common/log: Speed improvement for log#19100
Conversation
c0abf5f to
2f31179
Compare
beforeafter |
|
@tone-zhang the "make check" on aarch64 consistently fails recently. do you mind taking a look at your convenience? |
|
@tchaikov My pleasure! I will have a look ASAP. Thanks! |
|
@tchaikov I validated the PR and ran "make check -j16" command in my test bed, it looks everything is OK. |
|
@tone-zhang thanks for verifying this! probably this failure is not very reproducible, but in our jenkins slave, it happens frequently. see |
|
@tchaikov All the above failures are in smoke.sh. I remember I have seen such error before. |
|
Signed-off-by: Adam Kupczyk <akupczyk@redhat.com>
2f31179 to
27bd214
Compare
|
@aclamk i reverted the second change, and rebased against master. could you take another look? |
…usage Signed-off-by: Adam Kupczyk <akupczyk@redhat.com> Signed-off-by: Kefu Chai <kchai@redhat.com>
27bd214 to
b08ae65
Compare
aclamk
left a comment
There was a problem hiding this comment.
Before:
[ RUN ] Log.Speed_gather
[ OK ] Log.Speed_gather (1625 ms)
[ RUN ] Log.Speed_nogather
[ OK ] Log.Speed_nogather (52 ms)
After
[ RUN ] Log.Speed_gather
[ OK ] Log.Speed_gather (1273 ms)
[ RUN ] Log.Speed_nogather
[ OK ] Log.Speed_nogather (45 ms)
Everything looks good.
@aclamk i adapted your PR of #17938 so it works with the rest part of logging system. also did some cleanup.
ThreadLocalPrebufferedStreambuftoCachedPrebufferedStreambuf, shorter name, and it makes the intention more visible.ostreamout ofCachedPrebufferedStreambuf. so it looks more reasonable: it's kind of weird at the first glance that astreambufactually containsostream!CachedPrebufferedStreambufinherit fromstreambufinstead ofbasic_streambuf<>, less typeCachedPrebufferedStreambuf, as they are inherited from its parent class.traits_typeinstead oftraits_ty. the former is ready to use already.