[LLVMdev] 3.4.1 Release Plans

[Tom Stellard]

On Fri, Apr 04, 2014 at 04:29:04PM +0800, Jiangning Liu wrote:
> Hi,
> 
> 
> > > I think the latter option is the right choice. The pain of preserving the
> >
> > ABI belongs with our hard-working branch maintainers (and they always have
> > > the option of rejecting a change because it would break the ABI).
> >
> > Absolutely!
> >
> > --renato
> >
> 
> Sorry, I 'm not sure I'm following you guys correctly. Do you mean we
> should not allow any LLVMWhatever.so ABI broken, so we need to guarantee
> the ordering of those intrinsic enums generated by Intrinsicsxxx.td?
> 

Yes, we must make sure the shared object's ABI is not broken between
3.4.0 and 3.4.1.

> If yes, I don't think it's worthwhile tuning them with whatever tricky
> method to make abi-compliance-checker give 100% backward compatibility, so
> I'd want to give up adding the complete NEON support in 3.4.1 release.
> 
> Therefore, plan B is to port the followings to 3.4.1 release only,
> 
> CLANG:
> 
> 198940 Enable -fuse-init-array for all AArch64 ELF targets by default, not
> just linux.
> 
> LLVM:
> 
> 
> 
> 198937 Make sure -use-init-array has intended effect on all AArch64 ELF
> targets, not just linux.

I wasn't able to merge this commit, because it broke ABI compatibility of
the LLVM library shared object.  This is because it removes the function
llvm::AArch64LinuxTargetObjectFileInitialize ( llvm::MCContext&,
                                               llvm:TargetMachine const&);

Does the clang commit mentioned above require this LLVM change?  I
wasn't sure, so I didn't merge it.  If you can come up with a version of
this patch that doesn't break the ABI, then I can merge it.  Let me know
what you want to do.

> 198941 Silence unused variable warning for non-asserting builds that was
> introduced in r198937.
> 
> 199369 For ARM, fix assertuib failures for some ld/st 3/4 instruction with
> wirteback.
> 
> 201541 Fix a typo about lowering AArch64 va_copy.
> 201841 [AArch64] Add register constraints to avoid generating STLXR and
> STXR with unpredictable behavior.
> 204304 [ARM]Fix an assertion failure in A15SDOptimizer about DPair reg
> class by treating DPair as QPR.
> 

The rest of these commits were merged.

-Tom

> I tried them and the regression tests can all pass.
> 
> -- 
> Thanks,
> -Jiangning

> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits