[otbn, dv] Fix for regression failures in otbn_imem_err
Delay for waiting for completion of alert handshake has been
synchronized to async_clk instead of clk as the async_receiver_driver
drives the ack_p signal sunchronized to async_clk.
delay chosen = max_ack_delay + max_ack_stable + arbitrary delay for
alert_p to go down + 2 cyclks of clk_i = 50 clocks (approximately chosen
as it is difficult to quantify the last two delays)
Signed-off-by: Prajwala Puttappa <prajwalaputtappa@lowrisc.org>
diff --git a/hw/ip/otbn/dv/uvm/env/seq_lib/otbn_imem_err_vseq.sv b/hw/ip/otbn/dv/uvm/env/seq_lib/otbn_imem_err_vseq.sv
index 107fd83..13d5561 100644
--- a/hw/ip/otbn/dv/uvm/env/seq_lib/otbn_imem_err_vseq.sv
+++ b/hw/ip/otbn/dv/uvm/env/seq_lib/otbn_imem_err_vseq.sv
@@ -42,7 +42,7 @@
// doesn't actually fetch anything until that happens. This isn't as lax as it looks, though. We
// know that the alert is supposed to go out at the same time as STATUS changes, so can add a
// timeout there.
- cfg.scoreboard.set_exp_alert(.alert_name("fatal"), .is_fatal(1'b1), .max_delay(10000));
+ cfg.scoreboard.set_exp_alert(.alert_name("fatal"), .is_fatal(1'b1), .max_delay(100000));
cfg.model_agent_cfg.vif.invalidate_imem <= 1'b1;
@(cfg.clk_rst_vif.cb);
@@ -60,9 +60,11 @@
wait_alert_trigger("fatal", .wait_complete(1));
seen_alert = 1'b1;
end
- repeat (10) @(cfg.clk_rst_vif.cb);
+ begin
+ repeat (50) @(cfg.m_alert_agent_cfg["fatal"].vif.receiver_cb);
+ end
join_any
- `DV_CHECK_FATAL(seen_alert, "No alert after 10 cycles")
+ `DV_CHECK_FATAL(seen_alert, "No alert after 50 cycles")
disable fork;
end
join
