This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[patch/gfortran] Optimizing matmul-transpose combinations





This patch follows the previous discussion of this topic.
(http://gcc.gnu.org/ml/fortran/2004-11/msg00104.html)

As Paul Brook proposed, the preferred
way to handle matmul-transpose combinations is to
change the implementation of TRANSPOSE so that it will
create a new array descriptor with swapped strides.

There are two parts to this change -- compiler part
and library part.  This patch contains the library
part, which is just dead code without the accompanying
change in compiler.

My experiments with a hacked version of compiler show
more than a 100% improvement on galgel spec benchmark.

      Victor
--
  Victor Leikehman
  IBM Research Labs in Haifa, Israel

2004-11-30  Victor Leikehman  <lei@il.ibm.com>

      * m4/matmul.m4: Use different loop order in the special
      case of transposed first argument.
      * generated/matmul_??.c: Regenerated.
(See attached file: matmul.m4.diff)

Attachment: matmul.m4.diff
Description: Binary data


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]