Ashwani Raina has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/23064 )

Change subject: [tablet] avoid -Wpsabi warning in ComputeLowerAndUpperBound on 
Rocky ARM
......................................................................


Patch Set 3:

> Hi, could you add some description for the problem instead of
 > calling it. Do I understand that the following happens:
 >
 > In case of older g++, classes with a zero-sized subobject were not
 > treated as HFAs. Newer g++ versions have fixed this issue.
 >
 > This could cause problems if we compiled the function into a
 > library
 > with a newer g++, then tried to use it while compiling with an
 > older
 > compiler (or vice versa). Since the new code would use the
 > floating-point SIMD register, this would cause ABI incompatibility.
 > 
 > This is not the case in KUDU, since it is an internal function, so
 > this is just a warning.
 >
 > A custom structure instead of std::pair<double, double> was already
 > treated as an HFA with older compilers, so the incompatibility
 > disappears.
 >
 > This commit does not change anything with newer compilers, and with
 > older compilers KUDU will probably use a different register to pass
 > the return value (so practically nothing changes).
 >
 > ?

I agree with Zoltan here. The patch changes nothing except for suppressing the 
warning.
Maybe, std::pair is slightly more performant than custom struct but that is not 
a concern here. However, custom struct certainly adds more readability.


--
To view, visit http://gerrit.cloudera.org:8080/23064
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Iab4db585a8eaf152f1fa2884e72e38a4e7c5efa1
Gerrit-Change-Number: 23064
Gerrit-PatchSet: 3
Gerrit-Owner: KeDeng <[email protected]>
Gerrit-Reviewer: Ashwani Raina <[email protected]>
Gerrit-Reviewer: KeDeng <[email protected]>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Reviewer: Zoltan Martonka <[email protected]>
Gerrit-Comment-Date: Wed, 25 Jun 2025 08:13:39 +0000
Gerrit-HasComments: No

Reply via email to