NaCl docs: an update on AVX

NaCl docs: an update on AVX

NaCl now supports AVX2.

R=shyamsundarr@chromium.org
TBR=jmedley@chromium.org, shyamsundarr@chromium.org
BUG= https://code.google.com/p/nativeclient/issues/detail?id=3637

Committed: https://chromium.googlesource.com/chromium/src/+/8084a6a0ffe56840c29f725dec5d75c120fd838c

[JF, 2015-01-31 18:09:55]

Shyam, should I be more nuanced on AVX2 support? We should probably specify
which instructions we don't support.

Maybe we can also improve the x86 docs here:
  https://developer.chrome.com/native-client/reference/sandbox_internals/index
?

[shyamsundarr, 2015-02-02 16:33:45]

On 2015/01/31 18:09:55, JF wrote:
> Shyam, should I be more nuanced on AVX2 support? We should probably specify
> which instructions we don't support.
> 
> Maybe we can also improve the x86 docs here:
>   https://developer.chrome.com/native-client/reference/sandbox_internals/index
> ?

How about "except for all variants of VGATHER"?. Some of them can be supported,
but given that some of them can't be,it makes sense to be lazy and I will wait
for someone to ask for it.

[JF, 2015-02-02 16:47:47]

On 2015/02/02 16:33:45, shyamsundarr wrote:
> On 2015/01/31 18:09:55, JF wrote:
> > Shyam, should I be more nuanced on AVX2 support? We should probably specify
> > which instructions we don't support.
> > 
> > Maybe we can also improve the x86 docs here:
> >  
https://developer.chrome.com/native-client/reference/sandbox_internals/index
> > ?
> 
> How about "except for all variants of VGATHER"?. Some of them can be
supported,
> but given that some of them can't be,it makes sense to be lazy and I will wait
> for someone to ask for it.

Done.

Anything else to add? It would be good to update the sandbox internals docs (in
a different CL), since they're basically what we tell people to use when trying
to figure out how to write asm.

[shyamsundarr, 2015-02-02 17:25:30]

On 2015/02/02 16:47:47, JF wrote:
> On 2015/02/02 16:33:45, shyamsundarr wrote:
> > On 2015/01/31 18:09:55, JF wrote:
> > > Shyam, should I be more nuanced on AVX2 support? We should probably
specify
> > > which instructions we don't support.
> > > 
> > > Maybe we can also improve the x86 docs here:
> > >  
> https://developer.chrome.com/native-client/reference/sandbox_internals/index
> > > ?
> > 
> > How about "except for all variants of VGATHER"?. Some of them can be
> supported,
> > but given that some of them can't be,it makes sense to be lazy and I will
wait
> > for someone to ask for it.
> 
> Done.
> 
> Anything else to add? It would be good to update the sandbox internals docs
(in
> a different CL), since they're basically what we tell people to use when
trying
> to figure out how to write asm.

lgtm.

for the other document, i guess there haven't really been any changes to the
sandboxing rules? avx1/avx2/fma3 all follow the existing sandbox model.

[commit-bot: I haz the power, 2015-02-02 18:05:45]

Patchset 2 (id:??) landed as
https://crrev.com/8084a6a0ffe56840c29f725dec5d75c120fd838c
Cr-Commit-Position: refs/heads/master@{#314166}

[JF, 2015-02-02 18:06:50]

Committed patchset #2 (id:20001) manually as
8084a6a0ffe56840c29f725dec5d75c120fd838c (presubmit successful).