[llvmlinux] "make test" for x86_64 target just hung there, why?

Sedat Dilek sedat.dilek at gmail.com
Sat Sep 12 21:10:17 UTC 2015


On Sat, Sep 12, 2015 at 10:09 PM, Sedat Dilek <sedat.dilek at gmail.com> wrote:
> On Sat, Sep 12, 2015 at 6:50 PM, Daniel Sanders
> <Daniel.Sanders at imgtec.com> wrote:
>> That patch looks right to me. Given that it didn't cause GCC to show the bug too, the problem is unlikely to be the same kind of bug I had with the slab allocator.
>
> OK.
> Thanks for all the feedback.
>
> I have seen that the issue occurs with -O2 (two) and -O3 (three).
> Whereas -O0 (zero) and -O1 (one) and -Oz
> (CONFIG_CC_OPTIMIZE_FOR_SIZE=y) are fine.
>
> I rebuild a last Linux-kernel with CONFIG_CC_OPTIMIZE_FOR_SIZE=y.
>

The objdump for -Oz looked good but -O0 and -01 resulted in a bootable
Linux-kernel with CLANG v3.7.

Cannot say which "inline" compiler-flags are activated by default for >= -O2.

- Sedat -

> - Sedat -
>
>> ________________________________________
>> From: Sedat Dilek [sedat.dilek at gmail.com]
>> Sent: 12 September 2015 15:36
>> To: Daniel Sanders
>> Cc: pageexec at gmail.com; llvmlinux at lists.linuxfoundation.org; David Woodhouse
>> Subject: Re: [llvmlinux] "make test" for x86_64 target just hung there, why?
>>
>> I tried with the attached patch on top of Linux v4.2 + llvmlinux
>> patchset, but this seems not to disable "inline optimization".
>>
>> Re-compiling lib/bitmap with gcc v4.9 let's my kernel boot.
>>
>> Maybe pipacs is right when he says it might be a
>> CONFIG_ARCH_HWEIGHT_CFLAGS problem.
>>
>> For AMD64 and GCC...
>>
>> CONFIG_ARCH_HWEIGHT_CFLAGS="-fcall-saved-rdi -fcall-saved-rsi
>> -fcall-saved-rdx -fcall-saved-rcx -fcall-saved-r8 -fcall-saved-r9
>> -fcall-saved-r10 -fcall-saved-r11"
>>
>> FYI: This is not used with LLVMLinux.
>>
>> Empty head,
>> - Sedat -


More information about the LLVMLinux mailing list