[Libre-soc-bugs] [Bug 558] gcc SV intrinsics concept

[bugzilla-daemon at libre-soc.org]

https://bugs.libre-soc.org/show_bug.cgi?id=558

Luke Kenneth Casson Leighton <lkcl at lkcl.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|lkcl at lkcl.net               |oliva at libre-soc.org
           See Also|                            |https://bugs.libre-soc.org/
                   |                            |show_bug.cgi?id=550

--- Comment #1 from Luke Kenneth Casson Leighton <lkcl at lkcl.net> ---
alexandre, ok so i have an idea here: if i put in an NLnet grant request for
adding bitmanip and cryptographic intrinsics to gcc would you be interested to
add SV intrinsic support at "just the level above svp64"?

the idea would be to not actually add vector datatypes as was done for RVV
(except vec2/3/4), instead to "mark" variables as "SV Vectorised".  and also to
mark them as "saturated".

when these get used, svp64 prefixes are automatically added.

an array of uint64_t for example would automatically result in a "Vector"
EXTRA2/3 mark.

an intrinsic "Push Vectorisation Context" would contain the MAXVL. Pop removes
it.

the only actual intrinsic functions needed would be to support things like the
crypto primitives and anything unusual that doesn't cleanly map to c (set
before first opcode)


the rest of the gramt request would be for linux kernel module support.

a *separate* grant would be for hardware.

what do you think?

-- 
You are receiving this mail because:
You are on the CC list for the bug.