[Libre-soc-bugs] [Bug 713] PartitionedSignal enhancement to add partition-context-aware lengths

[bugzilla-daemon at libre-soc.org]

https://bugs.libre-soc.org/show_bug.cgi?id=713

--- Comment #114 from Luke Kenneth Casson Leighton <lkcl at lkcl.net> ---
(In reply to Jacob Lifshay from comment #112)
> After thinking a bit, the actual issue is that element-widths are the thing
> arithmetic needs to happen on -- you can come up with whatever algorithm you
> like for deciding what full simd width to use, as long as the element widths
> take priority.

with likewise sone thought, tending to agree with you, here.

the fixed_width scenario, after creating the
dictionary of Vector Element Widths (vec_el_widths in layout())
i *think* that from that point onwards it's ok to do sums
on those widths (basic ones!)

if people really really really want to do things based on
fixed_width then, well, they declare a new SimdShape.

this is based on the (preliminary) premise that if you
set a fixed_width of say 64, and combined that with part_count of
{1,2,4,8} to auto-create vec_el_widths of {64,32,16,8}
and you add say 2 to that (to create room for carry-in
and carry-out in the ADD ALU for example) you should reasonably
expect a new vec_el_widths of {64+2,32+2,16+2,8+2}

this *even though* you initially created the SimdShape with
a fixed_width.

i updated the url btw.
https://libre-soc.org/3d_gpu/architecture/dynamic_simd/shape/?updated

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