intelrdfpmath FTBFS on Alpha due to missing architecture support. It
is clear that Alpha was once supported but the header files for the
Alpha suppport are missing in the source file.  I could not find the
missing header files in any web archives.

The special header files were required because floating-point on
Alpha is not fully IEEE compliant without kernel support, however
the Debian gcc compiler on Alpha defaults to -mieee which adds in
all support to get IEEE compliant behaviour, so we don't actually
need the missing header files.

I therefore attach a patch that comments out the Alpha specific
header file selection stuff and adds Alpha architecture support in
the same vein as other arches have done.

With that intelrdfpmath builds to completion and the test suite
reports 3756 failed results, but that is the same number as reported
in the build logs for some of the release architectures.

Regards,
Michael.