Write the veto gate and two injectors for a VIP whose one hook may edit or drop a beat. The ledger must match the pins.
Hardware context: Lane VIP with a hook that may edit or drop a beat. Build a veto gate and stackable injectors with a ledger that matches the pins.
A lane VIP offers one hook, received, in the style of the UVM 1.1 guide. A callback may edit the beat, and returns 0 to drop it. The driver asks a gate object before each beat, and drops the beat when the gate says 0.
Write the gate, a drop injector and a parity injector, and install them on a lane. The scoreboard trusts the ledger. Every drop, corruption and sent beat recorded there must be exactly what reached the pins. A beat planned for both is dropped, never corrupted.
Full specification
// Asks every callback in order. The first one that returns 0 drops the beat, and the
// callbacks after it are not asked. Record each beat that will be sent in ledger.
class lane_gate extends lane_gate_base;
function new(string name = "lane_gate"); super.new(name); endfunction
function bit ask(lane_driver drv, lane_item it, lane_beat beat);
return 1;
endfunction
endclass
// Drops the items listed in on. Record each drop in ledger.
class drop_cb extends lane_cb;
int on[$];
inject_ledger ledger;
function new(string name = "drop_cb"); super.new(name); endfunction
endclass
// Sends the items listed in on with bad parity. Record each corruption in ledger.
class corrupt_cb extends lane_cb;
int on[$];
inject_ledger ledger;
function new(string name = "corrupt_cb"); super.new(name); endfunction
endclass
// Called once per lane with that lane's driver and injectors.
function void install(lane_driver drv, drop_cb d, corrupt_cb c);
endfunction
| Given | Meaning |
|---|---|
| lane_cb::received(drv, it, beat) | The hook. Return 1 to let the beat go on, 0 to drop it. The base returns 1. |
| ledger.note(kind, n) | Record "drop", "corrupt" or "sent" for item n. |
| lane_driver | Calls gate.ask(this, req, beat) after it builds the beat. On 0 it calls item_done and drives nothing. |
| on | The item numbers an injector acts on, set by the test. |
The gate asks the callbacks in queue order and stops at the first that returns 0. Later callbacks are not asked. It records "sent" exactly for the beats it lets through.
Examples
Example 1
Drop Item 2, Corrupt Item 4- Input
5 items; drop [2], corrupt [4]
- Output
items_on_pins = {1, 3, 4, 5}, bad_parity_on_pins = {4}, ledger_dropped = {2}, ledger_corrupted = {4}, ledger_sent = {1, 3, 4, 5}
Item 2 never reaches the pins; item 4 arrives with bad parity. The ledger lists the drop, the corruption and the four sent items.
Example 2
Item 3 Is Planned for Both- Input
4 items; drop [3], corrupt [1, 3]
- Output
items_on_pins = {1, 2, 4}, bad_parity_on_pins = {1}, ledger_dropped = {3}, ledger_corrupted = {1}, ledger_sent = {1, 2, 4}
The drop injector answers first and returns 0, so the parity injector is never asked about item 3. Only item 1 is corrupted.
Example 3
Nothing Planned- Input
3 items; drop [], corrupt []
- Output
items_on_pins = {1, 2, 3}, bad_parity_on_pins = {}, ledger_dropped = {}, ledger_corrupted = {}, ledger_sent = {1, 2, 3}
Every callback returns 1, the gate returns 1, and all three items are sent with good parity.
Constraints and Contract
- Keep every class and function name and signature; the bench creates and calls them.
- Flip beat.par to corrupt; do not touch the item.
- install is called once per lane, in connect_phase.
Hints
Hint 1
UVM has a callback macro that stops at the first callback returning a chosen value. Read what it expands to before you put code after it.
Hint 2
If the parity injector is asked before the drop injector, what does the ledger say about an item planned for both?
Hint 3
The macro returns from the function it sits in. Where can the "sent" record go so it runs only for beats that will be sent?
What the grader checks: Your gate and injectors run on six real lanes with hidden plans. The checks compare the items the sink took and their parity with your ledger's drops, corruptions and sent beats.
6 checks on UVM · server. Run the examples first, then check the full regression. Full checks use the server judge. Downloads contain public samples only. A passing result establishes the tested cases, not a proof for all possible inputs.
See all 1 problem on this path