11 1 9999 9999 9 0.14648437D-02 0.00000000D+00 0.1465D-02 0.0000D+00
11 2 9999 9999 2 0.10961114D+08 0.65333333D+00 0.1096D+08 0.6533D+00
11 2 9999 9999 5 -0.10000000D+01 0.10000000D+01 -0.2000D+01 0.1000D+01
11 2 9999 9999 7 -0.24414062D-03 0.00000000D+00 -0.2441D-03 0.0000D+00
11 2 9999 9999 8 0.10000000D+01 0.00000000D+00 0.1000D+01 0.0000D+00
11 2 9999 9999 9 0.34877232D-04 0.14285714D+00 -0.1428D+00 0.1429D+00
11 5 9999 9999 2 0.25165824D+06 0.15000000D-01 0.2517D+06 0.1500D-01
11 5 9999 9999 7 -0.24414063D-07 -0.10000000D-03 0.9998D-04 -0.1000D-03
11 5 9999 9999 9 0.24414062D-03 0.10000000D+01 -0.9998D+00 0.1000D+01
11 6 9999 9999 2 0.22369621D+06 0.13333333D-01 0.2237D+06 0.1333D-01
11 6 9999 9999 7 -0.24414062D+03 -0.10000000D+07 0.9998D+06 -0.1000D+07
11 6 9999 9999 9 0.24414062D-03 0.10000000D+01 -0.9998D+00 0.1000D+01
11 7 9999 9999 2 0.37529997D+05 0.22369621D-02 0.3753D+05 0.2237D-02
11 7 9999 9999 7 -0.59604645D-07 -0.24414062D-03 0.2441D-03 -0.2441D-03
11 7 9999 9999 9 0.59604645D-03 0.24414062D+01 -0.2441D+01 0.2441D+01

then the whole installation just hangs.

Is this a bug, or have i done something stupid? :)
I've done the same thing on another linux machine with gcc 4.3.2 and basically the same thing happens.
I would suspect that there's something wrong with my GotoBLAS2 installation - but as a standalone thing (the GotoBLAS2 seems to have automatically installed lapack 3.1.1 so that the blas library contains lapack 3.1.1 routines too...) the BLAS seems to work. Perhaps this is the problem?

Thanks in advance!

Peter

[admin]

Hi,
Actually I believe your GotoBLAS2 installation is fine.
We added back those DROTMG Testing into the Reference BLAS in the last release.
Those are for the Reference BLAS, even they should theoretically works for any BLAS, it might not be the case.
The only important thing is that your LAPACK testings are successful.
Julie

[plwm2_cam]

Thanks for the quick reply!
So are you suggesting that there all the lapack installation stuff is done first, then the lapack tests are built and executed, and then the BLAS tests are built then executed? In other words, by the time i've got to the problem i've had with DROTMG - have all the lapack tests been done and that i've nothing more to worry about?

Thanks :)

[admin]

Yes, exactly!
If you look in the makefile you have:

Code: Select all

all: lapack_install lib lapack_testing blas_testing


But I will take a closer look, because usually we had some minor problems with the BLAS testings with some vendors libraries but it was just basic failure. This is really bad that the program hangs.