M1 CPU features

[Jason Liu]

In addition to SIMDe, projects like Blender are using sse2neon to allow
them to compile Blender on arm64 while using libraries that use Intel
intrinsics and haven't been ported for use with Arm Neon.

https://github.com/DLTcollab/sse2neon

I plan to package sse2neon after I update a couple of the other ports I'm
maintaining. Maybe I'll try packaging SIMDe at the same time. However,
since I don't own an M1 Mac, I'm hoping that people on this mailing list
would be able to help test my portfile before I submit them. Any
volunteers? :)

-- 
Jason Liu


On Mon, Apr 26, 2021 at 3:16 PM Georges Martin <jrjsmrtn at gmail.com> wrote:

> And this article describes SIMDe:
>
>
> https://simd-everywhere.github.io/blog/2020/06/22/transitioning-to-arm-with-simde.html
>
> SIMD Everywhere (SIMDe) provides fast, portable, permissively-licensed
> (MIT) implementations of the x86 APIs which allow you to run code designed
> for x86/x86_64 CPUs pretty much anywhere, including on Arm (using NEON if
> available). With almost no source code changes, you can recompile your x86
> SIMD code for Arm (or POWER, or WebAssembly, etc.).
>
>
> ...that is not yet packaged for MacPorts ;-) ;-) ;-)
>
> G.
>
> Le 26 avr. 2021 à 20:56, Georges Martin <jrjsmrtn at gmail.com> a écrit :
>
> Aha, hw.optional! That's useful, thanks Georges!
>
>
> You're welcome :-) You also have:
>
> hw.optional.amx_version: 2
> hw.optional.arm64: 1
> hw.targettype: J313
>
> "amx" is the Neural Engine and I think "J313" is the code name for the M1.
>
> You may find this article very interesting:
>
> https://levelup.gitconnected.com/armv9-what-is-the-big-deal-4528f20f78f3
>
> It describes the new ARMv9 instruction set with SVE2 and how it compares
> to Intel/AMD MMX/SSE/AVX and NEON/SVE.
>
> Question is: would Apple adopt ARMv9 with SVE2 in a M2 for a future Mac
> Pro ? ;-)
>
> G.
>
> Le 26 avr. 2021 à 20:44, Jason Liu <jasonliu at umich.edu> a écrit :
>
> Aha, hw.optional! That's useful, thanks Georges!
>
> --
> Jason Liu
>
>
> On Mon, Apr 26, 2021 at 2:16 PM Georges Martin <jrjsmrtn at gmail.com> wrote:
>
>> $ sysctl hw.optional | grep -E 'neon|armv8'
>> hw.optional.neon: 1
>> hw.optional.neon_hpfp: 1
>> hw.optional.neon_fp16: 1
>> hw.optional.armv8_1_atomics: 1
>> hw.optional.armv8_crc32: 1
>> hw.optional.armv8_2_fhm: 1
>> hw.optional.armv8_2_sha512: 1
>> hw.optional.armv8_2_sha3: 1
>>
>> Le 26 avr. 2021 à 19:55, Jason Liu <jasonliu at umich.edu> a écrit :
>>
>> On Mon, Apr 26, 2021 at 1:42 PM Christopher Jones <
>> jonesc at hep.phy.cam.ac.uk> wrote:
>>
>>> Thats not at all surprising as those instruction sets are very much
>>> specific to X86_64 systems.
>>>
>>> RISC processors, Arm, do have their own sets of SIMD instructions (e.g.
>>> Neon), but they are entirely different to those on X86_64 machines.
>>>
>>> Whether or these are supported on Apple’s M1 processors I have no idea.
>>>
>>
>> It looks like the M1 supports Neon SIMD instructions, but not SVE SIMD
>> (which I guess is supposed to be similar to AVX?):
>>
>> https://discussions.apple.com/thread/252073619
>>
>> --
>> Jason Liu
>>
>>
>> On Mon, Apr 26, 2021 at 1:42 PM Christopher Jones <
>> jonesc at hep.phy.cam.ac.uk> wrote:
>>
>>>
>>>
>>> On 26 Apr 2021, at 6:28 pm, Jason Liu <jasonliu at umich.edu> wrote:
>>>
>>> Thanks Arno :)
>>>
>>> I'm kind of surprised that the M1 doesn't seem to support any SSE or
>>> AVX....
>>>
>>>
>>>
>>> Thats not at all surprising as those instruction sets are very much
>>> specific to X86_64 systems.
>>>
>>> RISC processors, Arm, do have their own sets of SIMD instructions (e.g.
>>> Neon), but they are entirely different to those on X86_64 machines.
>>>
>>> Whether or these are supported on Apple’s M1 processors I have no idea.
>>>
>>> Chris
>>>
>>>
>>>
>>>
>>> Does "sysctl machdep.cpu.features" return anything?
>>>
>>> --
>>> Jason Liu
>>>
>>>
>>> On Mon, Apr 26, 2021 at 1:23 PM Arno Hautala <arno at alum.wpi.edu> wrote:
>>>
>>>> > On 26 Apr 2021, at 13:20, Jason Liu <jasonliu at umich.edu> wrote:
>>>> >
>>>> > sysctl machdep.cpu.brand_string ; sysctl machdep.cpu | grep -i
>>>> "avx\|sse”
>>>>
>>>> $ sysctl machdep.cpu.brand_string ; sysctl machdep.cpu | grep -i
>>>> "avx\|sse”
>>>> machdep.cpu.brand_string: Apple M1
>>>>
>>>> --
>>>> arno  s  hautala    /-|   arno at alum.wpi.edu
>>>>
>>>> pgp b2c9d448
>>>>
>>>>
>>>>
>>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.macports.org/pipermail/macports-dev/attachments/20210427/566b7000/attachment.htm>