29 lines
		
	
	
		
			875 B
		
	
	
	
		
			Plaintext
		
	
	
	
			
		
		
	
	
			29 lines
		
	
	
		
			875 B
		
	
	
	
		
			Plaintext
		
	
	
	
| AUTHOR = "David Sharp <dhsharp@google.com>"
 | |
| NAME = "Tracing microbenchmark"
 | |
| TIME = "SHORT"
 | |
| TEST_CATEGORY = "Benchmark"
 | |
| TEST_CLASS = "Kernel"
 | |
| TEST_TYPE = "client"
 | |
| 
 | |
| DOC = """
 | |
| A simple benchmark of kernel tracers such as ftrace. Enables tracepoints in
 | |
| sys_getuid and makes 100,000 calls to getuid with tracing on and off to measure
 | |
| the overhead of enabling tracing. The intent for this benchmark is to not
 | |
| overflow the ring buffer, so the buffer is generously sized.
 | |
| 
 | |
| 
 | |
| tracer:  tracepoint enabled
 | |
| ------
 | |
| off:     n/a
 | |
| ftrace:  syscalls:sys_enter_getuid
 | |
| 
 | |
| Args:
 | |
|   tracer: see table above.
 | |
|   buffer_size_kb: Set the tracing ring buffer to this size (per-cpu).
 | |
|   calls: Set the number of calls to make to getuid.
 | |
| """
 | |
| 
 | |
| 
 | |
| job.run_test('tracing_microbenchmark', tracer='off', tag='off', iterations=10)
 | |
| job.run_test('tracing_microbenchmark', tracer='ftrace', tag='ftrace', iterations=10)
 |