==============================================================
High Performance Computing Linpack Benchmark (HPL)
HPL - 2.1 - October 26, 2012
==============================================================
History
- 09/09/00 Public release of Version 1.0
- 09/27/00 A couple of mistakes in the VSIPL port have been
corrected. The tar file as well as the web site were updated
on September 27th, 2000. Note that these problems were not
affecting the BLAS version of the software in any way.
- 01/01/04 Version 1.0a
The MPI process grid numbering scheme is now an run-time
option.
The inlined assembly timer routine that caused the compila-
tion to fail when using gcc version 3.3 and above has been
removed from the package.
Various building problems on the T3E have been fixed; Thanks
to Edward Anderson.
- 15/12/04 Version 1.0b
Weakness of the pseudo-random matrix generator found for pro-
blem sizes being power of twos and larger than 2^15; Thanks
to Gregory Bauer. This problem has not been fixed. It is thus
currently recommended to HPL users willing to test matrices
of size larger than 2^15 to not use power twos.
When the matrix size is such that one needs > 16 GB per MPI
rank, the intermediate calculation (mat.ld+1) * mat.nq in
HPL_pdtest.c ends up overflowing because it is done using
32-bit arithmetic. This issue has been fixed by typecasting
to size_t; Thanks to John Baron.
- 09/10/08 Version 2.0
Piotr Luszczek changed to 64-bit RNG, modified files:
-- [M] include/hpl_matgen.h
-- [M] testing/matgen/HPL_ladd.c
-- [M] testing/matgen/HPL_lmul.c
-- [M] testing/matgen/HPL_rand.c
-- [M] testing/ptest/HPL_pdinfo.c
For a motivation for the change, see:
Dongarra and Langou, ``The Problem with the Linpack
Benchmark Matrix Generator'', LAWN 206, June 2008.
-- [M] testing/ptest/HPL_pdtest.c --
Julien Langou changed the test for correctness from
||Ax-b||_oo / ( eps * ||A||_1 * N )
||Ax-b||_oo / ( eps * ||A||_1 * ||x||_1 )
||Ax-b||_oo / ( eps * ||A||_oo * ||x||_oo * N )
to the normwise backward error
|| r ||_oo / ( eps * ( || x ||_oo * || A ||_oo + || b ||_oo ) * N )
See:
Nicholas J. Higham, ``Accuracy and Stability of Numerical Algorithms'',
Society for Industrial and Applied Mathematics, Philadelphia, PA, USA,
Second Edition, pages = xxx+680, ISBN = 0-89871-521-0, 2002.
Note that in our case || b ||_oo is almost for sure
1/2, we compute it anyway.
- 10/26/2012 Version 2.1
Piotr Luszczek introduced exact time stamping for HPL_pdgesv():
-- [M] dist/include/hpl_misc.h
-- [M] dist/testing/ptest/HPL_pdtest.c
Piotr Luszczek fixed out-of-bounds access in data spreading functions. exact time stamping for HPL_pdgesv():
-- [M] dist/src/pgesv/HPL_spreadN.c
-- [M] dist/src/pgesv/HPL_spreadT.c
Thanks to Stephen Whalen from Cray.
==============================================================