[libre-riscv-dev] [Bug 154] Cell for Dependency Matrices is needed

[bugzilla-daemon at libre-riscv.org]

http://bugs.libre-riscv.org/show_bug.cgi?id=154

--- Comment #18 from Luke Kenneth Casson Leighton <lkcl at lkcl.net> ---
(In reply to Staf Verhaegen from comment #16)


> placement of the cells is. When scaling
> to smaller nodes, the placer has to be changed to do placement for
> optimizing timing and do proper buffer insertion.

i missed this earlier.  i remember now a discussion on comp.arch where people
warned about driving a 128 bit unary array from a single source.

that is effectively what each row (and column) inputs are driven by.

each input from the left side drives every entry in the DM Row

each input from the top drives every entry in a DM column.

thus you are absolutely right, there will be buffers needed.

i believe that timing *should* be ok because despite the size it is extremely
regular.  no interactions horizontally, no interactions vertically.  *only* at
the actual DependencyCell itself does interaction take place.

very much like an SRAM Matrix.


this will be, i assume and hope, that Jean Paul will be teaching the interns at
LIP6.fr about buffers and timing.

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