[Llvm-bgq-discuss] question about flags

Hal Finkel hfinkel at anl.gov
Wed Aug 27 01:09:31 CDT 2014


----- Original Message -----
> From: "John A. Biddiscombe" <biddisco at cscs.ch>
> To: "Hal Finkel" <hfinkel at anl.gov>
> Cc: llvm-bgq-discuss at lists.alcf.anl.gov
> Sent: Tuesday, August 26, 2014 5:54:52 PM
> Subject: Re: [Llvm-bgq-discuss] question about flags
> 
> Hal
> 
> I spent the rest of the day compiling clang and I’m quite surprised
> to say
> that after considerable trial and error, I now have my regex test
> passing
> with a freshly minted clang 3.6!

Ah, great!

> 
> I had to cheat and use your libc++ from the latest bgclang release
> because
> I didn’t realise that the libc++ was not compiled with clang itself,
> but
> I’ve set a compilation going with the libc++ (and libc++ABI or
> whatever it
> is). Once they’re done, then in principle I won’t need the bgclang
> stuff
> and I’ll stop annoying you with these emails. I expect you’ll be sad
> about
> that!

I don't use libc++ABI, and I'm not sure how well it works on PPC/Linux -- I generally build on top of libstdc++ -- so please let me know how it goes.

> 
> [Fear not though as I will still need to get stuff running on the
> compute
> nodes, but at least in principle I can now compile the bgclang
> toolset
> myself so I might even be able to help test your stuff before you
> release
> it].

Sounds good to me :-)

 -Hal

> 
> Thanks once more for all the help
> 
> JB
> 
> 
> 
> 
> On 26/08/14 16:55, "Hal Finkel" <hfinkel at anl.gov> wrote:
> 
> >Hi John,
> >
> >If you're compiling with -mcpu=a2 with the default Linux target then
> >-mno-qpx -mllvm -qpx-stack-unaligned -mllvm
> >-ppc-always-use-base-pointer
> >should really be unnecessary. I recommend that you try with the
> >upstream
> >trunk (or the latest release candidate for 3.5 -- 3.5 is due out
> >next
> >week) and see if that works. You may just be hitting some upstream
> >bug
> >that has been fixed now, and while I'll be putting together a new
> >bgclang
> >release shortly, since you're running on the IONs, you don't need to
> >wait
> >for that.
> >
> > -Hal
> >
> >----- Original Message -----
> >> From: "John A. Biddiscombe" <biddisco at cscs.ch>
> >> To: "Hal Finkel" <hfinkel at anl.gov>
> >> Cc: llvm-bgq-discuss at lists.alcf.anl.gov
> >> Sent: Tuesday, August 26, 2014 8:27:07 AM
> >> Subject: RE: [Llvm-bgq-discuss] question about flags
> >> 
> >> Hal
> >> 
> >> Linking. I see now that it produces the message when linking, so I
> >> will ignore it.
> >> 
> >> Unfortunately, the flags did not help me.
> >> I've got a problem with this test
> >> 
> >>https://raw.githubusercontent.com/biddisco/cpptest/master/regex-recursion
> >>.cxx
> >> 
> >> which hangs after the first boost_check_throw, the first bad regex
> >> test does not throw and hangs indefinitely. Since it's a possibly
> >> stack related problem inside regex, I thought the stack related
> >> flags might help.
> >> 
> >> I'm using
> >> -mcpu=a2 --std=c++11 -stdlib=libc++ -mno-qpx -mllvm
> >> -qpx-stack-unaligned -mllvm -ppc-always-use-base-pointer
> >> 
> >> but no joy yet.
> >> 
> >> Can you suggest any other stack related flags that I might
> >> experiment
> >> with, I'll be working my way through
> >> clang++ --help | grep stack
> >> 
> >> to get ideas. But it does mean a lot of tedious boost
> >> recompilation.
> >> 
> >> gcc4.8.2 compiles the test and runs ok (also using -std=c++11
> >> which I
> >> need for other reasons)
> >> So far I've tried boost_1_55 and 1_56
> >> 
> >> ta
> >> 
> >> JB
> >> 
> >> 
> >> -----Original Message-----
> >> From: Hal Finkel [mailto:hfinkel at anl.gov]
> >> Sent: 26 August 2014 14:17
> >> To: Biddiscombe, John A.
> >> Cc: llvm-bgq-discuss at lists.alcf.anl.gov
> >> Subject: Re: [Llvm-bgq-discuss] question about flags
> >> 
> >> John,
> >> 
> >> Does it say this when compiling, or just when linking? I'd expect
> >> them to be unused when linking.
> >> 
> >>  -Hal
> >> 
> >> ----- Original Message -----
> >> > From: "John A. Biddiscombe" <biddisco at cscs.ch>
> >> > To: llvm-bgq-discuss at lists.alcf.anl.gov
> >> > Sent: Tuesday, August 26, 2014 7:09:42 AM
> >> > Subject: [Llvm-bgq-discuss] question about flags
> >> > 
> >> > 
> >> > 
> >> > 
> >> > 
> >> > When I use
> >> > 
> >> > -mllvm -qpx-stack-unaligned
> >> > 
> >> > -mllvm -ppc-always-use-base-pointer
> >> > 
> >> > 
> >> > 
> >> > the compiler tells me that these flags are unused. I wanted to
> >> > experiment with them to try to work around a boost problem I’ve
> >> > got.
> >> > Are there equivalents that I should be using instead?
> >> > 
> >> > 
> >> > 
> >> > thanks
> >> > 
> >> > 
> >> > 
> >> > JB
> >> > 
> >> > 
> >> > 
> >> > 
> >> > 
> >> > --
> >> > 
> >> > John Biddiscombe, email:biddisco @.at.@ cscs.ch
> >> > 
> >> > http://www.cscs.ch/
> >> > 
> >> > CSCS, Swiss National Supercomputing Centre | Tel: +41 (91)
> >> > 610.82.07
> >> > 
> >> > Via Trevano 131, 6900 Lugano, Switzerland | Fax: +41 (91)
> >> > 610.82.82
> >> > 
> >> > 
> >> > _______________________________________________
> >> > llvm-bgq-discuss mailing list
> >> > llvm-bgq-discuss at lists.alcf.anl.gov
> >> > https://lists.alcf.anl.gov/mailman/listinfo/llvm-bgq-discuss
> >> > 
> >> 
> >> --
> >> Hal Finkel
> >> Assistant Computational Scientist
> >> Leadership Computing Facility
> >> Argonne National Laboratory
> >> 
> >
> >--
> >Hal Finkel
> >Assistant Computational Scientist
> >Leadership Computing Facility
> >Argonne National Laboratory
> 
> 

-- 
Hal Finkel
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory


More information about the llvm-bgq-discuss mailing list