pivoting, Linpack performance, tunable SPEC results

[Earl Killian]

   Date: Thu, 20 Jun 91 06:07:49 PDT
   From: dghavalidgh.com (David G. Hough on validgh)

   I guess the moral is that on the 1000x1000 Linpack benchmark (where
   rewriting the source code is expected) it would be OK to rewrite it
   in quad and throw away the pivoting.  Or maybe just throw away the
   pivoting.

I once submitted a 64-bit non-pivoted result for 1000x1000 to Jack
Dongarra, the benchmark's author, and he said come back when you can
do pivoting.  So I don't think your last sentence will fly.

On the other hand, LAPACK shows that pivoting need not get in the way
of blocking.  Of course, I wouldn't expect a compiler to turn LINPACK
into LAPACK automatically.