[PATCH][ARM][1/7] Convert FP mnemonics to UAL | mov patterns

Bin.Cheng amker.cheng@gmail.com
Fri Sep 12 02:53:00 GMT 2014


On Wed, Sep 10, 2014 at 6:21 PM, Bin.Cheng <amker.cheng@gmail.com> wrote:
> On Wed, Sep 10, 2014 at 4:57 PM, Kyrill Tkachov <kyrylo.tkachov@arm.com> wrote:
>>
>> On 10/09/14 09:40, Christophe Lyon wrote:
>>>
>>> Hi,
>>
>> Hi Christophe,
>>
>>>
>>> On 9 September 2014 13:02, Ramana Radhakrishnan
>>> <ramana.gcc@googlemail.com> wrote:
>>>>
>>>> On Tue, Aug 19, 2014 at 4:22 PM, Kyrill Tkachov <kyrylo.tkachov@arm.com>
>>>> wrote:
>>>>>
>>>>> Hi all,
>>>>>
>>>>> In this patch the move patterns are updated.
>>>>> For the fconst case where the constant is encoded in a decimal
>>>>> representation before going into the immediate field of the assembly
>>>>> instruction UAL syntax allows for the real operand to be output directly
>>>>> and leaves the assembler to do the encoding.
>>>>> This simplifies the logic in arm_print_operand a bit.
>>>>>
>>>>> fp_immediate_constant is deleted and it seems that the function was not
>>>>> meant to be used anyway (it returned "0" for all inputs!)
>>>>>
>>>>> Ok for trunk?
>>>>
>>>> Ok.
>>>>
>>>> Ramana
>>>>
>>> I've noticed that your patch makes
>>> gcc.target/arm/memset-inline-5.c
>>> fail when the compiler generates code for a57 + crypto-neon-fp-armv8.
>>
>>
>> I've managed to reproduce the failure as well (don't forget the
>> -mfloat-abi=hard), however
>> I tried with and without my patch and could still see the failure. Looking
>> at the test itself I don't
>> see what my patch could have changed in this department...
>>
>> I'm cc'ing Bin since he added the test for his recent memset inlining work.
>> Bin, could you have a look please?
> I will have a look tomorrow.
>
I think this is because I didn't enable memset inlining with neon on
aarch32 processors at the first place.  It has already been noticed
and patch is in internal review.

Thanks,
bin



More information about the Gcc-patches mailing list