blas test failures
Greetings! There appears to be nothing I can do to get blas to pass
its integrity tests on sparc (vore). Tried 2.95 and 3.0, turning off
all optimizations, -Wall, even an ultrasparc ISA. All fail in exactly
the same place. Mips had a similar problem in this place until I used
gcc-3.0 there instead of the default. I know its not an error in the
code, as it passes on all other arches (except hppa for other
explainable reasons), and is a *very* well established standard
upstream.
Anyway, the failures are in the single precision real and complex C
interface tests to dot, asum, and nrm2. Here is a sample error file:
Could go through with gdb, but don't have time right now.
=============================================================================
Real CBLAS Test Program Results
Test of subprogram number 1 CBLAS_SDOT
FAIL
CASE N INCX INCY MODE I COMP(I) TRUE(I) DIFFERENCE SIZE(I)
1 1 1 1 9999 1 0.38146987E-06 0.30000001E+00 -0.3000E+00 0.3000E+00
1 2 1 1 9999 1 0.21457691E-07 0.20999999E+00 -0.2100E+00 0.1600E+01
1 4 1 1 9999 1 0.11718752E-03 0.62000000E+00 -0.6199E+00 0.3200E+01
1 1 2 -2 9999 1 0.38146987E-06 0.30000001E+00 -0.3000E+00 0.3000E+00
1 2 2 -2 9999 1 -0.35652176E-11 -0.70000000E-01 0.7000E-01 0.1600E+01
1 4 2 -2 9999 1 0.15625017E-02 0.85000002E+00 -0.8484E+00 0.3200E+01
1 1 -2 1 9999 1 0.38146987E-06 0.30000001E+00 -0.3000E+00 0.3000E+00
1 2 -2 1 9999 1 -0.80078148E-03 -0.79000002E+00 0.7892E+00 0.1600E+01
1 4 -2 1 9999 1 -0.44921864E-03 -0.74000001E+00 0.7396E+00 0.3200E+01
1 1 -1 -2 9999 1 0.38146987E-06 0.30000001E+00 -0.3000E+00 0.3000E+00
1 2 -1 -2 9999 1 0.74386634E-06 0.33000001E+00 -0.3300E+00 0.1600E+01
1 4 -1 -2 9999 1 0.36250010E-01 0.12700000E+01 -0.1234E+01 0.3200E+01
Test of subprogram number 2 CBLAS_SAXPY
----- PASS -----
Test of subprogram number 3 CBLAS_SROTG
----- PASS -----
Test of subprogram number 4 CBLAS_SROT
----- PASS -----
Test of subprogram number 5 CBLAS_SCOPY
----- PASS -----
Test of subprogram number 6 CBLAS_SSWAP
----- PASS -----
Test of subprogram number 7 CBLAS_SNRM2
FAIL
CASE N INCX INCY MODE I COMP(I) TRUE(I) DIFFERENCE SIZE(I)
7 1 1 9999 9999 1 0.38146987E-06 0.30000001E+00 -0.3000E+00 0.3000E+00
7 2 1 9999 9999 1 0.30517578E-04 0.50000000E+00 -0.5000E+00 0.5000E+00
7 3 1 9999 9999 1 0.29296879E-03 0.69999999E+00 -0.6997E+00 0.7000E+00
7 4 1 9999 9999 1 0.97656222E-04 0.60000002E+00 -0.5999E+00 0.6000E+00
7 1 2 9999 9999 1 0.38146987E-06 0.30000001E+00 -0.3000E+00 0.3000E+00
7 2 2 9999 9999 1 0.30517578E-04 0.50000000E+00 -0.5000E+00 0.5000E+00
7 3 2 9999 9999 1 0.29296879E-03 0.69999999E+00 -0.6997E+00 0.7000E+00
7 4 2 9999 9999 1 0.97656222E-04 0.60000002E+00 -0.5999E+00 0.6000E+00
Test of subprogram number 8 CBLAS_SASUM
FAIL
CASE N INCX INCY MODE I COMP(I) TRUE(I) DIFFERENCE SIZE(I)
8 1 1 9999 9999 1 0.38146987E-06 0.30000001E+00 -0.3000E+00 0.3000E+00
8 2 1 9999 9999 1 0.29296905E-03 0.69999999E+00 -0.6997E+00 0.7000E+00
8 3 1 9999 9999 1 0.14062506E-01 0.11000000E+01 -0.1086E+01 0.1100E+01
8 4 1 9999 9999 1 0.78125000E-02 0.10000000E+01 -0.9922E+00 0.1000E+01
8 1 2 9999 9999 1 0.38146987E-06 0.30000001E+00 -0.3000E+00 0.3000E+00
8 2 2 9999 9999 1 0.29296905E-03 0.69999999E+00 -0.6997E+00 0.7000E+00
8 3 2 9999 9999 1 0.14062506E-01 0.11000000E+01 -0.1086E+01 0.1100E+01
8 4 2 9999 9999 1 0.78125000E-02 0.10000000E+01 -0.9922E+00 0.1000E+01
Test of subprogram number 9 CBLAS_SSCAL
----- PASS -----
Test of subprogram number 10 CBLAS_ISAMAX
----- PASS -----
=============================================================================
Any advice most appreciated!
--
Camm Maguire camm@enhanced.com
==========================================================================
"The earth is but one country, and mankind its citizens." -- Baha'u'llah
Reply to: