Re: [PATCH 0/3] cond-optab merge

[Paolo Bonzini <bonzini at gnu dot org>]

Ramana Radhakrishnan wrote:
> Hi Paolo,
> 
> [Parts of your email snipped].
> 
> Sorry about the rather slow response.
> 
>> * ARM: -mthumb/-march=armv5t -mthumb/-march=armv6t2 -march=armv5
> 
>> If there is any additional testing I can do, please tell. :-)
> 
> I just realized that these options however don't seem to test the
> hardware floating point bits - Would it be possible to double check
> that there are no assembly differences with using -mfpu=vfp
> -mfloat-abi=softfp for each of the above options so that  Hardware
> floating point in the ARM backend is also tested ? We're attempting to
> bootstrap and arm-linux on hardware that might have hardware floating
> point but it might take a couple of days for the results to show up.

Done (with a couple more patches which are especially floating-point
related).  There were no ICEs.  The additional patches introduce a lot
of churn at -O0; after removing -O0 differences I was able to look at
the changes, and the changes (improvements) are more or less exclusively
what I expected from the additional patches that I included in the test.

Sorry for the informal nature of the description above. :-)

Paolo