1180370 - Use Clang as default compiler on Linux

Status: RESOLVED DUPLICATE of bug 1253064

(Firefox Build System :: General, defect)

Description

[Godrod]

User Agent: Mozilla/5.0 (Windows NT 6.3; Win64; x64; rv:42.0) Gecko/20100101 Firefox/42.0
Build ID: 20150703030216

Comment 1

[Godrod]

Attachment: Clang Result Octane

Comment 2

[Godrod]

Attachment: GCC Result Octane

Comment 3

[Godrod]

As you can see, Firefox build with clang is 0.95X slower than GCC version(almost inexistant difference betwen those two).
Also clang has better diagnostic than gcc which make that you can fix bug easier. Finaly, clang compiler faster than gcc which take around 57 min to compile against clang which take 38min to compile whether around 2 times faster

Comment 4

[Godrod]

I've just forgot one thing.If we'd switch to clang as default compiler we'd have from 3 to 2 different compiler

Comment 5

[Mike Hommey [:glandium]]

Your benchmarks lack the most interesting information:
- what version of clang?
- what version of gcc?
- what mozconfig did you use?
- is the gcc build one from mozilla continuous integration or one you did yourself?
- if the former, what version? nightly? aurora? beta? release?
- if the latter did you enable PGO/FDO?
- did you enable PGO/FDO with clang?

Comment 6

[Godrod]

GCC 5.1.0
Clang/LLVM 3.6
mozconfig: attached file
version nighlty from mozilla-central
i don't PGO/FDO so i don't belive it was enable. but i can build another one with this enabled with need to

Comment 7

[Godrod]

Attachment: mozconfig

Comment 8

[Godrod]

* juste au cas ou je suis français et comme je n'ai pas un niveau très éleve en anglais...

Comment 9

[Godrod]

This bug have not been updated for almost a month but i have a question. Why still use GCC when we can have a compiler that can do things better: better diagnostic, faster compilation, easier implementation with IDE and also a complete support of the API use with GCC ?

Comment 10

[Gregory Szorc [:gps]]

I didn't know about this bug when I just filed bug 1253064 :/

Normally I'd dupe to the older bug. However, I sent out a mailing list post linking to the new bug. So duping to the new one.

Congratulations on being several months ahead of us on this proposal :)

Also, we're likely not going to switch automation builds from GCC as part of bug 1253064. If Clang pans out, we'll consider it.