RFR: 8265783: Create a separate library for x86 Intel SVML assembly intrinsics

[Andrew Haley]

On 5/20/21 12:34 AM, Paul Sandoz wrote:
> I don’t think this should be considered a generally acceptable approach for Vector API operations (most code for operations does not and should not follow this approach), nor is it generally acceptable for other kinds of intrinsic in HotSpot (I believe there are a few special cases under os_cpu). Thus we should dissuade the use of .S source for other intrinsic cases.

I've got nothing at all against .S files, as long as they are the real
preferred form. That is to say, they should be the actual source code,
as written by someone.

> Does this help alleviate some of your concerns?

Somewhat, but I wonder if this, as a matter of policy, is an area in
which the Governing Board should get involved. I don't want to hold up
progress, of course, but this is potentially a very important issue.

I guess I wouldn't mind as long as we had a "This far, and no further"
policy, with some hope that the library could be replaced by readable
and maintainable code.

-- 
Andrew Haley  (he/him)
Java Platform Lead Engineer
Red Hat UK Ltd. <https://www.redhat.com>
https://keybase.io/andrewhaley
EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671