[cairo] [cairo-announce] [ANNOUNCE] Stable pixman release 0.16.0 now available

[Koen Kooi]

On 02-09-09 12:49, Siarhei Siamashka wrote:
> On Wednesday 02 September 2009, Koen Kooi wrote:
>> On 02-09-09 00:49, Siarhei Siamashka wrote:
>>> On Monday 31 August 2009, Koen Kooi wrote:
>>>> And I still hate that pixman puts in it's own -march and -mcpu, but
>>>> that's a discussion for another day.
>>>
>>> Pixman puts -mcpu only into the flags for compiling some selected files,
>>> *after* configure script has already verified that gcc can accept this
>>> set of flags. If gcc would reject this combination of flags, it would not
>>> try to compile these sources at all. Still a missing part here is
>>> -mfloat-abi=softfp option for the source file with NEON optimizations.
>>>
>>> This behavior is consistent with the rest of pixman, except that gcc
>>> behaves a lot more ugly on ARM than on other platforms.
>>
>> This is an excerpt of the gcc commandline from pixman:
>>
>> arm-angstrom-linux-gnueabi-gcc \
>> -march=armv7-a -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp \
>> -mcpu=arm1136j-s
>>
>> I split it between what I'm using when building for armv7a and what
>> pixman adds.
>
> You omitted the most important detail: the file for which this
> additional -mcpu=arm1136j-s option is added.
>
> The fact is that it is enabled for 'pixman-arm-simd.c' exclusively, which
> contains nothing more, but ARMv6 specific optimizations.
>
>> So pixman adds too little to be usefull and what it adds is
>> not right for the target I'm targetting.
>
> This file with overridden cpu type contains the code, which is only important
> for the ARM cores which support ARMv6 but not ARMv7 instructions, it is not
> supposed to be used on your target at all (or at least it is not optimal
> for it anyway).
>
> You can actually reduce the size of compiled binaries a bit by
> adding --disable-arm-simd configure option, so this file will not
> be compiled in for your target.
>
>> Can we please add a
>> --I'm-using-ARM-but-actually-know-what-I'm-doing configure switch to
>> disable this behaviour?
>
> Yes. Can you propose some patch?

I'll have a look at cooking up a patch.

regards,

Koen