> On Sep 16, 2019, at 7:22 AM, Benoit Ganne (bganne) <[email protected]> wrote:
> 
>> I tried "perf record --call-graph fp" but the results seemed odd (e.g., I
>> see a tcp function symbol deep within a call-branch rooted in the dpdk
>> crypto device enqueue routine for the marvell mvsam cryptodev).
> 
> Just in case, did you try "--call-graph lbr" (on Intel CPU older than Haswell 
> I think) or "--call-graph dwarf"? The frame pointer is usually unreliable 
> with optimized code unfortunately.


Could these odd symbols I'm seeing be the result of link-time optimization?

Here's an example:

           |           --0.00%-- dpdk_update_link_state
           |                     rte_eth_link_get_nowait
           |                     |
           |                     |--0.00%-- mrvl_link_update
           |                     |          pp2_port_get_link_state
           |                     |          __socket
           |                     |          el0_svc_naked
           |                     |          sys_socket
           |                     |          __sock_create
           |                     |          inet_create
           |                     |          tcp_v4_init_sock
           |                     |          tcp_init_sock
           |                     |          tcp_assign_congestion_control

Thanks,
Chris.

> 
> Best
> ben
> 

Attachment: signature.asc
Description: Message signed with OpenPGP

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13990): https://lists.fd.io/g/vpp-dev/message/13990
Mute This Topic: https://lists.fd.io/mt/34128898/21656
Group Owner: [email protected]
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to