Ophir Friedler , Wisam Kadry,
Amir Nahir, Vitali Sokhin
{ophirf, wisamk, nahir, vitali}@il.ibm.com
Carla Ferreira, João Lourenço
{carla.ferreira, joao.lourenco}@fct.unl.pt
IBM Research Universidade Nova de Lisboa
1
WTM’13, Prague, April 14, 2013

2
Post Silicon
Post-silicon validation elements:
1.
2.
Stimulating the design under test
Detecting erroneous behavior
3.
4.
Localizing the root cause of the problem
Providing a fix.
WTM’13, Prague, April 14, 2013

3
Stimulation
1.
2.
3.
Test generation
Execution
Consistency checking
4.
Repeat … Forever!
Exerciser Image (Threadmill)
Test Template
Topology
Architectural
Model
Accelerator
Generation
Execution
Silicon
Checking
OS services
WTM’13, Prague, April 14, 2013

•
Detection
Consistency checking
Run the same test-case from the same initial architectural state.
• Expect the same final architectural state ori r10,r0,170 stb r10,0(r6) lbz r11,0(r6)
...
Initial State
R0 = 0x1, R1 = 0x2 …
Final State
R0 = 0xA, R1 = 0xB …
4
Micro-architectural state varies!
Caches, page misses, pre-fetching, thread priorities
WTM’13, Prague, April 14, 2013

Detection
And what if two different final states are manifested? ori r10,r0,170 stb r10,0(r6) lbz r11,0(r6)
...
Initial State
R0 = 0x1, R1 = 0x2 …
R0 = 0xA, R1 = 0xB … ori r10,r0,170 stb r10,0(r6) lbz r11,0(r6)
...
Initial State
R0 = 0x1, R1 = 0x2 …
R0 = 0xC, R1 = 0xB …
MIS-COMPARE
5
WTM’13, Prague, April 14, 2013

6
Localization approach
1. A test-case that produces a mis-compare is found
2. Fast-forward to that test-case on a software simulator
(a.k.a. Reference model)
3. Execute test case on the reference model instruction by instruction and extract information
WTM’13, Prague, April 14, 2013

7
Localization
 Reduce number of resources and instructions that might be the root cause of the mis-compare
 Study the effect of transactions in the test-case on the final state.

Justification:
Force erroneous behaviour on reference model and recreate the mis-compare results
WTM’13, Prague, April 14, 2013

:
1 4 1 2
R1
= suspicious instruction subset
I
2
: t e n
I RR
3 1 3
:
4 1 2
I R R
5 2 1
I t n
6
:
I R
7 3 1 2
8
R2 R3 R4

9
Concluding remarks
 Debug automation effectively reduces the debugging effort.
 Graph analysis holds the potential automate the localization of suspicious resources and instructions
-
-
Future work:
Study the impact of escaped stores in transaction aborts experiment with larger (real-world) cases
WTM’13, Prague, April 14, 2013

Questions
10
WTM’13, Prague, April 14, 2013