Summary of discussions with David.
I am not able to reproduce the 'Test Fail' issue. However, log indicates an 
issue in the test code. David has validated the patch and I will send it out 
soon.

The test case taking long time to complete happens when the code is compiled 
with '-O0 -g', but the test case itself completes. The test cases are currently 
tuned to take less time when compiled with '-O3'.

Thanks,
Honnappa

From: David Marchand <[email protected]> 
Sent: Friday, May 3, 2019 9:31 AM
To: Honnappa Nagarahalli <[email protected]>
Cc: Ananyev, Konstantin <[email protected]>; Stephen Hemminger 
<[email protected]>; [email protected]; Kovacevic, Marko 
<[email protected]>; dev <[email protected]>; Gavin Hu (Arm Technology 
China) <[email protected]>; Dharmik Thakkar <[email protected]>; Malvika 
Gupta <[email protected]>; Aaron Conole <[email protected]>
Subject: Re: [dpdk-dev] [PATCH v9 2/4] test/rcu_qsbr: add API and functional 
tests

On Wed, May 1, 2019 at 5:55 AM Honnappa Nagarahalli 
<mailto:[email protected]> wrote:
From: Dharmik Thakkar <mailto:[email protected]>

Add API positive/negative test cases, functional tests and
performance tests.

Signed-off-by: Malvika Gupta <mailto:[email protected]>
Signed-off-by: Dharmik Thakkar <mailto:[email protected]>
Signed-off-by: Honnappa Nagarahalli <mailto:[email protected]>
Reviewed-by: Gavin Hu <mailto:[email protected]>
Acked-by: Konstantin Ananyev <mailto:[email protected]>

Did not investigate, but this test always fails on my laptop.
Tried multiple times, I caught one instance when the test was still burning cpu 
after 5 minutes and I killed it.

Usually, it fails like this:

[dmarchan@dmarchan dpdk]$ ./master/app/test -c 0x1f -n 4 --no-huge
EAL: Detected 8 lcore(s)
EAL: Detected 1 NUMA nodes
EAL: Multi-process socket /run/user/114840/dpdk/rte/mp_socket
EAL: Probing VFIO support...
EAL: Started without hugepages support, physical addresses not available
EAL: PCI device 0000:00:1f.6 on NUMA socket -1
EAL:   Invalid NUMA socket, default to 0
EAL:   probe driver: 8086:15d7 net_e1000_em
APP: HPET is not enabled, using TSC as default timer
RTE>>rcu_qsbr_perf_autotest 
Number of cores provided = 4
Perf test with all reader threads registered
--------------------------------------------

Perf Test: 3 Readers/1 Writer('wait' in qsbr_check == true)
Total RCU updates = 945401635
Cycles per 1000 updates: 25157
Total RCU checks = 20000000
Cycles per 1000 checks: 396375

Perf Test: 4 Readers
Total RCU updates = 400000000
Cycles per 1000 updates: 6241

Perf test: 4 Writers ('wait' in qsbr_check == false)
Total RCU checks = 80000000
Cycles per 1000 checks: 21061

Perf test: 1 writer, 4 readers, 1 QSBR variable, 1 QSBR Query, Blocking QSBR 
Check
Reader did not complete #28 =  4097
Test Failed
RTE>>


-- 
David Marchand

Reply via email to