llvm dev - Sep 2014 - [LLVMdev] [cfe-dev] AARCH64 - compiler-rt missing some functions

Hi

Does anyone have a plan of attack for some things which libgcc is 
providing on AARCH64, but compiler-rt isn't yet?

We're hitting some Fortran issues with quad bits missing in glibc as a 
result.

Specifically this list
__addtf3
__eqtf2
__extenddftf2
__extendsftf2
__fixtfdi
__fixtfsi
__floatditf
__floatsitf
__gcc_qadd
__gcc_qdiv
__gcc_qmul
__gcc_qsub
__getf2
__gttf2
__letf2
__lttf2
__subtf3
__trunctfdf2
__trunctfsf2
__unordtf2

Comments/feedback appreciated. I'd like to get this upstream into 
compiler-rt asap.

./C

On Thu, Aug 28, 2014 at 09:48:33AM +0700, "C. Bergström"
wrote:
> __addtf3
Present.
> __eqtf2
comparetf2.c
> __extenddftf2
Present.
> __extendsftf2
Present.
> __fixtfdi
> __fixtfsi
Has a pending review, needs a bit work before commit.
> __floatditf
Present.
> __floatsitf
Present.
> __gcc_qadd
> __gcc_qdiv
> __gcc_qmul
> __gcc_qsub
Those are only used on PPC for the double-double format?
> __getf2
> __gttf2
> __letf2
> __lttf2
comparetf2.c
> __subtf3
Present.
> __trunctfdf2
Present.
> __trunctfsf2
Present.
> __unordtf2
comparetf2.c

Joerg

2nd round of missing stuff in compiler-rt which is just waiting for review
---------------
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20140210/204960.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20140210/204961.html

The code is pretty small - Can someone please review those patches and 
push it. This blocks quad math stuff on aarch64 from using compiler-rt 
instead of gnu libgcc

__fixtfsi
__fixunstfsi
__floatsitf
__floatunsitf