Adar Dembo has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/14964 )

Change subject: KUDU-3007. Support building Kudu on aarch64 platform
......................................................................


Patch Set 7:

(6 comments)

I started to review this but stopped; the issues I found suggest that the Kudu 
unit tests can't pass on aarch64. If that's true, it'll be pretty tough to 
review any of this; how do we know that the changes are correct?

http://gerrit.cloudera.org:8080/#/c/14964/7/src/kudu/cfile/binary_plain_block.cc
File src/kudu/cfile/binary_plain_block.cc:

http://gerrit.cloudera.org:8080/#/c/14964/7/src/kudu/cfile/binary_plain_block.cc@202
PS7, Line 202:       #ifndef __aarch64__
             :       p = coding::DecodeGroupVarInt32_SSE(
             :           p, &dst_ptr[0], &dst_ptr[1], &dst_ptr[2], &dst_ptr[3]);
             :       #endif //__aarch64__
This seems wrong; shouldn't we call the SlowButSafe function below? Meaning, 
shouldn't the ifdef be built into the condition on L201 in some way?


http://gerrit.cloudera.org:8080/#/c/14964/7/src/kudu/cfile/bitshuffle_arch_wrapper.cc
File src/kudu/cfile/bitshuffle_arch_wrapper.cc:

http://gerrit.cloudera.org:8080/#/c/14964/7/src/kudu/cfile/bitshuffle_arch_wrapper.cc@60
PS7, Line 60: #if !defined(__APPLE__) && !defined(__aarch64__)
We can't safely call CPU().has_avx2() to determine this on aarch64? Why not?


http://gerrit.cloudera.org:8080/#/c/14964/7/src/kudu/common/key_encoder.h
File src/kudu/common/key_encoder.h:

http://gerrit.cloudera.org:8080/#/c/14964/7/src/kudu/common/key_encoder.h@249
PS7, Line 249:     #ifndef __aarch64__
Shouldn't the aarch64 case return false so we go through the slow loop?


http://gerrit.cloudera.org:8080/#/c/14964/7/src/kudu/gutil/cpu.cc
File src/kudu/gutil/cpu.cc:

http://gerrit.cloudera.org:8080/#/c/14964/7/src/kudu/gutil/cpu.cc@271
PS7, Line 271:   has_broken_neon_ = false;
Isn't this the default value?


http://gerrit.cloudera.org:8080/#/c/14964/7/src/kudu/gutil/linux_syscall_support.h
File src/kudu/gutil/linux_syscall_support.h:

PS7:
What caused the changes in this file?


http://gerrit.cloudera.org:8080/#/c/14964/7/src/kudu/gutil/spinlock.h
File src/kudu/gutil/spinlock.h:

PS7:
I don't really understand the changes here.



--
To view, visit http://gerrit.cloudera.org:8080/14964
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I2953519c5d28de17e6b2bb7094abab0c1cd12c97
Gerrit-Change-Number: 14964
Gerrit-PatchSet: 7
Gerrit-Owner: liusheng <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Reviewer: Tim Armstrong <[email protected]>
Gerrit-Reviewer: liusheng <[email protected]>
Gerrit-Comment-Date: Mon, 06 Jan 2020 06:05:00 +0000
Gerrit-HasComments: Yes

Reply via email to