Re: adding binary comparisons to MIPS testsuite

[cgd at sibyte dot com (Chris G. Demetriou)]

Nick Clifton <nickc@redhat.com> writes:
> Assuming of course that the binary values in the .d file are correct
> and have not just been produced by running objdump --show-raw-insn and
> then converting the output into a .d file.

Heh.  "No cheating," eh?  8-)

Actually, that's what i've been doing to get the test files, but i
follow it with:

(1) hand-comparison with the binary dumps we'd previously produced by
hand, in a different format, and

(2) actual checking of both against the relevant MIPS architecture
documentation.

generating the test files that way is the easiest way to get the data
in the right places in the file -- anything else requires a lot of
tedious cut and paste repetition or error-prone typing in addition to
paper checking.  8-)


> Given that gas/testsuite/gas/mips/mips.exp already has lots of
> conditional tests, I see no reason why you should not add more.  Try
> to test on several variants though, to make sure that there are no
> hidden gotchas.

yah.


OK, well, I've got some other things going on right now, so it'll be a
bit before you see any result from this question...

My current road-map looks something like:

* produce and submit some diffs to add the remainder of the
instructions added as part of MIPS32, which haven't been covered in
previous patches.  (Unfortunately, the patches based on the 4K[cpm]
aren't complete for generic MIPS32 -- for instance, the 4K cores don't
include FP support.)

* produce and submit diffs to add MIPS64 instructions.

and then after that who knows.  We've had diffs to do those things
("and more!!!" 8-) in our development tree for ... a while now.



chris