Transcript Accelerated Code Coverage on HW Emulator
Coverage Solutions on Emulators
Ravi P Gupta
Agenda
• Functional Verification Overview (~5 Mins) • • Traditional Verification Challenges Future requirements and Solutions • Code coverage with PXP (~5 Mins) • Performance analysis with coverage enabled (~5 Mins) • PXP Code coverage Result analysis • Support of Functional coverage 2 Presentation Title 26/04/2020
Functional Verification
• Definition of a test plan.
• Implementation using random test generators that produce a large number of test cases.
• Comparing the result to the expected results in order to say if the test passed.
• Analyze how much functionality of the design has been verified.
• coverage tools – measures the percentage of design functionality covered. • Detect the occurrence of events in the test plan.
• Provide information related to the progress of the test plan.
• Analysis of the coverage reports allows the verification team to modify the directives for the test generators in Verification 3 Presentation Title 4/26/2020
Challenges in Functional Verification
4 • Around 70% of product development cycle time is consumed in verification activity.
• Simulation based verification is very slow and no longer sufficient to meet the demands of a complex IPs and SOC.
• Dedicated hardware solutions are too expensive to develop.
• Considerable effort is invested in finding ways to close the loop of coverage analysis and test generation.
Presentation Title 4/26/2020
Requirement & Solution
Requirements
• Accelerating functional verification • Closing verification process Can verification closure be accelerated??
Solution
• Hardware accelerated simulation 5 Testbench
Module 0 Module 1 Module 2
Hardware Accelerator
Module 2 is synthesized & compiled into Custom processors
Presentation Title 4/26/2020
Hardware Accelerated Simulation
6 •
Pros
• Much faster than simulation.
• • Provides simulation like verification flow.
Debugging is easier than customize hardware. •
Cons (Obstacles to overcome)
• • • Mapping RTL design into the hardware can be substantial longer. SW-HW communication speed can degrade the performance. Tool cost is much more than simulators.
Presentation Title 4/26/2020
Idea!!
7 • The addition of coverage to emulators can accelerates the detection of inadequate functional verification and augments the efficiency of the verification engineer in writing test cases by focusing on uncovered areas in the design.
Presentation Title 4/26/2020
Metric driven verification flow
8 Presentation Title 4/26/2020
Simulation vs. Emulation
•
Design has two main sub-module
• Downstream • Upstream 9 Presentation Title 4/26/2020
Steps to generate Toggle database
10 • Instrumentation add
+tcov
and
+tcovType+
to elaboration command • Run the design irun –R
+ixccWorkDir+
<>
+ixccTest+
<> • Report generation (Equivalent to .res file of internal solution) ixcc
–inputDir
<>
-input
<> -summary | tee <>.log
NOTE: tcovType can be
only ports
or
ports, Verilog & VHDL variables, Verilog nets & VHDL signals
Presentation Title 4/26/2020
Result analysis (1/3)
Testcase
Sx_ipbypass Sx_ipreseqbypass 19.1% Sx_macbypass 17.6% Sx_macipbypass
IES Coverage
17.4% 19.3%
PXP Coverage
17.2% 18.9% 17.5% 19.1% 11 Presentation Title 4/26/2020
Result analysis (2/3)
•
Coverage results come in txt file as below : Instance Name: mem_tb.DUT
File Name: mem.v
Hit(Full) Hit(Rise) 1 1 Hit(Fall) 1 0 1 0 rst <><><><><><><><><><><><><><><><><><><><><><><><><><><><> : Instance Name: mem_tb.DUT.addr_inst
File Name: mem.v
Hit(Full) Hit(Rise) Hit(Fall) Signal 1 0 1 1 1 0 Signal clk a[7] out[7]
Presentation Title 4/26/2020 12
Result analysis (3/3)
LOG = Local Overall Grade LOC = Local Overall Covered OG = Overall Grade OC = Overall Covered Instance LOG LOC OG OC -------------------------------------------------------------------------------------------------------------- mem_tb.DUT 81.5 44 / 54 87.5 77 / 88 mem_tb.DUT.addr_inst 97.1 33 / 34 97.1 33 / 34
•
Merging of different test coverage database is possible.
•
iccr support likely to come in next release.
Presentation Title 4/26/2020 13
Performance impact
Config. Type
Only ports
Resource used before toggle enabled
62% Ports+internal 62%
Resource used after toggle enabled
63% 88%
Frequency before toggle enabled
1.33MHz
1.33MHz
Frequency after toggle enabled
1.33MHz
0.97MHz
14 Presentation Title 4/26/2020
Steps to generate functional cov. database
• Instrumentation add
+sv
and
–functional coverage
options to elaboration command • Run the design irun –R
–gui –cov.work/scope/test
15 Presentation Title 4/26/2020
Conclusion
• Palladium-XP is able to support most of the SV functional coverage constructs and toggle coverage • Coverage on Emulation without much penalty on performance & area utilization.
• Traditional flow for result analysis.
PXP, not only just accelerate the verification but also verification closure.
16 Presentation Title 4/26/2020
Presentation Title 4/26/2020 17
Presentation Title 4/26/2020 18