diff options
author | Wade Fife <wade.fife@ettus.com> | 2021-09-10 15:49:40 -0500 |
---|---|---|
committer | Wade Fife <wade.fife@ettus.com> | 2021-09-13 13:20:37 -0500 |
commit | 71aedc922d9cd5348ad4829cf323e569688f660f (patch) | |
tree | 5a3d8d3d392343340b5d1e03c454b527bb40c616 /fpga/usrp3/lib/control | |
parent | 09d94529e5dbfa992cb6012cc0dc38d6f3e7d57d (diff) | |
download | uhd-71aedc922d9cd5348ad4829cf323e569688f660f.tar.gz uhd-71aedc922d9cd5348ad4829cf323e569688f660f.tar.bz2 uhd-71aedc922d9cd5348ad4829cf323e569688f660f.zip |
fpga: lib: Update example constraint in synchronizer
Diffstat (limited to 'fpga/usrp3/lib/control')
-rw-r--r-- | fpga/usrp3/lib/control/synchronizer.v | 58 |
1 files changed, 40 insertions, 18 deletions
diff --git a/fpga/usrp3/lib/control/synchronizer.v b/fpga/usrp3/lib/control/synchronizer.v index 177b1219c..6ac16012d 100644 --- a/fpga/usrp3/lib/control/synchronizer.v +++ b/fpga/usrp3/lib/control/synchronizer.v @@ -1,9 +1,47 @@ // -// Copyright 2014 Ettus Research LLC -// Copyright 2018 Ettus Research, a National Instruments Company +// Copyright 2021 Ettus Research, a National Instruments Brand // // SPDX-License-Identifier: LGPL-3.0-or-later // +// Module: synchronizer.v +// +// Description: +// +// This is a double-synchronizer module. However, it can can implement a +// synchronizer of any depth (double, triple, etc.) and width. +// +// A double synchronizer is typically used to cross a single-bit glitch-free +// signal from one clock domain to another. +// +// *WARNING*: The input signal must be glitch-free. In other words, it should +// be driven by a register and NOT combinational logic. Otherwise +// you could capture a glitch instead of the intended signal. +// +// *WARNING*: When WIDTH is not 1, the multiple bits are not guaranteed to be +// coherent. In other words, they can arrive on the output at +// different times. This module should not usually be used to +// cross a multi-bit signal. Consider using the handshake module +// instead. +// +// When crossing between unrelated clock domains, we typically don't want the +// timing analyzer to consider the path between clock domains. To make +// writing this constraint easier, the FALSE_PATH_TO_IN parameter controls +// the name of the synchronizer_impl instance. The following XDC constraint +// is used to ignore all instances of this false path. +// +// set_false_path -to [get_pins -hierarchical -filter \ +// {NAME =~ */synchronizer_false_path/stages[0].value_reg[0][*]/D}] +// +// Parameters: +// +// WIDTH : Width of the synchronizer (1 by default). +// STAGES : Number of synchronizer stages (2 by default, for a +// standard double-synchronizer). +// INITIAL_VAL : Initial value of the output register (0 by default). +// FALSE_PATH_TO_IN : Set to 1 if the input should be considered a false path +// and ignored by the timing analyzer. Set to 0 to let the +// tool analyze this path. +// module synchronizer #( parameter WIDTH = 1, @@ -17,22 +55,6 @@ module synchronizer #( output [WIDTH-1:0] out ); - //Q: Why do we have a separate impl and instantiate - //it with a different instance name based on this - //arbitrary parameter FALSE_PATH_TO_IN? - //A: To make constraining these synchronizers easier. - //We would like to write a single false path constraint - //for all synchronizers when the input is truly async. - //However other cases might require constraining the input - //of this module. - //To enable this, all clients that hook up async signals to - //the "in" port can set FALSE_PATH_TO_IN=1 (or use the default) - //and all clients that want the "in" delay to be constrained can - //set FALSE_PATH_TO_IN=0. - //In the XDC we can write the following async constraint: - //set_false_path -to [get_pins */synchronizer_false_path/stages[0].value_reg[0]/D] - //and this will take care of all instances of this module with FALSE_PATH_TO_IN==1 - generate if (FALSE_PATH_TO_IN == 1) begin synchronizer_impl #( .WIDTH(WIDTH), .STAGES(STAGES), .INITIAL_VAL(INITIAL_VAL) |