Bug#919628: Apply Julia's LLVM patches

[Sylvestre Ledru]

Hello,

I started the work ( 
https://salsa.debian.org/pkg-llvm-team/llvm-toolchain/commit/408f329cd84ad41cef7fc41ee4ac2b4b4573945f 
) on that but:

* some patches didn't apply

* some are breaking the builds

I will try to finish this week end

Anyway, this should not prevent you to move back to llvm packages as

I took this fix  "Fix a baseline violation on armhf (Closes: #914268)"

Cheers,

S


Le 22/02/2019 à 00:55, Mo Zhou a écrit :
> Hi Sylvestre,
>
> Any chance for getting this into Buster? If there is any, I'd like
> to apply for freeze exception early, especially for the next Julia
> LTS release 1.0.4 (likely to come out before 1st March)
>
> On Fri, Feb 08, 2019 at 08:46:06AM +0100, Sylvestre Ledru wrote:
>> Wahou, better than I was expecting! Many thanks!
>>
>> I will take as much as possible! Thanks
>>
>> S
>>
>>
>> Le 08/02/2019 à 07:47, Mo Zhou a écrit :
>>> Hi Sylvestre,
>>>
>>> Please cherry-pick at least: (8) (12) (13) (14) (15)
>>>
>>> Recommended to include: (1) (2) (4) (5) (11)
>>>
>>> Feel free to ignore: (6) (9) (16) (17)
>>>
>>> I have no idea about: (3) (7) (10)
>>>
>>> https://github.com/JuliaLang/julia/tree/master/deps/patches
>>> I've listed patches for llvm 6.0.1
>>>
>>> ========================================================
>>>
>>>
>>> (1) [unwind] llvm-D27629-AArch64-large_model_6.0.1
>>>
>>>       Fix unwind info relocation with large code model on AArch64
>>>       https://reviews.llvm.org/D27629
>>>
>>> (2) [performance] llvm-D34078-vectorize-fdiv
>>>
>>>       Enable support for floating-point division reductions
>>>       https://reviews.llvm.org/D34078
>>>
>>> (3) [nvptx] llvm-6.0-NVPTX-addrspaces
>>>
>>>       No idea about this patch.
>>>
>>> (4) [performance regression] llvm-D42262-jumpthreading-not-i1
>>>
>>>       For details see Julia commit: e94a1f8b08e0bc3b8093d8f1dc2bf3c8f5d59519
>>>       merged upstream: https://reviews.llvm.org/D42262
>>>
>>> (5) [???] llvm-PPC-addrspaces
>>>
>>>       merged upstream: https://reviews.llvm.org/D43781
>>>
>>> (6) [ignore: mingw] llvm-6.0.0_D27296-libssp
>>>       [ignore: mingw] llvm-6.0-D44650
>>>
>>> (7) [???] llvm-D46460
>>>
>>>       still under review: https://reviews.llvm.org/D46460
>>>
>>> (8) [???] llvm-rL327898
>>>
>>>       https://github.com/JuliaLang/julia/blob/master/deps/patches/llvm-rL327898.patch
>>>       Fixes Julia issues: #27055 #27080 #27032 #27603
>>>
>>> (9) [ignore: compiler complain] llvm-6.0-DISABLE_ABI_CHECKS
>>>
>>> (10) [profiling] llvm-OProfile-line-num
>>>
>>> (11) [profiling] llvm-D44892-Perf-integration
>>>
>>>        merged upstream: https://reviews.llvm.org/D44892
>>>
>>> (12) [bug fix] llvm-D49832-SCEVPred
>>>        [bug fix] llvm-rL323946-LSRTy
>>>
>>>       Add LLVM patches for bugs introducing illegal ptrtoint
>>>       rL323946 [LSR] Don't force bases of foldable formulae to the final type.
>>>       D49832   [SCEV] Don't expand Wrap predicate using inttoptr in ni addrspaces
>>>
>>> (13) llvm-D50010-VNCoercion-ni
>>>
>>>       Fixes julia issue: #28360 (#28362)
>>>       https://reviews.llvm.org/D50010
>>>
>>> (14) llvm-D50167-scev-umin
>>>
>>>       Add LLVM patch to explicitly represent umin in SCEV (#28403)
>>>       Fix mix-type arithmetic detection in umin/max expansion (#28465)
>>>       Fixes #28464
>>>       Fixes #28379
>>>       Fixes #28388
>>>
>>> (15) llvm-rL326967-aligned-load
>>>
>>>       Fixes incorrect codegen: #28726
>>>
>>> (16) [ignore: win64] llvm-D51842-win64-byval-cc
>>>
>>> (17) [...] llvm-D57118-powerpc
>>>
>>>       https://reviews.llvm.org/D57118
>>>