[libre-riscv-dev] pipeline sync issues

Luke Kenneth Casson Leighton lkcl at lkcl.net
Fri Apr 12 02:59:32 BST 2019

On Fri, Apr 12, 2019 at 2:24 AM Luke Kenneth Casson Leighton
<lkcl at lkcl.net> wrote:

> FIFOControl has been tested with a simple Signal(16) as input and
> output, i'm just about to try chaining the above with a FIFOControl
> that takes the Example2OpRecord as input and output.

 oleeeee! :)

 ok so that introduces 2 clock cycles, where one is all that's
necessary.  the FIFOControl introduces a clock cycle delay, and so
does the adder (using SimpleHandshake).

 if it were possible to allow FIFOControl to perform *processing* of
its input data (pre and/or post), there's no need for separate
pipe-stages, and, crucially, BufferedHandshake can be deleted.

 if on the other hand the Stage API *was* the Pipe API - i.e. if
Stages contained *BOTH* data control *AND* data processing, then
hypothetically it would be possible to use combinatorial chaining
between FIFOControl and SimpleHandshake to eliminate the extra clock

 was this what you have been referring to, jacob?


More information about the libre-riscv-dev mailing list