robin wrote:
>Any arithmetic operation that uses
>registers to carry out arithmetic renders the
>earlier contents unusable.
>
Most of this thread has degenerated into pointless bickering. However,
when wildly false generalizations like the one above are made, I get
concerned that some reader might actually believe it to be true and end
up worse off then if the posted message had never been read.
As an example, for registers a1, a2, and a3, the following is a
perfectly fine assembly language statement (at least in our assembler):
a1 a2 & a3
The value previously in a1 is obviously overwritten. However, the
statement above suggests that a2 and/or a3 also somehow no longer
contain their previous values, which is utter rubbish. The instruction
above is what would be generated for a statement of the form i =
iand(j,k).
Robin may have found some non-RISC hardware architecture where his claim
is at least partially true, but it is certainly detrimental to the
usefulness of the mailing list to make sweeping generalizations based on
that limited information.
Cheers,
Bill
--
Bill Long [log in to unmask]
Fortran Technical Support & voice: 651-605-9024
Bioinformatics Software Development fax: 651-605-9142
Cray Inc., 1340 Mendota Heights Rd., Mendota Heights, MN, 55120
|